JP2003132060A - Search support apparatus, search support method, and search support program - Google Patents

Abstract

(57) [Problem] To provide a search support device capable of automatically obtaining an output result of a search engine specialized in a specific field without imposing a burden on a user for selecting a search engine. A search input sentence obtained from a client terminal device is parsed and matched with a case sentence registered in advance. These case sentences are associated with a search engine specialized in a specific field, whereby a search input sentence and a search engine specialized in a specific field can be selected. Next, a keyword is extracted from the search input sentence, and if necessary, a predetermined process such as converting an abbreviated name into a formal name is performed to create a search request sentence. Then, a search is requested to the previously selected search engine using the search request sentence, and the search result is presented on the client terminal device.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a search support device, for example, a device that supports a search for a website.

[0002]

2. Description of the Related Art With the rapid spread of the Internet in recent years, it has become possible to easily access a website from a personal computer (hereinafter referred to as a PC) or a mobile phone. On the Internet, the URL (Unif
A meta search engine (hereinafter referred to as a search engine) for searching orm Resource Locators is provided.

A user can use these search engines by accessing a predetermined meta search site. These meta search engines include those that perform a simple search using a keyword input by the user without specifying the field, and those that support searches specialized in various fields such as stock prices and weather. Hereinafter, among the meta search engines and other search engines, those that do not specialize in a specific field and perform general searches are called general-purpose search engines.
Those specialized in a specific field such as weather will be called a dedicated search engine.

When a search is executed by a general-purpose search engine, search results associated with the entered keyword can be obtained regardless of the field. For example, if you search for "stock price of company A" with a general-purpose search engine, news about the stock price of company A, a personal home page about the stock price of company A, etc. will be displayed.
Those related to Company A's stock price will hit regardless of whether or not there is a site. And it is hard to come up with a chart that shows fluctuations in stock prices.

On the other hand, in the case of a dedicated search engine, information hits within the range preset by the search engine. For example, if you search for "stock price of company A" with a dedicated search engine specialized for the stock market, you will get information specific to company A's stock, such as the stock chart of A company, daily trading volume, or stock price fluctuations. To be

[0006]

As described above, a user who wants to efficiently obtain information specialized in a specific field needs to use a dedicated search engine for the dedicated field.
However, since the user manually selects this dedicated search engine, it is a heavy burden especially for Web beginners.

[0007] Also, the dedicated search page for stock prices, weather, etc. is easy to use, and while the desired information can be accurately obtained, the user may forget that the search page is provided on the Internet, or may not know it. And I often look it up with a general-purpose search engine. In particular, since there are countless search fields, this problem becomes more serious as the number of fields supported by dedicated search engines increases.

Therefore, an object of the present invention is to provide a search support device, a search support method, and a search support program that can automatically obtain the output result of a dedicated search engine without burdening the user with selecting a search engine. Is to provide.

[0009]

In order to achieve the above-mentioned object, the present invention provides a search input sentence acquisition means for obtaining a search input sentence, claim 1, and a case sentence prepared beforehand. Matching means for matching the search case, a search engine specifying means for specifying a search engine associated with the matched case sentence, an extracting means for extracting a keyword from the search input sentence, and the specifying using the extracted keyword Request a search from the search engine
A search support device comprising: a search requesting unit that acquires a search result from the search engine; and a presenting unit that presents the acquired search result. The matching means according to claim 2 is characterized in that the similarity between the search input sentence and the case sentence is calculated, and the case sentence having the similarity higher than a predetermined threshold value is matched with the search input sentence. A search support device according to claim 1 is provided. In the invention according to claim 3, further comprising a search formula generating means for generating a search formula by using the extracted keyword, and when there is no case sentence matching with the search input sentence by the matching means, the case Search request means
The search support apparatus according to claim 1 or 2, wherein a search is requested to a predetermined search engine using the generated search formula. In the invention described in claim 4,
The presenting means synthesizes and presents search results of the plurality of search engines when there are a plurality of search engines specified by the search engine specifying means for the case sentence. A search support device according to any one of claims 2 and 3 is provided. The invention according to claim 5 further comprises an ordering unit for ordering the search results acquired from the plurality of search engines, and the presenting unit presents the search results in the order given by the ordering unit. A search support device according to claim 4 is provided. According to a sixth aspect of the present invention, in a computer including a search input sentence acquisition unit, a matching unit, a search engine specifying unit, an extraction unit, a search request unit, and a presentation unit, the search input sentence acquisition unit is provided. A search input sentence acquisition step of acquiring a search input sentence by means, a matching step of matching the acquired search input sentence with a case sentence prepared in advance by the matching means, and the search engine specifying means, A search engine specifying step of specifying a search engine associated with the matched case sentence; an extracting step of extracting a keyword from the search input statement by the extracting means; and a step of using the extracted keyword by the search requesting means A search request page that requests a search from the specified search engine and obtains search results from the search engine. And-up, in the presentation unit, to provide a search supporting method characterized in that it consists of a presentation step of presenting the obtained search result. In the invention according to claim 7, a search input sentence acquisition function for obtaining a search input sentence, a matching function for matching the obtained search input sentence and a prepared case sentence, and the matched case sentence are used. A search engine specifying function for specifying a search engine by using a search function, an extraction function for extracting a keyword from the search input sentence, a search request for the specified search engine using the extracted keyword, and a search result from the search engine. A search support program for realizing a search request function to be acquired and a presentation function to present the acquired search result, or a computer-readable storage medium storing the search support program.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of the present invention will be described in detail below with reference to FIGS. 1 to 11. [Outline of Embodiment] The search support server according to the present embodiment is
The dedicated search engine is automatically selected according to the search input sentence from the user, and the search is performed by the selected search engine. The dedicated search engine is selected by matching a case input sentence and a search input sentence that are associated with the dedicated search engine in advance. As a matching method, any of various existing methods may be used. In the present embodiment, as an example, sentence structures peculiar to a predetermined field are prepared in advance as case sentences, and matching is performed using these and the similarity of the sentence structure of the search input sentence. On the other hand, an optimum query sentence (query) is created by appropriately processing by extracting a keyword from the search input sentence. Then, a search is requested by the created inquiry sentence to the selected dedicated search engine, and the search result is presented to the user.

For example, if the user inputs "Watanabe Electric Co., Ltd.", it matches the case sentence such as "<company><share or achievement>". This example sentence is assumed to be associated with a dedicated stock price search engine. On the other hand, the keyword “Watanabe Denki” is extracted from the search input sentence and converted to the official name “Watanabe Denki Co., Ltd.” if necessary. Then, a search is requested to a stock price specialized search engine using "Watanabe Electric Co., Ltd." as a keyword. Then, the search result sent from the search engine is presented to the user.

[Details of Embodiment] FIG. 1 is a diagram showing an example of a network configuration using the search support server 1 of the present embodiment. The search support server 1 has client terminal devices 2, 2, 2, ...
, And the search servers 3, 3, 3, ... Note that, hereinafter, the client terminal devices 2, 2, 2, ..., The search servers 3, 3, 3, ... Are simply referred to as the client terminal device 2 and the search server 3 unless otherwise distinguished. These devices are, for example, TC
P / IP (Transmission Control)
Protocol / Internet Protocol
They can communicate with each other using protocols such as l). The search support server 1 and the search server 3 can be accessed from the client terminal device 2 at a predetermined URL.

The client terminal device 2 is, for example, P
C, mobile phone, internet-accessible word processor, digital television, and other devices. The client terminal device 2 includes an input device such as a keyboard, a mouse, and a remote controller, an output device such as a display and a speaker, a CPU that performs various arithmetic processes, a storage device that stores a browser and other programs, the Internet 5, and a client terminal. A communication control device for controlling communication with the device 2 is provided.

