JP2003345829A - Method and apparatus for retrieving information, and computer program for information retrieval - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a system or a method which is more convenient in operation for retrieval, and to provide a computer program. <P>SOLUTION: When a database is retrieved based upon retrieval information entered into an area 1201, a plurality of labels for narrowing down retrieval results are extracted and displayed in an area 1204, a label is specified to further select retrieval results according to the specified label, and the results are displayed in an area 1203. When the specification of the label is changed, the selection results in the area 1203 are changed. <P>COPYRIGHT: (C)2004,JPO

Description

DETAILED DESCRIPTION OF THE INVENTION

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and system for searching a database or a computer program for searching.

2. Description of the Related Art As an example of searching for target information from a large amount of stored information, a search key including
Alternatively, a method of searching for a document similar to a search key is known. In the document search, there is a method of searching for a document including a specified search key like a full-text search. There is also a concept search method for finding a document similar to the feature vector of a search key by using the frequency of words used in the document as a feature vector of the document. Such a technique is described in, for example, Japanese Patent Laid-Open Publication No. Hei 7-114572. Also, Japanese Patent Application Laid-Open No. 2002-12355
No. 1 describes a content for preventing a reduction in search processing speed and a decrease in operability that occur when a large number of search words exist.

There is a need for a system and method that is easier to use in search processing. An object of the present invention is to provide a system, a method, and a computer program in which a search operation is more convenient.

According to the present invention, a plurality of labels as described in the following embodiments are extracted based on search information for searching for a search target such as a database, and a specific label is extracted from these labels. By designating a label, sorting is performed based on the designated label, and the final search result of the database is output. According to the present invention, it is possible to optimize the search result by further changing the label designation while looking at the output result.