The client terminal device 2 can access the search support server 1, the search server 3, and other server devices connected to the Internet 5 to send and receive data. For example, when a URL is input from the input device, the HTML (Hypertext Markup Lan) is input from the website specified by the URL.
You can download the files in the guage) format. Then, the screen defined by the HTML file is displayed on the display. These communication functions and display functions are exhibited by the CPU operating according to a communication program or a browser.

The search server 3 is a server device that provides a search engine and provides a search service to the search support server 1 and the client terminal device 2. There are various search engines provided by the search server 3 from a general-purpose search engine that does not specify a search field to a dedicated search engine that specializes in a specific field such as stocks, weather, restaurants, hospitals, movie theaters, and department stores. Note that one search server 3 can also provide a plurality of search engines.

The search server 3 is a client terminal device 2
When a search request is received from the search support server 1 or the search support server 1, a web site corresponding to the search formula is searched, and a list of URLs and abstracts of these web sites is returned according to the search order calculated by a predetermined method. The search engine provided by the search server 3 is accessible by a predetermined URL. Therefore, the search engine to be used can be selected by the URL.

The search support server 1 is a server device that automatically selects a dedicated search engine in accordance with a search input sentence received from the client terminal device 2 and substitutes a search request for these search engines. If you cannot select the optimum dedicated search engine from the search input sentence,
Make a search request to a general-purpose search engine. The search engine selected for one search input sentence may be singular or plural.

When the search support server 1 is accessed by the client terminal device 2, the search support server 1 transmits the search input screen data to the client terminal device 2 and causes the display of the client terminal device 2 to display the search input screen. The search support server 1 receives the search input text input from the search input screen, selects the search server 3 and creates a search query text, and uses the search query text created in the selected search server 3 and the like. Request a search. And the search server 3
Search results returned from the client terminal device 2
Send to.

The screen data transmitted by the search support server 1 to the client terminal device 2 is defined in, for example, HTML or another language. Then, since the search support server 1 selects the optimum search engine for the search,
The user can input a search input sentence in the same manner as a general-purpose search engine. When a plurality of search engines are searched, the search results are ranked according to a predetermined procedure, and a screen as if one search engine was searched is presented.

FIG. 2 is a diagram showing an example of a logical configuration of the search support server 1. These configurations are generated by the CPU included in the search support server 1 according to a predetermined program. This logical configuration has a search input sentence acquisition unit 1 that transmits search input screen data to the client terminal device 2 and receives a search input sentence input by the user.
1. Case sentence matching 12 for matching a search input sentence and a case sentence to select a search site, a conversion unit 13 for extracting a keyword from the search input sentence and performing a predetermined process to generate a search command, and a search command A parameter conversion rule database 14 that stores the parameter conversion rules at the time, a search request unit 15 that requests a search from a selected search engine and receives the search results, and generates search result screen data from the received search results. It is composed of an output unit 16 for transmitting to the client terminal device 2. The functions of these components will be described below.

The search input sentence acquisition unit 11 sends an HTML file defining a search input screen to the accessed client terminal device 2. The search input screen is provided with an input field for inputting a search input sentence. When the search input text is input in the input field and the search button is clicked, the search input text is transmitted from the client terminal device 2 to the search support server 1. The search input sentence acquisition unit 11 acquires a search input sentence from the client terminal device 2 and stores it in a RAM (Random Access).
It is stored in a storage device such as Memory. Thus, the search input sentence acquisition unit 11 constitutes a search input sentence acquisition means.

The case sentence matching unit 12 is a matching unit that reads a search input sentence from the storage device, parses the sentence, and calculates the degree of similarity with a predefined case sentence. The example sentence is, for example, “<company name><stock price or
It is prepared as a sentence structure such as "Achievements>" or "Buy <products>". Then, these case sentences are stored in advance in the storage device as a database.

The case sentence matching unit 12 parses the search input sentence to obtain a sentence structure, and calculates the similarity between this sentence structure and the case sentence (defined as the sentence structure). Then, the case sentence whose similarity is equal to or higher than a certain threshold is specified. As a matching method, for example, unpublished Japanese Patent Application 2
The natural sentence matching device described in 001-95891 can be used. In this matching method, the search input sentence is decomposed into parts of speech, and part-of-speech information (nouns, verbs, etc.), semantic concept information (animals, documents, etc.), dependency information (surface layer, deep layer, relational) The sentence structure of the search input sentence is obtained by adding the part of speech and meaning of the source and the contact. Then, the similarity between the search input sentence and the case sentence is obtained by using the commonalities between the sentence structure of the search input sentence and the sentence structure of the case sentence, and the size and similarity of the part-of-speech dependency tree structure. The similarity is represented by, for example, a decimal number from 0 to 1, and the closer it is to 1, the larger the similarity is, and the threshold is, for example, 0.8.
And The outline of this matching method will be described later in detail.

Each case sentence is associated with an abstract command composed of a combination of information specifying a dedicated search engine as a search destination and a search word. The abstract command is a template for generating a search command used when the search support server 1 requests a search from a dedicated search engine. Then, a search command is generated from the abstract command and the search word extracted from the search input sentence. The abstract commands are provided in association with various dedicated search engines such as a stock search engine and a weather search engine. The search command is later converted into an HTML document by the conversion unit 13,
Used to request a search from a search engine.

For example, the search input sentence is "Watanabe stock price".
If it is, the sentence structure “stock price of <company name>” is generated by the syntax analysis. As a result of the similarity calculation, this matches the example sentence “stock price of <company name>”, for example. Then, the abstract command “[search destination] stock price search [search word] <company name>” associated with this example sentence is searched. And
By substituting the company name for <company name>, a search command “[search destination] stock price search [search word] <company name> Watanabe Electric” is generated.

In addition to the above method, the matching between the search input sentence and the case sentence is performed by, for example, IIF (Intelligent Interface) disclosed in Japanese Patent Laid-Open No. 11-31149.
There are various publicly known methods such as a method and a method of matching with a keyword. The matching is not limited to the above, and any one may be used.

When there is a case sentence whose similarity is equal to or higher than a certain threshold value, the conversion unit 13 extracts a keyword from the search input sentence using the position in the sentence structure of each phrase obtained by the syntax analysis as an index. . In this way, the conversion unit 13 constitutes an extraction unit that extracts a keyword. Then, the conversion unit 13
When the extracted keyword matches the conversion rule registered in the parameter conversion rule database 14, the conversion is performed according to the conversion rule. If there is no case sentence that matches the search input sentence, the search input sentence is decomposed into words and operators, and a search expression is created while referring to the conversion rules.

The parameter conversion rule database 14 stores conversion rules for converting the keywords extracted by the conversion unit 13. There are the following four conversion rules, for example.

(1) Rule for Converting Common Name to Official Name This conversion rule is to convert an abbreviation or a common name into a formal name. The parameter conversion rule database 14 stores a set of various abbreviations / common names and official names. Then, the conversion unit 13 searches the parameter conversion rule database 14 for the keyword set in the [search word] of the search command, and if there is an official name,
Convert keywords to official names. For example, abbreviations such as "Watanabe Electric", "Tokyo Metropolitan Government", and "National Diet" are converted into formal names such as "Watanabe Electric Co., Ltd.", "Tokyo Metropolitan Government Office", and "National Diet Building".

(2) Conversion of search command to HTML document Rule search command is converted to an HTML document, and the search result by an appropriate search engine is automatically obtained by the client terminal device 2
Is to be displayed in. For example, the URL of the search site corresponding to the [search destination] is embedded in the parameter conversion rule database 14, the keyword specified in the [search word] can be inserted, and the HTML has a function of displaying the search result. The document is stored as a template. In the parameter conversion rule database 14, both are associated with each other so that the HTML document specified by [search word] is specified, and constitutes search engine specifying means.

Search sites corresponding to [search destination] include search sites that store general-purpose search engines, as well as search sites that store dedicated search engines for various fields such as stock price search, product search, building search, etc. . These search sites are selected by the URL in the HTML document. The conversion unit 13 searches for an HTML document corresponding to [search destination],
Insert the search keyword into this.

FIG. 3 shows an example of the created HTML document. The HTML document 41 is obtained by converting the search command “[search destination] stock price search [search word] <company name> Watanabe Electric Co., Ltd.” into an HTML document according to the conversion rule.
The example in the figure shows a case where the search is requested to two stock price specialized search sites. The HTML document 42 requests “Abc stock price search” and the HTML document 43 requests “Info”.
Request a stock price search ".

In the HTML document 42, the URL 44 of "Abc stock price search", the parameter 45 for starting the search engine, and the keyword 46 are set in advance. The URL 44 and the parameter 45 are preset, and the keyword 46 is generated from the [search word] of the search command.

Similarly, the HTML document 48 is an HTML document generated for the search command "[search destination] product search [search word] thermos". "Rakuraku merchandise search" and "sale merchandise search", which are specialized product search sites, are designated as search destinations.

(3) Conversion Rule from Search Input Sentence to Search Expression Using Dependency This conversion rule is used when there is no case sentence that matches the search input sentence, that is, when the similarity of the search input sentence is any case. It is used when converting the dependent sentence structure of the search input sentence into a search expression when the threshold value is not reached for the sentence as well. Figure 4
It is a figure showing an example of a conversion rule. There are three types of conversion patterns, in which the deep case of the origin is disjunctive parallel, the place or the target, and the adnominal modification, and the search expression is defined for each narrowing level.

As the narrowing-down level, the narrowing-down level 3 is set as a standard by default, and the user can change it from the search input screen as needed. For example, in the case of disjunctive parallel, the following is performed. If the input search sentence is "a place where you can see baseball or soccer," the reception is as follows. "Baseball" (deep personality: target) →
"See", "Soccer" (deep case: target) → "See" In this way, "baseball" and "soccer" are in a disjunctive parallel relationship.

When the narrowing level is 1, 2 or 3, the retrieval formula is "baseball" or "soccer", and the retrieval word containing at least one of "baseball" and "soccer" is hit. When the refinement level is 4 or 5, the search formula is "baseball" and "soccer", and the search terms that include both "baseball" and "soccer" are hit, and the refinement level is 1, 2 or 3. It is possible to obtain more narrowed search results than in the case.

In the case of a place or an object, it is as follows. It is assumed that the search input sentence is “Hiroshima department store”.
In this case, the contact is "Hiroshima" (deep personality: place or target) →
It is a "department store". In the case of narrowing down levels 1 and 3, the search formula is “Hiroshima” or “department store”. In the case of the narrowing down level 2, it is “Hiroshima”. And when the narrowing level is 4 or 5, "Hiroshima" and "Department store"
Becomes

In the case of modification with a continuum, it is as follows. It is assumed that the search input sentence is "dancing store". In this case, the relationship is "dance" (deep case: adornment modification) → "store". In the case of the narrowing levels 1 and 2, the search formula is “danceable” or “store”. When the narrowing level is 3 or 4, it is "danceable". And when the narrowing level is 5, it becomes "danceable" and "store". In this way, the conversion unit 13 also constitutes a search formula generating means for generating a search formula using the keyword of the search input sentence.

(4) Word Expansion Interpretation Rule This conversion rule is used to expand the word words of the input search sentence and increase the number of cases and pages that are hit. FIG. 5 is a diagram for explaining a word expansion interpretation rule. The expansion range of words can be set by the user according to the narrowing level. The default is the narrowing level 3.

An example of the word "store" will be described with reference to FIG. When the narrowing level is 1 or 2, the word "store" is "store", "store", "store",
"Shop", "restaurant", "department store", "stall", "convenience store" ... When the narrowing down level is 3, narrow the expansion range,
"Store", "Store", "Shop", "Restaurant"
Becomes When the narrowing level is 2, the expansion range is further narrowed to become “store”, “store”, “store”, and “shop”, and when the narrowing level is 5, “store” is set.

In addition to this, for example, the word "vegetable" means "carrot", "potato", "onion", "burdock",
Expanded interpretation of "radish", "spinach", "cabbage", ... These expanded interpretation rules are set for various words and are stored in the parameter conversion rule database 14.

Next, returning to FIG. 2, the search request unit 15 will be described. The search request unit 15 requests the search server 3 to search using the HTML document generated by the conversion unit 13,
Get the search result page. If the search input sentence matches the case sentence, a search request is issued to the dedicated search engine associated with the case sentence. If the search input sentence does not match the case sentence, it is stored in advance in a storage medium or the like and registered. Request a search from a generic search engine. Thus, the search request unit 15 constitutes a search request means.

The output unit 16 creates search result screen data for displaying the search results acquired by the search request unit 15 on the client terminal device 2. The created search result screen data is transmitted to the client terminal device 2 via the Internet 5. The output unit 16 creates a search result screen using the search result while referring to the rules stored in the output screen rule database 17. This screen is transmitted from the search support server 1 to the client terminal device 2 and presented on the client terminal device 2. In this way, the output unit 16 constitutes a presentation unit.

The output screen rule database 17 stores search result page interpretation rules and URL ranking rules. The search result page interpretation rule is a rule used to extract a URL or a page summary from a search result page obtained from a search engine of a search request destination. Figure 6
Shows an example of the search result interpretation rule.

Is a rule for extracting the search result information from the HTML file obtained as a result of the search on the building-specific search site A. This rule is matched with the HTML file of the actual search result. "Number" in the rule
Indicates that any number can be matched at that position. Further, "character string: URL" indicates that an arbitrary character string can be collated at that position and the collated character string is extracted as a URL. The HTML file of the actual search result and the search result information extraction rule are collated by the above-described method, and if the collation is successful, the link destination URL 91, the page title 92, and the page summary 93 are read out. These pieces of information are used for ranking search results. Relates to a search site B specialized in buildings. In this way, rules can be set for each search site.

The URL ranking rule is used to combine the results obtained from a plurality of search engines and display the ranked results as if they were obtained by a single search engine. Equation 1 in FIG. 7 shows a point calculation formula for each URL. The meaning of each symbol in the formula is as follows.

U: a mark for uniquely identifying a URL, i: a mark for uniquely identifying a case sentence matched with a search input sentence, j: a search registered for a case sentence Mark for uniquely identifying the engine, k: Mark for uniquely identifying the rank of the search result for each search engine, Pu: Point for each URL, Mj: Matching score for the search input sentence of the case sentence, Wij: The weight of the search engine registered for the case sentence (Specify whether to set the initial value uniformly or set individual numerical values according to the answer setting of the case sentence. After that, modify by the method described later. Fijku: 1 if the specified URL appears in the specified rank in the result of the specified search engine searched from the specified case sentence, 0 otherwise.

By defining the points for each URL as described above, URLs that appear in the higher ranks in the results of a plurality of search engines can obtain a high point. By displaying URLs with the highest points in order, it is not necessary to browse the results of many search engines,
You can see the ranking results combined in order.

The URL of the ranking result is "http /
/ Metasearch engine URL? Url = identifier of the hit case sentence & j = identifier of the search engine used to pull the URL & k = identification of the rank in the search engine result &
"l = identifier of rank in synthesized result" or another format containing equivalent information, and this URL
The URL selected by the user is notified to the search engine side by transferring the information to the actual result URL from the search engine side.