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments according to the present invention
explain. FIG. 1 shows the configuration of the information retrieval system. Book
The stem includes a plurality of operating devices 102 and these operating devices 1.
02 that executes a search based on a search request from
Device 101. The operation device 102
Each operation device 102 is a device for performing a search operation.
Keyboard and mouse as force devices and output devices
It has a display. Also, the search processing device 101
Means information specifying means (not shown) and similarity calculating means (not shown)
No) and information storage means (not shown),
Furthermore, the index information database 103 and the accumulated information
Database 104 and file information database 105
It is connected. The index information database 103,
Storage information database 104, file information database
Database 105 is used to execute a search process.
Is. FIG. 2 shows the information retrieval system performing a retrieval process.
When performing similar information search processing,
Configuration and the relationship between the operation of the above configuration and the use of the database.
It is operation | movement explanatory drawing which shows a relationship. Search key input means 209 and
The output device 215 such as a display is provided by the operation device 102.
It is a configuration to do. Also, the search processing device 101 inputs information.
Means 201, feature extracting means 202, information specifying means 20
3. A similarity calculation unit 204 and an information storage unit 205 are provided.
Furthermore, the search key corresponding information search means 210 and the activation value setting
Setting means 211, activation value propagation means 212, selection means 213
And output means 214. Here, the information
The operating device 102 has the force means 201, and the operating device 102
To input information to the search processing device 101 from
Good. Also, the output unit 214 of the search processing device 101
Generates output information data for displaying
The data is transmitted to the operating device 102 and displayed by the display unit 216.
It may be displayed. The feature extracting means 202 which is the above means,
Information specifying means 203, similarity calculating means 204, information storage
Means 205, search key corresponding information search means 210, activation value
Setting means 211, activation value propagation means 212, selection means 21
3 and the output means 214 constitute the search processing device 101
Computer to implement the above functions on a computer
It is realized by loading and executing a program. I
Index information database 103 and stored information database.
And the file information database 105 search processing.
Although it may be provided in the internal memory of the device 101, the configuration and
In order to make it easier to understand
Described as being outside. Information retrieval system is a search key
Search processing based on input information input from input means 209
It has a function to perform the processing and output the result. example
If the information retrieval system is input from the retrieval key 209,
Documents that describe similar content to the input information are stored
Search from the information that is
Collect knowledge from accumulated documents, or
It has a function to perform pattern search and the like. Information retrieval system
The processing content of the system can be roughly divided into
Search processing and output processing can be divided. Accumulation
The processing at the time of loading and searching is mainly performed by the search processing apparatus 101 and the interface.
Index information database 103 and accumulated information database
104, a file information database 105.
The search processing device 101 and the index information database
103, the accumulated information database 104, the file information
The database 105 operates as a server machine. Accumulation
The configuration of the search processing device 101 related to the processing at the time of stacking
Report input means 201, feature extraction means 202, information identification means
203, similarity calculation means 204, information storage means 205
It is. The configuration related to the processing at the time of the search is the search key
The input means 209 and the search key corresponding information search means 210,
Sex value setting means 211, activation value propagation means 212, selection means
213. Output processing is mainly performed by the client machine.
The operation is performed by the operation device 102 which is a user. However, the server side
The search processing device 101, which is a thin, performs processing up to output processing.
It is also possible to In the following, the accumulation processing and
The search processing and the output processing will be described in this order. In this description, “Information
Using a document as an example, a search key, ie, a search key
, Related to information (search information) such as documents and words,
An example of searching similar documents collectively
Will be explained. FIG. 3 and FIG. 4 show the flow of processing at the time of accumulation.
In step 301 of FIG. 3, the information input means 20 shown in FIG.
1. Document or directory or document such as URL
Is input to the information retrieval system. Step
In step 302, the information input by the feature extracting means 202
Are extracted from the document. This feature extraction operation
Then, when targeting a document, words such as morphological analysis
Extraction means can be used. In step 303,
Using the extracted features, the information specifying means 203
Information input as needed is captured in smaller units
It is. For example, if the input information is text,
This can be done for the whole document, or for chapters and sections, formal paragraphs and lexical
Identify semantic units in units such as chains
I do. As described above, the information specifying unit 203
By specifying small units, these units are meaningful
It is possible to use the unity of search as a unit of search. This
As a result, the search target can be more accurately selected. sand
That is, since redundant parts other than the search processing part are reduced,
If search keys are given properly, interference from redundant parts
Search from the part of the document with less influence
Sorting can be performed accurately and highly accurate searches can be performed. Also the whole document
The body is not obtained as a search result.
Since you can search in fine units, read the search results document
You don't have to read the whole thing, you can identify relevant parts of the document
You. Therefore, the burden on the searcher after the search can be reduced.
Wear. In steps 304 and 305, the information specifying means 203
Of the document in smaller units (hereinafter simply the document
) Is sent to the similarity calculating means 204
More similarity calculation is performed. In the similarity calculating means 204, the information
Document passed from the report identification means 203 and accumulated up to that time
Similarity, which is the degree of similarity of content between the document and
Calculate the degree. When calculating similarity, multiple documents
Word pairs (particularly important words) and similar word pairs
Documents with have the property of representing similar contents. This gender
Utilizing quality, similarity measures the word in each document.
Dot product of vectors when importance is a vector element
It is expressed by the mutual relationship of words such as and dependency. here
Where the word importance is the number of documents in which the word appears and the total number of documents
, Such as log (total number of documents /
Number of occurrences) and the word in that document
It is a value represented by the product of the number of occurrences. Next,
A processing method effective for reduction will be described. Similarity calculation means
If new information is stored by 204,
It is necessary to calculate the degree of similarity with the stored information.
You. In addition, the difference between stored information due to changes in the importance of words, etc.
When it is necessary to recalculate the similarity, the similarity between stored information
The degree of similarity between any two stored information to calculate the degree
Needs to be calculated. When the number of stored information is n, calculation
Since the amount is n × n, the number of stored information
Calculation time is required. Therefore, the similarity between all information
Instead of calculating the
Calculate similarity only between groups of classified information
By doing so, the calculation time can be reduced
You. In addition, if there is an environment where processing can be distributed,
The similarity between the classified information can be calculated in parallel,
Calculation time can be reduced. How many information