For each URL selected by the user, Mi and Wij are modified by the equations 2 and 3 in FIG. However, Mi is corrected when it can be corrected. In the formula, a and b are constants, and l is the rank in the result of combining the URLs selected by the user.
Expressions 2 and 3 roughly increase the matching score between the search input sentence and the case sentence, which contributes to the point of the URL selected by the user, and the weight of the search engine for the case sentence, and the rank in the result synthesized from them. It is modified to reduce the weight of case sentences and search engines that contributed to the points of higher URLs. In this way, the output unit 16 can also rank and combine the search results of the plurality of search engines, and thereby generate the search result screen. In this way, the output unit 1
Reference numeral 6 also serves as an ordering means for ordering the search results.

FIG. 8 is a diagram for explaining the contents of the operation performed by the search support server 1 when the search input sentence and the case sentence match. The search input sentence acquisition unit 11 causes the search input sentence 6 which is “Watanabe Electric stock price” from the client terminal device 2.
It is assumed that 1 is acquired. The case matching unit 12 parses the search input sentence 61 and then matches the case sentence with the case sentence 62 of “stock price of <company>”.
In the example sentence 62, “[search destination] stock price [search word] <company>”
Is associated with the abstract command 63. As described above, the case sentence matching unit 12 includes a case sentence and a set of abstract commands associated with the case sentence. Further, like the search engine 67 in the case sentence 62, a predetermined dedicated search engine is associated with each case sentence.
This association is based on the parameter conversion rule database 1
HTM with embedded search destination URL stored in 4
This is done by associating the case sentence with the L document.

After identifying the abstract command 63, the case sentence matching unit 12 substitutes the keyword “Watanabe Denki” extracted from the search input sentence 61 into the abstract command as a search parameter to generate the search command 64. Next, the conversion unit 13 checks whether or not the search command 64 corresponds to the parameter conversion rule, and if it does, converts it. In the case of the search command 64, the common name “Watanabe Electric” is converted into the official name “Watanabe Electric Co., Ltd.” to generate the converted search command 65, and then the HTML document 6
6 is generated. In the example of FIG. 8, one search input sentence matches one case sentence, but the present invention is not limited to this, and one search input sentence is configured to match a plurality of case sentences. You can also In this case, a search request is made to each of the dedicated search engines associated with the plurality of case sentences.

FIG. 9 is a block diagram showing the configuration of the search support server 1. The search server 1 includes a control unit 71, an input device 72, an output device 73, a communication control device 74, an input / output interface 81, a storage device 82, and the like on the bus line 7.
It is connected through 5.

The control unit 26 includes a CPU (Central).
Processing Unit), ROM (Read
It is composed of an only memory), a RAM, and the like, and provides a search support service in response to a search request from a user or controls the entire search support server 1 according to a search support program or the like.

The ROM is a read-only memory (read-only memory) that stores various programs, data, parameters and the like for the CPU to perform various calculations and controls. The RAM is a random access memory used by the CPU as a working memory. CPU is RA
Programs and data can be written in and deleted from M. In the present embodiment, the RAM can secure an area for storing a search input sentence, parsing a sentence, matching a case sentence, converting a parameter, requesting a search from a search engine, and the like.

The input device 72 is composed of, for example, a keyboard and a mouse. The keyboard is a device for inputting information such as letters and numbers to the search support server 1. The keyboard is composed of keys for inputting kana and English characters, ten keys for inputting numbers, various function keys, cursor keys and other keys.

The keyboard is used, for example, when the administrator of the search support server 1 inputs a login ID or password for logging in to the search support server 1, or when operating the search support server 1 for maintenance or the like. .

The mouse is a pointing device.
GUI (Graphical User Interface)
ace) or the like to operate the search support server 1, it is possible to input predetermined information by clicking a button or icon displayed on the display device with a mouse.

The output device 73 is composed of, for example, a display device and a printing device. The display device is a device for presenting information on a screen, for example, a CRT (Cat).
A hode ray tube) display, a liquid crystal display, a plasma display, and the like. The display device can display a keyboard or mouse input result, log data recording user access, and other screens.

The printing device is a device for printing data and the like in the form of text on a printing medium such as paper. The printing device is
For example, it is configured by various printer devices such as an inkjet printer, a laser printer, a thermal transfer printer, and a dot printer.

The communication control device 74 is a device for connecting the search support server 1 to the Internet 5, and is composed of a modem, a terminal adapter and other devices. The communication control device 74 is controlled by the CPU,
For example, it communicates with the client terminal device 2 and the search server 3 according to a predetermined protocol such as TCP / IP.

The communication control device 74 can receive the search input sentence and send the search result to the client terminal device 2, and can request the search server 3 and receive the search result. Further, the communication control device 74 does not directly connect to the Internet, but uses a LAN (Lo).
It can also be configured to connect to the Internet 5 via a Cal Area Network or the like.

The storage device 82 is composed of a readable / writable storage medium and a drive device for reading / writing programs and data from / to the storage medium. A hard disk is mainly used as the storage medium, but other storage media such as a magneto-optical disk, a magnetic disk, and a semiconductor memory can be used instead.

The storage device 82 stores a search support program and O
S (Operating System), a program storage unit 83 that stores various programs such as a communication control program, a parameter conversion rule database 14, a search result analysis rule 17, and a log that stores communication records with the client terminal device 2 and the search server 3. The data storage unit 81 stores files and other databases.

The search support program can be loaded into the CPU and executed to cause the search support server 1 to exert the search support function. When the search support program is executed, the CPU implements the case sentence matching unit 12, the conversion unit 13, the search request unit 15, and the output unit 16 illustrated in FIG. 2 by software.

The storage medium drive device 81 is a drive device for driving a removable storage medium to read and write data. Examples of removable storage media include magneto-optical disks, magnetic disks, magnetic tapes, semiconductor memories,
There is a paper tape punched with data, a CD-ROM, etc. Note that the CD-ROM can only be read.
By driving the storage medium by the storage medium drive device 81, the search support server 1 can upgrade, for example, the search program or the OS, or output the log file to the storage medium.

The input / output interface 44 is, for example,
It is composed of a serial interface and other standard interfaces. By connecting an external device corresponding to the input / output interface 44 to the interface, the function of the search support server 1 can be expanded. Examples of such external devices include a storage device such as a hard disk, a speaker, and a microphone.

FIG. 10 is a flow chart for explaining the operation of the search support server 1. First, the CPU of the search support server 1 receives from the client terminal device 2 via the Internet 5 and stores it in the RAM or the like (step 10). Next, the CPU parses the search input sentence stored in the RAM and specifies the sentence structure of the search input sentence (step 13).

Next, the CPU matches whether or not the sentence structure of the syntactically analyzed search input sentence matches the registered case sentence pattern (step 15). Next CPU
Determines whether there is a case sentence that matches the search input sentence (step 20). When there is a case sentence that matches the search input sentence (step 20; Y), the CPU acquires the abstract command associated with the case sentence (step 25).

Next, the CPU extracts words (keywords) of the defined structure portion in the case sentence pattern from the search input sentence (step 30). Next, the CPU inserts the extracted keyword into the abstract command to generate a search command (step 35). Next, the CPU converts the extracted words and phrases in accordance with the conversion rules defined for each case sentence (step 40).

Next, the CPU sends search request data to the search engine registered for the case sentence using the word / phrase converted in step 40 as a parameter to request a search (step 45). Next, the CPU receives the search result from the search engine. Then, this is analyzed by the result page analysis rule to generate the search result display screen,
The search result is transmitted to the client terminal device 2 and presented to the user (step 50).

When the CPU determines in step 20 that there is no case sentence that matches the search input sentence (step 20; N), the CPU decomposes the structure of the search input sentence into words and operators and extracts the search expression. A search is requested to those created and registered as a standard search engine (step 55).

(Example 1) It is assumed that the client terminal device 2 inputs "I want to know the stock price of Nabeden". After the search input sentence acquisition unit 11 acquires the search input sentence, the search support server 1 verifies whether the case sentence matching unit 12 matches the registered case sentence. At this time, if a case sentence “stock price of <company>” is registered in the case sentence matching unit 12, the search input sentence can be hit to this case sentence. Then, the case sentence matching unit 12 searches for the abstract command “[search destination] stock price search [search word] <company name>” and inserts the company name, and a specific search command “[search target] stock price search [search Word] <company name>
"Pot pan" is generated.

Then, in the conversion unit 13, "Nabeden" is changed to "Watanabe Electric Co., Ltd." according to the conversion rule from the common name to the official name.
, And the search command changes to “[search destination] stock price search [search word] <company name> Watanabe Electric Co., Ltd.”. Further, the conversion unit 13 converts the search command into an HTML document to convert it into an HTML document, and automatically presents a search result in an appropriate stock price search engine to the user.

(Embodiment 2) It is assumed that a search input sentence "I want a pet robot" is acquired from the client terminal device 2. Then, the case matching unit 12 can make the search input sentence hit the case sentence “I want <noun>”. Furthermore, the case sentence matching unit 12 uses a plurality of abstract commands “[search destination] product search [search word] <noun>”, “[search target] auction search [search word] <noun, which are associated with this case sentence. ＞ ”,“ [Search target] Sweepstakes search [Search word] <Noun> ”, specific search commands“ [Search target] Product search [Search word] <Noun> Pet robot ”,“ [Search target] Auction search "[Search word] <noun> pet robot", "[search destination] prize search [search word] <noun> pet robot" are generated. afterwards,
Similar to the first embodiment, the search command is an actual HTML.
It is converted into a document, and the search result in an appropriate search engine is automatically presented to the user.

(Third Embodiment) It is assumed that a search input sentence "I want to get married" is acquired from the client terminal device 2. Then, the case sentence matching unit 12 can hit the case sentence "I want to get married". At this time, since "married" is a different sentence mode (desire and past), even if there is a case sentence "married" (past), the case sentence matching unit 12 distinguishes it by syntax analysis. It is possible. Then, the associated abstract command “[search destination] building search [search word] <noun> wedding hall” is extracted.

After that, the search command is converted into an actual HTML document in the same manner as in the first embodiment, and the search result in an appropriate search engine is automatically presented to the user. After that, the search results are summarized in the format of "search engine: rank: URL" by the search result page interpretation rule. Then, according to the URL ranking rule, the results of multiple search engines are combined and presented to the user.
Also, the URL selected by the user is learned, and the weight of the search engine or the matching rule is corrected.

(Embodiment 4) From the client terminal device 2, "a restaurant where you can eat French or Italian food in Tokushima"
It is assumed that there is no matching case sentence as a result of acquiring the search input sentence, and matching by the case sentence matching unit 12. The dependency of the search input sentence is analyzed as follows.

"Tokushima" (deep case: location) → "eat", "French cuisine" (deep case: target) → "eat", "Italian food" (deep case: target) → "eat""Eatable" (deep case: adornment modification) → "store",
In addition, "French cuisine" and "Italian cuisine" are in a disjunctive parallel relationship.

A search expression "Tokushima and (French food or Italian food) and shop" is generated by this modification and the conversion rule from the search input sentence using the modification to the search expression. After that, if there is the above-mentioned parameter (word expansion) conversion rule, by expanding and interpreting the word according to it, the search expression becomes "Tokushima and (French cuisine o
r Italian food) and (store or store or store or shop or restaurant) ". This search formula is used as a query to request a search from a search engine registered in a storage medium as a standard Web search engine, and the search result is presented to the user.

Further, here, when it is desired to narrow down the number of searches, or conversely when there is a request from the user to increase the number of searches, the conversion rule is corrected by adjusting the narrowing down level, and the search query is facilitated. Can be modified to If you want to increase the number of searches, set the refinement level to 1 and the query will be "Tokushima and
(French food or Italian food) and (store or store or store or shop or restaurant or ...) "
Therefore, more search results can be expected.
When the number of searches is to be narrowed down, the query becomes “Tokushima and (French or Italian) and store” by setting the narrowing level to 5, and the query can be expected to have more narrowed down search results.

FIG. 11 shows an example of the search result screen 100 displayed on the display device of the client terminal device 2. The search text input field 101 is a field in which the user inputs a search text. When the search button 102 is clicked by a mouse operation or the like, the search support server 1 performs the search support operation. The search engine column 103 displays a list of search engines used for the search. In the search result display field 104, URLs that appear in each search engine are displayed in descending order of points.

The following modifications are also possible. In the search engine display field 103, dedicated search engines suitable for searching a search input sentence are displayed in order from the most suitable one. Then, in the search result display field 104, the result of the dedicated search engine most suitable for the search of the search input sentence is displayed, and when the user wants to use another dedicated search engine, it can be selected in the search engine field 103. .

Next, details of the matching method performed by the case sentence matching unit 12 whose outline has been described above will be described. FIG. 12 is a diagram showing the configuration of each component and dictionaries of the case sentence matching unit 12. These components are realized as software by executing the search support program stored in the program storage unit 83 (FIG. 9) by the CPU of the control unit 71. The dictionaries are stored in the data storage unit 84.

An outline of processing contents of each component of the case sentence matching unit 12 will be described below. First,
The search input sentence input by the user from the keyboard or the like is input to the morphological analysis component 117.
The search input sentence may be composed of a natural sentence such as “I want to go to an amusement park”.

The morphological analysis component 117 performs morphological analysis of the input natural sentence and outputs the result to the bunsetsu analysis component 118 as a morphological list.
A morpheme is a phrase that is more detailed than a bunsetsu and is divided into levels of independent words and adjuncts. The phrase analysis component 118 creates a phrase list from the morpheme list. To create a bunsetsu, the bunsetsu basically consists of independent words and adjuncts in the morpheme list. Later, information such as specific numerical values, person names, and place names obtained from the results of morphological analysis is added to the clause list, which is necessary when processing concepts such as person names and place names. It also normalizes alphabets, katakana, and symbols. The normalization process is a process of aligning full-width and half-width character codes, uppercase and lowercase letters of alphabets, and variants of Chinese characters into a certain constant. For example, consider the process of aligning alphabets to lowercase one-byte characters, half-width katakana to full-width characters, and variant characters to common kanji.
"et" can be converted into half-width characters "Alphabet", half-width "katakana" can be converted into full-width characters "katakana", and "Watanabe" can be converted into "Watanabe".

The vocabulary processing component 119 acquires the phrase list from the phrase analysis component 118, and uses the vocabulary dictionary 125 to add semantic information to the phrase list. Examples of semantic information include synonyms, synonyms, polysemous words, homonyms, and conceptual information. These pieces of information include conceptual information stored in the vocabulary dictionary 125 as a table, such as colors that are concepts such as red and blue, directions that are concepts such as west and east, and place names and person names. There is a special concept. As will be described later, in this embodiment, a special concept is used so that a numerical value, a person's name, a place name, etc. can be conceptually processed at the time of morphological analysis. Further, as will be described later, an expression regarding time such as 9:20 can also be included in the concept.