When categorizing to or
Classification methods can be used. In addition, patent information
Already classified, use the classification information
be able to. Documents into several groups in advance
When splitting, statements that can belong to multiple groups
The book is considered to belong to multiple groups, and each group
Calculate the similarity within the group. With the above processing, the similarity
Calculation time can be reduced. Next, index information
The accumulation of data in the database 103 will be described. information
The accumulating unit 205 calculates the similarity greater than or equal to the calculated threshold.
Together, the documents are registered and stored in the storage information database 104
I do. The extracted features are stored in the index information database.
File 103, and the information on the file name of the document is stored in the file
The information is stored in the information database 105. Stored information database
The format of the data stored in the source 104 is shown in FIG. Affection
The report id is attached to the information specified by the information specifying unit 203.
Number. As described in this explanation, the sentence
If a document is used, it is uniquely assigned to a small part of the document.
The number given as the information id. From which position in which document
A sentence uniquely attached to the document as to whether the information was issued
The start position of information according to the writing id and the number of characters from the beginning
Expressed by the end position. Words in the document and their importance are also preserved
And used for similarity calculation. Here, st in the data
p is information for indicating the end of the list.
It is a fixed value (minus 1). In other FIGS. 6 and 7, st
p is similar information and indicates the end of the list. Degree of similarity
The similarity calculated by the calculation means 204 is shown in FIG.
Thus, the weight of the link together with the information id
For example, it is held as a “weight value”. Calculation target
Information id is linked as stored information similar to each other.
Having. The link weight depends on the similarity of each information id.
Is set by the value. Index information database 103
FIG. 6 shows the format of data stored in. The word id is special
Uniquely attached to the feature extracted by the feature extraction means 202
Number. Words that mean actual characteristics are strings
And hold. The position of the word used for searching is
A list of the document ids that appear and their locations in the document
Is represented by File information database 105
FIG. 7 shows the format of the data to be stored. Document id is stored
This is a number that is uniquely assigned to the document. Document file
File name is stored as a character string. FIG. 8 shows the flow of the search process.
You. In the search process, the information input method of the operation unit which is the operation device
For example, a search is performed by the search key input means 209.
Search information is entered. In step 801, the search information
Search key corresponding information search means 2 as a search unit based on the information
A document corresponding to the search information is searched by 10. Search
The unit 210 simply extracts the search information by the feature extracting unit 202.
Extract words and store them in the index information database 103
Refers to the information database 104 and corresponds to the search information
Search for documents. Documents that correspond to the search information
Information, or documents containing search term co-occurrence or synonyms
It is. In step 802, the activation value providing unit 211
Of documents that correspond to the search information
The initial activity value is given. Initial activation value is searched in the document
The sum of the frequencies of words that correspond to information and the search
The sum of the importance of the word corresponding to the report, the vector between the search information and the document
This is the inner product of the torque. In step 803, search information
When the initial activation value is assigned to all the documents corresponding to
The weight of the link is added to the initial activation value given by the value
Process to propagate the activated value to the link destination and add the activation value
Work. The propagated activity value is determined by the activity value providing unit 211.
Is the initial activity value given by Active value propagation processing
It may be repeated several times. If you repeat multiple times,
The activation value propagated by the activation value propagation process of
And the same processing is repeated. In step 804, the activity
When the value propagation processing is completed, the activation state determination means 213
Information with an activity value that is higher than the threshold is collected and collected
The obtained information is passed to the output processing together with the activation value.
The flow and operation of the output process will be described with reference to FIGS.
explain. The collected document is a pointer to the document
The operation unit 1 is output from the output unit 214 in the form of a node 1001.
02 is output to an output device 215 such as a display.
Nodes repel other nodes, but have similar documents
The nodes representing the chiefs have similarity represented by the link weight.
Arrange nodes on a two-dimensional screen to attract each other with proportional force
Place. With this arrangement, similar information groups can be collected.
Therefore, the searcher can visually know the desired information.
Displayed nodes are highlighted according to their activation values
Display in the form. When a node is selected, the destination document
It is displayed on another screen. Also, when you display the node
Label, which is an important word in the document that characterizes the node
File 1002 is also displayed together with the node. Simply displayed
A word (label) is, for example, a word (label) in a document.
The product of the word's importance and the document's activity value, including the word
The sums found for all documents
And ask. Labels use the size of importance as an attraction
Placed near the node. La included in multiple documents
Bells are displayed near nodes representing those documents
Therefore, the content represented by the collection of nodes can be known.
Of the displayed labels, those related to the target document
Narrow down the target document by selecting words
Can be. In other words, important words (labels)
Are displayed two-dimensionally based on their relationship to each other,
By selecting this word, further narrow down from search results
Becomes possible. Also, if the user is not a search expert
Search documents so that they can be used without difficulty.
Instead of displaying them on the screen in the form of
View the document title and contents on a separate screen in
Multiple related words (labels) are visually noticeable
Is displayed, and these multiple words (labels)
Users find it appropriate in cut-and-try format
Word (label), and the document is selected based on this.
This makes it very easy to use. This place
If the label is displayed at the node (display is
Is determined by locating in the vicinity after determining the position
it can. Also, the degree of association between the label and the node
Generates a vector of labels with the elements of the
Labels repel each other using similarity as an attractive force.
Bells may be placed. In addition, the display of the output information and
Operations such as narrowing down will be described later. Using a specific example
The similar information search process will be described. For example, user
Made the following search request:
You. The search request states, "When cholesterol is low,
I want medical information about what, what kind of illness
I ". Before this search, the next data is stored in the database in advance.
Data is retained. That is, the following three documents are stored information
This shows a case where the information is stored in the database 104. Document A: Risk of Cerebral Hemorrhage When Low Cholesterol
There is. Document B "Hyperlipidemia and Hypertension Complicate Blood Vessels
Reduce thyroid and nutritional status as it weakens and increases the risk of cerebral hemorrhage
To check. Document C: Low-cholesterol foods include tofu, udon, and bite
Such as rags. Here, all sentences from document A to sentence C are described in one sentence.
However, this has been shortened to make the description concise and easy to understand.
It does not have to be a single sentence. Document storage
At times, features are extracted from the stored documents. Feature extraction
Step 202 extracts words of nouns and adjectives using morphological analysis.
When issued, the characteristic words extracted from each stored document are
The words are shown as the following features A to C. Feature A is