The case frame processing component 120
The phrase list to which the meaning is given is acquired from the vocabulary processing component 119, and the word or phrase considered to be the object of the verb is determined from the surface case and the concept. At that time, information on the deep case can be added to the clause list. For example, if the search input sentence is "search for a department store in Ginza", the verb is "search", and the object for this verb is surface case "wa case" and deep case "target case". It's a department store. Usually, the object of a verb is written in the form of "to", which is called "wa" in the surface case.
In addition, since the object of a verb is the target of the action of the verb in terms of meaning, it is called a "target case" in the deep case.

Further, the case frame dictionary 126 stores case frames corresponding to various words and phrases. A case frame is, for example, the phrase "search" is "~
In addition, the deep case has a frame (structure) of “search for (target case) by (place)”, and the concept of place name corresponds to “de case” and “place case”. A table is stored to indicate that a department store or the like corresponds to the “case case” and the “target case”. The case frame processing component 120 determines a word or phrase that is considered to be an object of the clause list, and then refers to the case frame dictionary 126 to determine how much the search input sentence matches the case frame.

When determining which case frame the search input sentence matches from the case frame information of the case frame information dictionary 126, there is a case where information is insufficient or vocabulary information is insufficient. In such a case, for example, the processing is performed by loosening the matching condition, such as matching using only deep case information. In this way, by loosening the matching conditions, even if the matching is originally difficult, the matching can be performed as it is.

When the search input sentence is matched with the case frame, if the surface case and the concept (or notation) match,
The rank 1 is set, and if there is no rank 1 one, only the concept (or notation) that matches, or the one that matches only the surface layer is searched, and this is set to rank 2. If there is no rank 1 or rank 2, general dependency information is adopted, and this is rank 3. The general dependency information is information such that "wa case" relates to a verb and "ni case" relates to a verb or a sahenun. As a result of the case frame processing, the case frame processing component 120 generates a sentence structure having information on the deep case of the case frame and information indicating in which rank the two match.

The attribute assigning component 121 uses the information of the sentence structure (clause) depending on the search support device 1 or the like,
For example, command parameter information and the like are added.
These pieces of information are stored in the attribute dictionary 127. As for the value when the special concept is used as an attribute, the value of the special concept can be directly used as the attribute value. For example, for a person's name, place name, numerical value, time, etc. processed by the concept, the input value can be directly used as an attribute.

The sentence structure acquired by the matching processing component 122 includes vocabulary information, case frame information, rank when matched with a case frame, attribute information, and the like. The matching processing component 122 calculates the degree of coincidence between the search input sentence and the case sentence and adopts the response of the case sentence having a high degree of coincidence as an answer. The degree of coincidence between the search input sentence and the case sentence is how much the two sentence structures match each other, and is calculated from the phrase information and the dependency information.

More specifically, the sentence structure to which the attribute is added is received from the attribute adding component 121, and the training corpus (a set of a search input sentence assumed to be input by the user and an answer corresponding thereto) is set. The case sentence and the search input sentence are matched using the answer dictionary 128 created from the case sentence 129. Then, the case sentence having a high degree of coincidence is specified, and the answer obtained from this is output as the output information 123. The answer is a set of a command specified from a case example, a parameter added by the attribute adding component 121, and the like. That is,
When the case sentence is specified by matching, the command or parameter is specified. In the case of the present embodiment, the abstract command is specified.

FIG. 13 is a diagram showing the structure of the vocabulary dictionary 125. The vocabulary dictionary 125 includes a synonym part 131 in which synonyms are registered, a synonym part 132 in which synonyms are registered, a polysemous part 133 in which synonyms are registered, a homonym part 134 in which homonyms are registered, and, for example, time and person name. It is composed of a concept part 135 and the like in which the concept is registered. The contents of these units will be described together with the following processes.

Next, a specific example of the analysis performed by the vocabulary processing component 119 will be described. [Synonym Analysis Example] FIG. 14 shows a synonym table 141 created in the synonym section 131. Synonym section 13
Reference numeral 1 is a collection of terms (synonyms) having the same meaning, and each synonym is tabulated in association with a representative word representing them. For example, “buy” is a synonym for “buy”, and “use” is a synonym for “use”. Now, for example, assume that the search input sentence is “buy a personal computer”. First, by the morphological analysis component 117, morphological analysis → PC / bu / buy /
And then the phrase parsing component 118.
By this, bunsetsu analysis → buy / buy a personal computer, it is divided into bunsetsu. Then the vocabulary processing component 119
A representative word (used here) of the synonyms is added to the clause list, such as "/ buy a computer (synonym = buy)".

[Example of Analyzing Synonyms] FIG.
Synonym table 14 when "use" is used as a representative word
3 is shown. In the synonym section 132, synonym information for processing expressions that are semantically similar but not completely the same is stored as a table. For example, "search", "search", "search", "search"
Are synonyms for each other, and "use", "use",
“Use” and “use” are synonyms for each other. In the synonym section 132, one representative word is selected from the synonyms (for example, the most commonly used phrase is selected), and the synonym information is tabulated and registered as synonym information.

For example, when the search input sentence is "search for a folk costume store", the synonyms are analyzed as follows in the same manner as the synonyms. Morphological analysis → folk costume store /
/ Search / Search, Phrase analysis → Search / Search folk costume stores,
Vocabulary processing → Search for a folk costume store (synonym = search),
As described above, the lexical processing adds synonym information to the clause list.

[Example of Polysemous Word Analysis] FIG. 17 is a diagram showing an example of the polysemous word table 145 of FIG. A polysemous word is a phrase that has multiple meanings. For example, the verb "pull" includes multiple meanings such as "subtract" and "pull". In the polysemous word part 133, these polysemous words are tabulated and registered. For example, when the search input sentence is “a tool for drawing a line”, the polysemous word is analyzed as follows. Morphological analysis → line / to / draw / ku / tool,
Phrase analysis → Line / draw / tool, vocabulary processing → Line / draw (polysemous word = subtract, pull) / tool. In this way, information on polysemous words is added to the clause list.

[Example of Analysis of Homophones and Synonyms] FIG. 17 is a diagram showing an example of the homophones and synonyms table 147. The processing for the homonyms is performed by adding the information of the homonyms to words and phrases that may be interpreted as different meanings in the voice recognition. As the homonyms, there are, for example, “contrast”, “object”, “symmetry” and “multiply”, “lack”, “draw”, “write” and the like. In the homonyms section 134, these homonyms are tabulated and registered.

For example, when the search input sentence is "PC classroom for the elderly", the homonyms are analyzed as follows. Morphological analysis → Elderly / Target / Target /
Yes / PC class, bunsetsu analysis → Aged / targeted / targeted / PC class, vocabulary processing → Aged / targeted (homonyms = contrast, symmetry) / Yes / PC class, in the phrase list Information on homonyms is added. By adding the information of the homonyms, for example, in the case of voice input, the possibility of erroneous analysis can be reduced.

[Analysis Example of Concept Information] FIG. 18 is a diagram showing an example of the concept table 149. In the concept part 135, for example, in order to collectively handle those having the same meaning contents such as color and direction, they are registered as a table as shown in the concept table 149. For example, "up",
The conceptual information of "bottom", "right", and "left" is "direction",
The concept information of “red”, “green”, and “blue” is “color”. Special concepts include person names, place names, numbers, and time. Among these special concepts, a person name, a place name, and a numerical value create concept information based on information given at the time of morphological analysis, and a time is a time table 151 registered in a time concept dictionary (not shown).
Create a time value from. FIG. 19 shows the time table 1
It is the figure which showed 51. As shown in FIG. 19, the special concept can also have a value. Items that span multiple clauses are separated by commas.

For example, when the search input sentence is "blue clothes", the concept information is analyzed as follows. Morphological analysis → Blue / I / Clothes, Phrase analysis → Blue / Clothes, Vocabulary processing → Blue (concept = color) / Clothes,

Further, when the search input sentence includes time such as "search for a new car released on August 3," the concept is analyzed as follows. Morphological analysis → 8 / month / 3 / day / to / sales / saved / new vehicle / of / search / phrase analysis → 8 / month / 3 / day / sold / new vehicle / search vocabulary processing → 8 (concept = numerical value · 8) / month / 3 (concept = numerical value; 3) day / updated / file / searching In this way, the numerical concept can have a value. It is also possible to process multiple clauses (8 / month / 3 / day) as follows. Vocabulary processing → 8 / month / 3 / day (concept = time; 2000/8/3) / updated / file / search

Furthermore, when the search input sentence includes a person's name which is a special concept such as "searching Socrates documents", it is analyzed as follows. Morphological analysis → Socrates / of / reference / search / search / phrase analysis → Socrates / search / reference / lexical processing → Socrates' (concept =
Person name; Socrates) / searches / references, like this,
The person name concept and the place name concept can have values.

[In the case of multiple clauses / multi-stage processing] The vocabulary processing component 119 processes a plurality of clauses in the same manner as in the analysis of the time concept, and
Information such as synonyms and synonyms can be added from the synonym portion 131, the synonym portion 132, and the like. For example, when the search input sentence is “apparatus for outputting a photograph on paper”, the following analysis is performed. Morphological analysis → Paper / to / photo / to /
Publish / apparatus / phrase analysis → Paper / photograph / append / apparatus,
Vocabulary processing → On paper (synonym = paper) / Photo / Issue (paper +
Publish → Synonym = Print) / Device, vocabulary processing component 119 causes synonym processing result “paper” and “issue”
Is further processed in multiple stages and "print" is added. This is because the synonym table 131 gives the synonym information "paper" to "paper" by the synonym table shown in FIG. 20 (a), and the synonym table shown in FIG. The synonym information “print” is added to “print” and “print”.

Next, the case frame processing component 120
A specific example of the analysis performed in 1. will be described. FIG. 21 shows
It is a figure showing an example of case frame information table 154 stored in case frame dictionary 126. For example, the phrase “export” (predicate) is usually used in the form of “export (target) to (country name)”. The phrase corresponding to (country name) is called "ni case" in the surface case and "opposite case" in the deep case, and the phrase corresponding to (target case) is "wa" in the surface case and "target case" in the deep case. It is called a case. In the case frame information table 154, "export", "wa",
“Industrial product” is registered as a notation corresponding to “target case”, and “country name” is registered as conceptual information corresponding to “case” and “other case”. The case frame processing component 120 matches the vocabulary-processed clause list with the case frame information stored in the case frame dictionary 126, and ranks the degree of matching.

For example, when the search input sentence is “resell an industrial product exported to the United States”, morphological analysis to lexical processing are analyzed by the following procedure. Morphological analysis → America / to / export / to / industrial products / to resell, bunsetsu analysis to / to the United States / to resell industrial products / vocabulary processing → to America (concept = country name) / to export / industry Product / resale

The case frame processing component 120
The phrase list subjected to the vocabulary processing as described above and the case frame information are matched in the following steps. Step 1: Using the case frame information table 154, the case frame information obtained from the "resell" clause is acquired. Regarding the target case, the case frame information table 1 of FIG.
Referring to the “Target Case” column of the “Export” column of 54,
"Industrial product" → "Resale" and the notation (industrial product) matches, so this match is rank 1.
・ ・ ・ (1) Regarding the counterpart, referring to the “Counterpart” column in the “Export” column, “To America (concept = country name)” → “Exported”, and the concept (country name) is consistent. Therefore, this match is ranked 1. ... (2)

Step 2: Acquire the case frame information obtained from the phrase "industrial product". Since there is no corresponding phrase in the “phrase” column of the case frame information table 154, there is no case frame information obtained from this clause. "phrase"
Words that can be predicates are registered in the column, and industrial products are nouns because they are nouns.

Step 3: Acquire the case frame information obtained from the "exported" clause. Referring to the “other case” column of the “export” column of the case frame information table 154, “America (concept = country name)”
→ It becomes “exported”, and the concept matches, so this match is ranked 1. ... (3)

Step 4: Examine the clause in which the dependency is not generated. The clauses that have no dependency are “exported”. In other words, there is no word that is in the position before "exported" and that is an opponent or target case of "exported". On the other hand, “sent” is a noun, a sahen noun, or an unregistered word because it is a continuum of verbs, that is, a form in which body phrases (nouns and pronouns) are connected. Here, we adopted a general dependency and "exported" →
"Industrial products". "Exported" means "industrial product"
It is a modifier of and is ranked 3. ... (4)

[0114] Dependence information (in this example, (1), (3), and (4) are adopted) by adopting the ones having a high rank from (1) to (4) and those having a close bunsetsu distance. , Create sentence structure. Case frame component 1 in FIG.
20 shows a sentence structure created by 20. As shown in FIG.
The meaning of the natural sentence input by the user is to "resell" an "industrial product", and the "industrial product" is "exported" to "America."

Next, a specific example of the processing performed by the attribute assigning component 121 will be described. Information regarding values depending on the search support server 1 and the like is added to the sentence structure created by the case frame processing component 120. Conceptual information such as a person's name, a place name, a numerical value, and time can be used as an attribute as it is. If these conceptual information are used as attributes, the vocabulary processing component 1
The value of the concept information acquired when the concept information is processed in 19 can be used as the attribute value. Further, when processing the attribute information in the attribute assigning component 121,
A method of independently acquiring the value of the conceptual information may be adopted.

FIG. 23 is a diagram showing an example of the attribute table 57 stored in the attribute dictionary 127. In the attribute table 57, a phrase, concept information indicating the concept of the phrase, and a parameter corresponding to the phrase are stored as a set. For example, “thin line” is “line type” as conceptual information, and thin line is parameter 1 of “line type”.
It is represented by. That is, the "thin line" is represented by "line type (1)". Similarly, the phrase "red" is "color (0x0000ff)"
Corresponding to.

For example, the search input sentence is "a tool for drawing a thin line".
It was. This search input sentence is processed by the following procedure. Morphological analysis → thin line / / / draw / ku, phrase analysis → thin line / draw, vocabulary processing → thin line / draw, case frame processing → (predicate) draw, (target case) thin line (case), attribute processing →
(Predicate) draw, (target case) thin line (case) (line type attribute =
1) In this way, the attribute assignment component 121
Attribute information (line type attribute = 1 in this case) is added to the sentence structure acquired from the case frame information component 120 by referring to the attribute dictionary 127.

If the search input sentence is "draw a thin line", it is processed as follows. Morphological analysis → Fine / I / Line / Of / Draw / Cut, Phrase analysis → Thin / Line / Draw, Lexical processing → Thin / Line / Draw, Case frame processing → (Predicate) draw (Target case) Line ( Case) (modifier) thin (adjective modifier), attribute processing → (predicate) draw (target case) line (wa) ・ ・ ・ (1) (modifier) thin (adnominal modifier) ・ ・ ・ (2) → From (1) and (2) (line type attribute = 1), it is also possible to acquire attribute information from a plurality of clauses in this way.

Next, the matching processing component 12
A specific example of the process performed in 2 will be described. The matching processing component 122 includes the sentence structure of the case sentence in the training corpus 129 and the attribute assignment component 121.
The sentence structures acquired from are compared to calculate the similarity (coincidence). The similarity measures how much the two sentence structures match, and as an example, as in the following formula,
A matching index may be obtained from the number of phrases in the answer candidate and the number of phrases in agreement with the answer candidate. (Matching index) = (Number of matching phrases) / (Number of segments of answer candidates) When deciding the ranking of answer candidates, not only the matching index but also the number of "matching phrases" is as shown in the formula below. Prioritize and decide the ranking. (Number of matching phrases)> (Matching index) That is, the answer with the largest number of matching phrases (numerator of the matching index) is adopted as the matching result, and if the number of matching phrases is the same, the size of the matching index. To judge.