"Cholesterol, low, cerebral hemorrhage, danger". Special
Sign B is "hypolipidemia, hypertension, complications, blood vessels, cerebral hemorrhage, danger
Sex, thyroid, nutrition, condition, check. Feature C is
"Cholesterol, food, tofu, udon, kamaboko"
is there. The above information is in the format shown in FIG.
It is stored in the database 103. Information specifying means 203
Lexical, which is a range of formal paragraphs or similar words
A semantic unit such as a chain or the whole document is a unit of information
Become. In the case of documents A to C given as an example, this sentence
For example, it is a unit. Specified by the information specifying means 203
For the unit, similarity calculation means 204 calculates similarity
I do. Total similarity between information in similarity calculating means 204
As a calculation method, the importance of words in the document (appearance frequency, t
vector whose elements are values obtained by the f-idf method or the like)
An inner product between them can be used. Similarity in document
As the dot product of the vector with the occurrence frequency of the word
Calculate similarity between documents A to C
Then it looks like this: Similarity AB = 2 (cerebral hemorrhage + risk) Similarity AC = 1 (cholesterol) Similarity BC = 0. The information storage means 205 determines this similarity between documents.
Stored information database 1 as link weight (weight value)
04. The documents targeted for calculation are mutually
There is a link as similar accumulated information. Then specific
Explain the search. In response to documents A through C,
For example, a search is performed using search information
Will be explained. By the search key corresponding information search means 210
Documents A and C containing "cholesterol"
It is. The activation value assigning means 211
Activity value. In this example, the search information in the document
To assign an activity value according to the frequency of occurrence of a word
Take. By this method, the active values of both documents A and C are
1.0 is given. The assigned activation value is the activation value propagation
Means 212 propagates to the linked document. Propagated
Is the product of the link activity value and the link weight.
The value to be represented. From document A to document B, a value of 2.0
However, a value of 1.0 is propagated and added to the document C. Document C
Are propagated and added to the document A by a value of 1.0. In addition,
Propagation of gender values is performed in parallel. Active value propagation is over
Thereafter, the final active state is determined by the active state determination unit 213.
Is determined. In this example, (activity value of document A) is 2.
0, (Activity value of document B) is 2.0, (Activity value of document C)
Is 2.0, and documents A, B, and C with large activity values are detected.
It will be collected as a search result. This allows traditional search
Document B that cannot be searched can be collected as a search result.
Next, the processing of the output unit 214 will be described. Output means 214
Receives information id and activation value for documents A to C
Output to an output device 215 such as a display. This
In the case of, the information document is represented as a node, and this node
When a code is selected by clicking or dragging,
Searched on the screen or in a different display area on the screen
The content of the information is displayed. Nodes repel each node
And similar according to the link weight
The nodes are arranged to attract each other. Given as an example
In documents A to C, "cerebral hemorrhage and danger" exist between documents AB
With regard to "cholesterol" between documents AC
Since the link is provided, as shown in FIG.
The word “cerebral hemorrhage / danger” is displayed near A and document B.
You. In addition, “cholesterol” is distributed near Document A and Document C.
Is placed. In the figure, links between nodes are represented by dotted lines.
You. Important words in each document are also displayed near the document.
Indicates the content of the document. FIG.
0 means how many words are extracted by the feature extraction process.
Is displayed. Of these words, the medical content
Searchers, such as "cerebral hemorrhage" or "high blood pressure"
By selecting words that are relevant to the information you ’re looking for,
Is the medical information for low cholesterol document A?
And that relevant information can be obtained from Document B.
The user can easily understand the collection of information
Wear. Note that this screen shows the relationship between nodes and labels.
Both node 1001 and label 1002
Is displayed, but the node is displayed on another screen or
It can be displayed in another display area on the surface. This
May be easier to operate. Also the node
Output the document in a format that outputs the title of the document.
Can be displayed in the area of
It is displayed in list format, while there are no
The label can be displayed in a format that does not display the code.
This is easier to understand and the subsequent operations are easier
There are cases. For a detailed description of such an output screen
Will be described next. This has been described with reference to FIGS. 1 to 10 described above.
In search processing and storage processing, explain sentences as examples.
However, the basic operation is the same for literature. Writing
Provides more granular search results, and
The text is taken as an example because it is easy to use. Then operated by multiple users
A search process performed from the operation unit as the device 102 will be described.
I do. However, the basic operation is the same as the above operation. First
In step 1101, similar information search performed by the client
FIG. 11 shows a flow showing the operation of (1). Client side
Information retrieval program on the operating device 102 which is a machine
When you launch a ram, that is,
Execute the search program by selecting, etc.
Then, a screen as shown in FIG. 12 is displayed on the display. This
Screen has the following three display areas and operation buttons.
Provided, making it easy for users to use. table
The display area 1201 stores search information to search for information.
This is a field to be input, that is, a display area. Display area
Information for search is ordered 1201 according to user input.
Will be displayed. You can correct, delete, or add by looking at this display.
It is possible. Operation button 1202 for sending search start instruction
Is displayed. This operation button is used for other printing and other
Even if both are provided at positions further above the area 1201
good. When displayed as in this embodiment, other menus are displayed.
Display larger than the menu, making it easy to use for inexperienced people.
There is fruit. A search button 1202 for performing a search
When selected, a search execution instruction is sent based on this and the search is
The rope starts. A display area 1203 displays information of the search result.
This is an area for displaying, and a display area 1204
Is an area for displaying a label. Search results and LA
The bells can be displayed in contrast and updated while selecting the label.
The operation of narrowing down to is very easy. The above description
An example will be described with reference to FIG.
Search request that is search information "cause when cholesterol is low
Want medical information about what
New ". Search area 1201 based on input
Check the information displayed and select the search button 1202
I do. Based on the search instruction, area 1203 is displayed in FIG.
The search result 1001 which is the node of is displayed in the form of a list
Is done. That is, sentences A, B, and C are displayed. With it
In the area 1204, a word related to the search information
A bell is displayed. That is, the words "tofu", "kamaboko",
"Cholesterol", "danger", "cerebral hemorrhage", "blood
"Tube", "thyroid", and "hypertension" are displayed. This label
If you select the word "tofu" from the text display, sentences A, B,
A sentence C is selected from C. The word "cholestero"
Is selected, sentences A and C are selected. Yet another
A specific example will be described with reference to FIG. Area 1201 in the figure
Then, input “quality plan” as search information. FIG.
Is an example where search information is displayed based on the above input
You. Displayed as described in step 1102 of FIG.
Search button 1202 is selected. Here, search information
The words entered as information are both "quality" and "plan"
Expect documents to be included or documents that contain either
And input it. Of course, as mentioned above,
Search for similar documents related to "quality" and "plan"
A search method may be used. When the search button is pressed,
Step 1102, the search information is sent to the server and the step
1103 search processing is executed. In step 1104,
Search result data from server to client side operation unit
102. In step 1105, the server-side device
14 based on the search result data received from the
The screen is displayed. The display area 1203 in FIG.
The search results for the search information "quality plan" are displayed.
You. “Narrowing down” displayed in the search result display area 1201
At the bottom of the screen of the
The title list 1403 is shown. By title alone
Document summaries and the relevance of search information to documents.
In addition, it may be displayed. Users can use this narrowed down document
Select the target document from the list and click the file name
File selected based on the selected operation
Can be displayed or obtained. "Related information" displayed in FIG.
1404 was searched but not narrowed down
In this example, the search (including all selected labels)
Document) is displayed. In this figure, the search results
Documents collected by search are documents containing all search information
Therefore, all documents are "refinement information". Search
When the similar information search method described above is used as the method
Immediately after the search does not contain all the search information,
Document will be displayed under “Related Information”.
You. The list of documents under "Refinement information" and "Related information"
If you select "Refinement information" and "Related information", for example,
Display and non-display can be switched by clicking and other operations
It works. Nodes are displayed on the label display screen 1204.
Words that characterize the document, eg, labels 1405, are displayed
It is. The displayed label is the searched node (document)
Characterize them for one or more documents
Is the word automatically selected as The word is it
It may be an icon that features. displayed
Register labels that are not required in the database in advance
By doing so, it can be removed from the display target. Label is
It is possible to select by clicking or dragging.
Also, deselect by reselecting the selected label
It is also possible. In the state immediately after the search shown in the figure,
Words entered as search information are selected as labels
ing. Indicates that the selected label is in the selected state
The display is visually distinguishable, that is, highlighted.
In this example, it is surrounded by a thick line. Making a label selection
To narrow down the documents containing the selected label
Document and other documents are classified and the result is displayed.
It is displayed in the information display area 1203 on the left side of the screen. label
The display position of the related label (characterizing similar documents)
Bells are near each other, and the distance increases as the relationship decreases
It is arranged so that it becomes. For this reason, searchers are the target
Finding multiple labels that characterize a document at once
Labels can be applied to every corner of the screen
The search effort can be reduced. Labels on the display screen
Label while maintaining the relative positional relationship in two dimensions.
The image is displayed enlarged to the maximum size that can fit within the display screen.
You. You can also manually move the labels to the preferences of the searcher.
Both are possible. Displayed in the label display area 1204
By selecting an appropriate label from among the labels
Can be narrowed down (step 110 in FIG. 11).
6). FIG. 15 shows this state. For example,
Of the search results of “quality plan” in FIG.
Learn about quality planning to reduce quality and increase poor picking
If you want to change the label from the label in the label display area,
By selecting "Good" and "Extract" in order,
It shows how quality plans can be extracted. This
The searcher looks at the contents of the searched documents before narrowing down in order.
To search for the desired information or to enter new search information.
The effort of performing a refined search can be reduced. Figure
Reference numeral 15 denotes the label “quality” 1501 and the label “total” in FIG.
Image 1502 is selected and the label "bad" 15
03 is shown. "Refinement information" 1
In 504, all the selected labels "quality, plan and defect" are displayed.
The document 1505 including the document is displayed. "Related information" 15
06 is the document of “refinement information” among the retrieved documents
Other documents 1507 are displayed. From the refinement
It is off, but you want to display or get this information
If you want to, you can operate from this area. FIG.
6 further shows the state where the label “extract” 1601 is selected.
Show. Note that all the selected labels
Instead of a document containing
You may. The specification method is the same as the general search method.
Separates search information with symbols such as "+" and "|"
It can be specified by a method or the like. Similarly, of the quality plan, I
A place to know about the quality system in preparation for SO9001
FIG. 17 shows the case. "Quality" 170 from the label display screen
1, "Plan" 1702, "Internal" 1703, "Audit" 1
704 and the target "quality system
"Maintenance and management" 1705 etc. can be extracted. Initial state
Select the label corresponding to the selected search information
To remove the label displayed on the label display screen.
By selecting a more appropriate label,
It is also possible to gather appropriate nodes based on the information. FIG.
By narrowing down from 5 to 17
`` Quality planning defect extraction '' for the purpose of obtaining a quality plan
Search information to identify the target document
Search with the search information that came up with, for example, “quality plan”
And select the label associated with the desired document.
Going to get a target node, for example, a document
Can be. With this operation, the searcher can search
Book can be classified,
Compared to cases where documents are classified by clustering or other methods
In addition, a necessary and sufficient document group consisting only of
Easy to obtain. In FIG. 13, without inputting search information
By performing a search or other means, the tendency of the document to be searched can be determined.
It is also possible to know. For all documents to be searched
By presenting representative labels that have been
It is possible to know whether such documents are stored.
Classify the results into word processing software, tabular software, html or
xml and other file formats to output documents
Document that summarizes the trend of the company
Every creation is easy. 14 to 17, in FIG.
When selecting a label, enter the selected label in the search information input field.
Selected label, for example by reflecting it in the field
It is also possible to search again with. Search by this operation
The user does not have to enter the search information again,
A collection of documents centered on files and retrieved documents
The label to attach, that is, the label associated with the selected label
Can get a group of
Documents and their trends. Searched conversely
By selecting one or more documents
Labels can be selected. This
The information contained in the selected document is represented by a label
It is possible to predict the contents before reading the contents of the document.
This makes it easier to find the target document,
Reading required documents can be suppressed. FIG.
Some more functions are added to the embodiment described using
The startup screen at the time of addition is shown. Include search information
Similar search function that can search
When a button 1801 for performing similarity search is selected,
When you can't think of the search information that identifies the document you want
Also spread similar documents with somewhat related search information.
Many searches can be performed. In addition, many documents
Even if it is searched, you can narrow down the documents by selecting the label
Is there a target document in the search results?
It can be suppressed that the user does not understand. Ma
Search information to select the source of the document to be searched.
The search target is switched by selecting the report selection field 1802.
By replacing, only documents for that search target can be
Obtaining important search results can be suppressed. Ma
In addition, the source of the document to be searched is switched
If you perform a trend analysis of what documents are
You can know if it is stored. Also, display
Label attribute selection field 1 for selecting the label attribute to be used
The attribute of the label displayed by 803 (for example, a place name or a person name)
), You can specify the attributes of the labels to be displayed.
So that searchers can easily select a label
Become. For example, if you are seeking documentation on the type of beef
If you select a place name as a label attribute, search for beef
If the search is performed as shown in FIG.
Is displayed as the label 1901,
Can easily select a desired document. FIG.
"Refinement information" and "Related information" on the information display screen
Operation area 180 for setting the order of documents to be displayed
4 is provided, and by this setting, the score order and selection of search results
In the order that contains many labels (in the order of label rank in the figure),