The number of bunsetsus with matching matching indexes is 0.3 as a measure of how many bunsetsu matches are in deep case and 0.7 in notation / concept. What was done was calculated as 0.4).

Further, for a case where the rank is low due to the degree of coincidence of case frames (for example, in the case of rank 3), the value of the number of clauses is halved. In addition, in order to reflect the dependency information of the bunsetsu, if the bunsetsu before and after the bunsetsu match the answer candidate, the bunsetsu number is left as it is, and only the one after the bunsetsu is matched. If not, further reduce the number of clauses to 2/3. If both before and after dependency do not match, reduce the number of clauses to 1/3.
The degree of similarity can be calculated as described above. However, this is just an example and may be calculated by other methods.

The matching method performed by the case sentence matching unit 12 has been described above, but this is an example.
The search input sentence and the case sentence may be matched by another method.

As described above, the following effects can be obtained in this embodiment. It is possible to automatically obtain the output result of the search engine specialized in the field without burdening the user with selecting the search engine. The search support server 1 can analyze the structure of the search input sentence input by the user and automatically search the search site in an appropriate field. Since the search results output by the plurality of search engines are combined and displayed on the client terminal device, the user does not have to search around the target site with the plurality of search engines. Even if there is no case sentence that matches the search input sentence, the search conditional expression can be generated from the search input sentence. The user can set the narrowing level. Even web beginners can perform highly accurate searches without effort.

Further, in the meta-search method executed by a computer, when the similarity with the prescribed sentence structure closest to the user's question sentence is equal to or more than a prescribed threshold value, the prescribed sentence structure is associated with the prescribed sentence structure. It is possible to provide a meta search method for requesting a search from a third party search engine group of. Furthermore, the meta search method is
A phrase can be extracted using a position in a similar regulatory sentence structure as an index, and if there is a prescribed conversion rule associated with the prescribed sentence structure, it can be used as a search keyword after conversion. As conversion rules, all of the part-of-speech information (nouns, verbs, etc.) for each part-of-speech, semantic concept information (animals, documents, etc.), and dependency information (part-of-speech, deep-case, part-of-speech and meaning of the origin and destination) Alternatively, it can be composed of two parts: a conditional part consisting of a part and a conversion part outputting a character string into which the information of the conditional part can be embedded by a prescribed method.

Although one embodiment of the present invention has been described above, the present invention is not limited to the described embodiment, and various modifications can be made within the scope of the claims.

[0126]

According to the present invention, output results of a search engine specialized in a specific field can be automatically obtained without imposing a burden on the user for selecting a search engine.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of a network configuration using a search support server according to the present embodiment.

FIG. 2 is a diagram showing an example of a logical configuration of a search support server.

FIG. 3 is a diagram showing an example of an HTML document when a search command is converted into an HTML document.

FIG. 4 is a diagram showing an example of a conversion rule.

FIG. 5 is a diagram for explaining a word expansion interpretation rule.

FIG. 6 is a diagram showing an example of a search result interpretation rule.

FIG. 7 is a diagram showing a point calculation formula and the like for each URL.

FIG. 8 is a diagram for explaining the content of the operation performed by the search support server when a search input sentence and a case sentence match.

FIG. 9 is a block diagram showing a configuration of a search support server.

FIG. 10 is a flowchart for explaining the operation of the search support server.

FIG. 11 shows an example of a search result screen.

FIG. 12 is a diagram showing a configuration of a natural sentence matching system.

FIG. 13 is a diagram showing an example of the configuration of a vocabulary dictionary.

FIG. 14 is a diagram showing an example of a synonym table.

FIG. 15 is a diagram showing an example of a synonym table.

FIG. 16 is a diagram showing an example of a polysemous word table.

FIG. 17 is a diagram showing an example of a homonym table.

FIG. 18 is a diagram showing an example of a concept table.

FIG. 19 is a diagram showing an example of a time table.

FIG. 20 is a diagram for explaining multistage processing performed in vocabulary processing.

FIG. 21 is a diagram showing an example of a case frame information table.

FIG. 22 is a diagram showing an example of a sentence structure created by a case frame processing component.

FIG. 23 is a diagram showing an example of an attribute table.

[Explanation of symbols]

1 Search support server 2 Client terminal device 3 search server 5 Internet 11 Search input sentence acquisition part 12 Case sentence matching section 13 Converter 14 Parameter conversion rule database 15 Search Request Department 16 Output section 17 Output screen rule database 71 control unit 72 Input device 73 Output device 74 Communication control device 75 bus line 80 I / O interface 81 Storage medium drive 82 storage device 83 Program storage section 84 data storage 117 Morphological Analysis Component 118 clause parsing component 119 Vocabulary Processing Component 120 case frame processing component 121 Attribute assignment component 122 Matching processing component 125 vocabulary dictionary 126 case frame dictionary 127 attribute dictionary 128 answer dictionary 129 Training Corpus

Claims (7)

[Claims] 1. A search input sentence acquisition unit for obtaining a search input sentence, a matching unit for matching the obtained search input sentence with a prepared case sentence, and a search engine associated with the matched case sentence. A search engine specifying unit that specifies a keyword, an extraction unit that extracts a keyword from the search input sentence, a search request to the specified search engine using the extracted keyword, and a search result that is obtained from the search engine A search support device comprising: requesting means; and presenting means for presenting the acquired search results. 2. The matching means calculates a degree of similarity between the search input sentence and a case sentence, and matches a case sentence having a degree of similarity larger than a predetermined threshold with the search input sentence. The search support device according to claim 1. 3. The search request generating means further comprises a search expression generating means for generating a search expression using the extracted keyword, and when there is no case sentence matching with the search input sentence by the matching means, the search requesting means is The search support apparatus according to claim 1 or 2, wherein a search is requested to a predetermined search engine using the generated search formula. 4. When there are a plurality of search engines specified by the search engine specifying unit for the case sentence, the presenting unit synthesizes and presents the search results of the plurality of search engines. The search support device according to claim 1, wherein the search support device is any one of claims 1, 2, and 3. 5. The apparatus further comprises an ordering unit for ordering the search results acquired from the plurality of search engines, and the presenting unit presents the search results in the order given by the ordering unit. Item 4. The search support device according to item 4. 6. A computer provided with a search input sentence acquisition means, a matching means, a search engine specifying means, an extraction means, a search request means, and a presentation means, wherein the search input sentence acquisition means performs a search. A search input sentence acquisition step of obtaining an input sentence; a matching step of matching the obtained search input sentence with a prepared case sentence by the matching means; and a matched case sentence of the search engine specifying means. A search engine identifying step for identifying a search engine associated with the search engine; an extracting step for extracting a keyword from the search input sentence by the extracting means; and a search identified by the search requesting means using the extracted keyword. A search request step for requesting a search from an engine and obtaining search results from the search engine; And a presenting step of presenting the acquired search result by the presenting means. 7. A search input sentence acquisition function for obtaining a search input sentence, a matching function for matching the obtained search input sentence with a prepared case sentence, and a search engine using the matched case sentence. A search engine specifying function for specifying, an extraction function for extracting a keyword from the search input sentence, a search request for requesting a search to the specified search engine using the extracted keyword, and acquiring a search result from the search engine A search support program for realizing a function and a presentation function for presenting the acquired search result on a computer.