You can select the date order to be displayed in the newest date order. Figure
Reference numeral 20 denotes a modified example of the information display screen. Multiple documents
If you want to stop and select, check the box before the document
Select the label to narrow down by checking
If you select it, you can operate it with the keyboard or right-click
Display the document that is the selected node in the menu 2301
Can be realized by adding items such as
You. FIG. 21 shows labels 2101 and 210 on the information display screen.
2 also shows an example incorporating the same. Label display screen is displayed
Even on simple search screens that cannot be performed,
By checking the check box 2103,
The same function as selecting a label on the label display screen
Can be used. The label list 2101 on this screen is
Click on the “Label” 2102 to display or
Switching of non-display is possible. FIG. 22 is similar to FIG.
Label is selected while label 2201 is displayed
Then, the document 2202, which is the node related to the label,
It is a state in which it is displayed. This allows searchers
Can easily find the document associated with the label. This
Function can be used while pressing the shift key on the label display screen.
Pop-up window by clicking on the bell
It can also be realized by displaying in windows
You. Once the documents related to the label are available, the directory
It is also easy to classify structured files (documents).
You. For example, the administrator of company documents can classify and
Help maintain document classification for directory search
Can be FIG. 23 shows the first information display screen.
, For example, when a right click is pressed.
The menu 23 is displayed by the first operation such as the case of the right click.
01 is displayed and functions related to the operation of the selected document are displayed.
It is shown. In the example, the document 2302 of “Quality Plan”
On the other hand, the following functions can be selected. "This document
"Display": Display the contents of the selected document on the screen.
"Detailed information": title and author of the document, beginning of the document
Detailed information table, such as a summary of the part or selected label
Indication function. "Change font size": Font size
The ability to change. "Change Color": Type of document displayed
Function to change the background color and font color of the tor. "Similar
Information search ": similar to selected document as search information
Ability to search for documents. In addition, these functions are keyboard
Operation is also possible. FIG. 24 shows a modification of the label display screen.
It is an example. Search information among labels 2401 and 2402
By enclosing a label 2401 corresponding to
It is a standing one. This allows the searcher to search the search information
Bells can be easily found, search information and other labels
You can understand the relationship with the file. Figure 25 is a label table
13 is a modified example of the display screen. Each label 2501 shows
The number of documents in parentheses after the label.
You. This gives the searcher the distribution of the document indicated by the label
You can know. Also, instead of the number of documents,
The importance (score) of the file may be displayed. Also, FIG.
It is a modification of a label display screen. Important label 2601
It is a large representation. Figure 27 shows the label display screen
This is a modified example. Label importance, label attributes, etc.
For example, place names and personal names, or labels 2701, 27
The colors are classified according to the relevance of 02, 2703.
FIG. 28 shows a modification of the label display screen. Label 280
By making the display of 1 three-dimensional instead of two-dimensional,
Search the bells for better understanding of their mutual positional relationship
Can easily select related labels. This screen is
Rotate in any direction by dragging on the screen
Can visually understand the distance in the depth direction.
Wear. Fig. 29 shows that when right-clicking on the label display screen is pressed
Screen. Right-click for menu 2901
Displayed and functions related to the operation of the selected label
, The functions related to the entire operation are displayed on the lower side.
In the example, for the label “Inside” 2902,
Function can be selected. "Don't show this label
I ": Function to hide the selection label. Separate instead
It is also possible to give the label of the searcher,
You can continue selecting labels for the documents you want
You. "Detailed label information": Number and weight of documents represented by the label
Detailed information display function such as necessity. "Font size change
Update ": Function to change font size. "Change color":
Function to change the background color and font color of the label. Also
The following functions can be selected as the whole operation.
"Re-search": A function to perform a re-search on the selected label.
"Change number of displayed labels": Change the upper limit of displayed labels
Function to do. Change the number of labels near the label display screen
To prepare radio buttons, slide bars, etc.
feasible. "Add label": displayed on the screen
A machine that can add a label that you do not want to display
Noh. You can redisplay hidden labels,
Ability to add and display unlabeled labels. "UND
O ": Function to return the label selection state to the previous state. (1
0) "Last search result": Return to the previous search state
function. In addition, these functions can also be operated with the keyboard
It is. Displayed when the window is scaled
The relative position of the labels
Scale the distance between labels according to the size of c
Can prevent searchers from losing labels
You. Further, the relative positions of the labels displayed as shown in FIG.
Scale the window without changing the physical relationship
With the new area that can be created when the window is enlarged
Search by adding and presenting a new label 3001
It is also possible to increase the range of label selection by the user. In addition,
The area selected by dragging the bell display screen is enlarged and displayed.
It can also be shown. This allows searchers to
Only the gathering of bells can be seen enlarged. FIG.
Indicates that the retrieved document is also a label display image as the node 3101.
This is an example displayed on the screen. This screen will allow you to
Can be visually understood. Label display screen
Select the document displayed on the screen, for example,
By clicking, dragging, etc., the selected
Acquisition and display of documents, display area for selected documents
Expanding the area, redisplaying labels only for those documents
And so on. The number of documents to be displayed on the label display screen
Added to right click menu (Fig. 29)
It can be adjusted by such means as performing. Also selected
You can choose to display only the documents related to the labels you have created.
FIG. 32 shows documents and labels in the format of Table 3201.
Things. Select documents, labels and right click
By switching to this screen by operation, the relation between the document and the label
The searcher can find the document
It is easier to select and analyze documents. FIG.
The operation will be further described using 3 to 37. here
The same reference numerals indicate the same contents as those described above.
In FIG. 33, the documents extracted as a result of the search in the quality plan
This is displayed in area 1203. Search results for the above quality plan
Labels extracted based on fruits are displayed in area 1204
are doing. By changing the label selected here, area 12
03 is changed. This situation is shown in FIG.
Will be described. FIG. 34 shows an example in which more labels are selected.
As a result, the documents are narrowed down. Conversely reduce labels
Then, the documents narrowed down by the label increase. Figure
35 and 36 are partly enlarged to make label display easier to see
It is an example of a display. The area to be enlarged in FIG.
When a selection is made by a tag or the like, as shown in FIG.
Area is enlarged and displayed. If you want to revert,
Select an area that does not contain Other methods may be used. Ma
If you want to increase or decrease the number of displayed labels in the display state of
The number of labels in the label number selection area 3301
It becomes possible by doing. FIG. 37 shows the label number selection
This is a case where the number of labels 25 is specified in the rear 3301. Yo
More detailed and more appropriate narrowing down becomes possible. In the above description
Has been described by exemplifying words as labels. Word is content
It is good to simply show
Icons and marks can be used outside. The form of the above implementation
According to the state, similar information can be obtained without having to prepare search information.
It is easy to collect at once. Also collected
It is easy to select the desired information from the similar information group
is there.

According to the present invention, the operability of search processing is improved.

[Brief description of the drawings]

FIG. 1 is a system configuration diagram of a similar information search device.

FIG. 2 is a processing flowchart of the entire system.

FIG. 3 is a flowchart of an accumulation process.

FIG. 4 is a flowchart of a similarity calculation process.

FIG. 5 is an explanatory diagram of data of a storage information database.

FIG. 6 is an explanatory diagram of data of an index information database.

FIG. 7 is an explanatory diagram of data of a file information database.

FIG. 8 is a flowchart of a search process.

FIG. 9 is a flowchart of an output process.

FIG. 10 is an explanatory diagram of an output screen.

FIG. 11 is a processing flowchart on the client side.

FIG. 12 is an explanatory diagram of a search screen.

FIG. 13 is an explanatory diagram of a search.

FIG. 14 is an explanatory diagram of an output screen.

FIG. 15 is an explanatory diagram of an output screen.

FIG. 16 is an explanatory diagram of an output screen.

FIG. 17 is an explanatory diagram of an output screen.

FIG. 18 is an explanatory diagram of an output screen.

FIG. 19 is an explanatory diagram of an output screen.

FIG. 20 is an explanatory diagram of an information display screen.

FIG. 21 is an explanatory diagram of an information display screen.

FIG. 22 is an explanatory diagram of an information display screen.

FIG. 23 is an explanatory diagram of an information display screen.

FIG. 24 is an explanatory diagram of a label display screen.

FIG. 25 is an explanatory diagram of a label display screen.

FIG. 26 is an explanatory diagram of a label display screen.

FIG. 27 is an explanatory diagram of a label display screen.

FIG. 28 is an explanatory diagram of a label display screen.

FIG. 29 is an explanatory diagram of a label display screen.

FIG. 30 is an explanatory diagram of an output screen.

FIG. 31 is an explanatory diagram of an output screen.

FIG. 32 is an explanatory diagram of an output screen.

FIG. 33 is an explanatory diagram of an information display screen.

FIG. 34 is an explanatory diagram of an information display screen.

FIG. 35 is an explanatory diagram of an information display screen.

FIG. 36 is an explanatory diagram of an information display screen.

FIG. 37 is an explanatory diagram of an information display screen.

[Explanation of symbols]

101 server 102 Client 102 Index information database 103 accumulated information database 104 File Information Database 201 Information input means 202 Feature extraction means 203 Information specifying means 204 Similarity calculation means 205 Information storage means 209 Search key input means 210 Search key applicable information search means 211 Activity value giving means 212 Active value propagation means 213 Active state determination means 214 output means 215 Output device such as display 216 Display means 1403 Refined document title 1801 similarity search start button 2403 circular marker

   ────────────────────────────────────────────────── ─── Continuation of front page    (72) Inventor Toshiya Sato             2-16-10 Honcho, Aoba-ku, Sendai City, Miyagi Prefecture             Hitachi Tohoku Software Co., Ltd. F term (reference) 5B075 ND03 NK32 PP13 PP24 PQ13                       PR06

Claims (11)

[Claims] 1. A search device for searching information in a database based on input search information and outputting a search result, performing a search based on the search information. By extracting a plurality of labels having a predetermined relationship and outputting them, by selecting a specific label from the plurality of output labels, a search based on the selected specific label is included in the search information. An information search apparatus characterized in that a search result based on a search result is output. 2. A search device for searching information in a database based on input search information and outputting the search result. The search device includes an input device, an output device having a display, and a processing device. An operation screen for performing a search is displayed on the display; the operation screen has an area for inputting search information; and the processing is performed based on the search information input to the input area. The apparatus performs a search, and further, the processing apparatus extracts a plurality of labels having a predetermined relationship with the search information from the search information, and outputs the extracted labels to the display. If a label is further specified from among the plurality of labels, the search results associated with the specified label are selected, and the plurality of labels and the label And another search result retrieval apparatus of information, characterized in that so as to display together on the display screen of the display. 3. A database based on input search information.
Search device that searches for search information and outputs the search results.
The search device was equipped with an input device and a display.
Having an output device and a processing device; for performing a search
Display the operation screen on the display, and the operation screen
It has an area for entering search information.
Processing device based on the search information input to the input area
Performs a search, Further, the processing device performs the search on the search information.
Extract multiple labels that have a predetermined relationship with search information
Output the extracted labels to the above display, Specify more labels from the multiple labels output
Select the search result associated with the specified label.
Separated The above-mentioned labels are sorted and related to the labels.
Search results are displayed on the display screen of the above display.
As shown, The above search result is provided on the display screen of the display.
Displayed in the search result display area, and
Is the search results filtered based on the specified label
And both unfiltered search results
An information retrieval apparatus characterized in that: 4. The information retrieval apparatus according to claim 2, wherein the plurality of labels displayed on the display screen of the display are a designated label and a non-designated label. Characterized in that the information is displayed so as to be visually different. 5. The information retrieval apparatus according to claim 2, wherein the plurality of labels displayed on the display screen of the display are displayed in a predetermined area on the display screen, and the plurality of labels are displayed on the display screen. The display position of each label in is
An information retrieval apparatus characterized by being determined in relation to the relationship between each label and the retrieval result. 6. A computer program having a computer provided with an input device, an output device provided with a display, and a processing device having the following processing functions; an operation screen for performing a search is displayed on the display. And
The operation screen has an area for inputting search information, and the processing device performs a search based on the search information input to the input area. A plurality of labels having a predetermined relationship with the search information are extracted, the extracted plurality of labels are output to the display, and further labels are specified from among the plurality of output labels. The search results associated with the selected label are selected, and the plurality of labels and the selected search results associated with the specified label are displayed on the display screen of the display. 7. A computer program for causing a computer to have the following functions: a computer program having a function of searching for information in a database based on input search information and outputting a search result; And
Further, a search is performed based on the search information, a search result is output, a plurality of labels having a predetermined relationship with the search information are extracted and output, and the plurality of labels are selected. Thus, the search results based on the selected label are narrowed down and the results are output. 8. An information search method for searching a search target database using search information representing contents to be searched, comprising: a step of setting up an operation unit for search; and an operation screen for search. Displaying the search information in an operation information display area of the operation screen in accordance with the input of the search information representing the content to be searched and the display step; and a search process based on an instruction regarding execution of the search process based on the search information. And a step of outputting a search result and a plurality of labels for narrowing down the search result based on the execution of the search processing. The output of the search result and the search result narrowing label includes:
A search result display area and a search result narrowing label display area are provided on the screen so that they are displayed in the above-mentioned areas, respectively, and the label is specified by further specifying a label from among the displayed plural labels. An information search method, wherein a search result is selected in relation to the selected label, and the selected search result is displayed. 9. A display device according to claim 1, wherein a display area for a search result, a display area for a search result narrowing label, and an input area for a search key are provided on the screen and displayed in each area. Of the above labels,
An information search method characterized by displaying a search result related to a selected and specified label by specifying a search key in addition to selecting a specific label. 10. A search method for outputting a search result obtained by searching database information based on input search information, wherein the search result is output based on the search information in addition to outputting the search result based on the search information. Outputting a plurality of labels for narrowing down, further selecting the output plurality of labels, the search result is selected and output based on the selected labels, and is output. How to find the information you want. 11. A method for retrieving information according to claim 10, wherein when presenting the collected information as a search result, similarity is determined according to the number of labels that match the search information and the relationship between the plurality of labels. An information presentation method that makes it easy for a searcher to select desired information by visually classifying displays in which pieces of information are arranged in the vicinity.