JP2004054431A - Structured document generation device and structured document generation program - Google Patents

Abstract

A structured document including a semantic structure that can be easily recognized by a computer from a natural sentence is generated.
A meaning extraction unit (30) stores a concept dictionary (31) in which an attribute value extraction pattern and an attribute value output format are stored for each of various concepts, and performs pattern matching to determine an optimal item concept from a plurality of item concept dictionaries. A pattern matching unit 32 for selecting a dictionary, an attribute value extracting unit 33 for extracting attribute values using the selected item concept dictionary, and a shaping unit for shaping the extracted attribute values and corresponding tags into an XML document and outputting the XML document 34.
[Selection diagram] FIG.

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a structured document generation device and a structured document generation program, and more particularly to a structured document generation device and a structured document generation program suitable for generating data that can be directly used for retrieval.
[0002]
Problems to be solved by the prior art and the invention
Japanese Patent Application Laid-Open No. 2001-290801 discloses a structure documentation system, a structure documentation program, and a computer readable storage medium (hereinafter referred to as “conventional”) that input a text document and output a structured document using the document structure. Technology 1 ”).
[0003]
The prior art 1 aims at reconstructing a text document in order to display it on a Web page. In order to decompose the structure of the document, a document decomposition definition file is defined in advance, and the document definition is defined. The document is structured using the patterns described in the file. The document structure generated in this manner has a surface format such as a title, a creator, and a date.
[0004]
However, a part written in a natural sentence as a text is output as it is without any conversion, and is not in a format usable by a computer system. For this reason, the structured document output from the prior art 1 has a problem that only the surface information is used in the database, but specific contents such as the text written in natural sentences are not used in the database. Was.
[0005]
SUMMARY OF THE INVENTION The present invention has been proposed to solve the above-described problems, and has a structure and a structure of a structured document generating apparatus that generates a structured document having a semantic structure that can be easily recognized by a computer from a natural sentence. It is an object of the present invention to provide a document generation program.
[0006]
[Means for Solving the Problems]
In order to solve the above-mentioned problem, the invention according to claim 1 provides a structured document constituted by a sentence input unit for inputting a sentence including a natural sentence, and a semantic structure of the sentence input by the sentence input unit. And a structured document generating means for generating a document.
[0007]
Here, the invention described in claim 1 is configured by installing the invention described in claim 4 on a computer.
[0008]
That is, the invention according to claim 4 generates a structured document constituted by a sentence inputting step of inputting a sentence including a natural sentence into a computer and a semantic structure of the sentence input in the sentence inputting step. And a structured document generation program for executing a process including a structured document generation step.
[0009]
Therefore, according to the first and fourth aspects of the invention, based on a sentence including a natural sentence, a structured document constituted by the semantic structure of the sentence can be automatically generated.
[0010]
According to a second aspect of the present invention, in the first aspect of the present invention, the structured document generation unit includes a phrase representing the concept based on a concept of the text input by the text input unit; , And a structured sentence composed of a semantic structure having
[0011]
Here, the invention described in claim 2 is configured by installing the invention described in claim 5 on a computer.
[0012]
That is, in the invention according to claim 5, in the invention according to claim 4, the structured document generation step includes: a phrase representing the concept based on a concept of the sentence input in the sentence input step; And generating a structured sentence composed of a semantic structure having an attribute value explaining the phrase.
[0013]
Therefore, according to the second and fifth aspects of the present invention, based on the concept of a sentence, a main word representing the concept and an attribute value for explaining the word are extracted, and a semantic having the main word and the attribute value is extracted. A structured sentence composed of a structure can be generated.
[0014]
The invention according to claim 3 is the invention according to claim 2, wherein a phrase representing a concept, an attribute value explaining the phrase, a pattern in which the phrase and the attribute value are described in a sentence, Further comprising a concept dictionary storage unit for storing a concept dictionary representing the following. The structured document generation unit compares a sentence input by the sentence input unit with a concept dictionary stored in the concept dictionary storage unit. Thus, a structured document constituted by the semantic structure is generated.
[0015]
Here, the invention described in claim 3 is configured by installing the invention described in claim 6 on a computer.
[0016]
That is, according to a sixth aspect of the present invention, in the invention of the fifth aspect, the structured document generating step describes the sentence input in the sentence input step, a phrase representing a concept, and the phrase. A structured document constituted by the semantic structure is generated by comparing an attribute value with a concept dictionary representing a pattern in which the word and the attribute value are described in a sentence.
[0017]
Therefore, according to the third and sixth aspects of the present invention, a phrase and an attribute value are extracted from a document by using a concept dictionary in which a pattern in which the phrase and an attribute value are described in a sentence is represented in advance. , A structured document composed of a semantic structure can be generated.
[0018]
The invention according to claim 7, wherein phrase analysis is performed on a document including a natural sentence, and a phrase analysis unit that outputs a phrase-analyzed document in which words constituting the document are associated with parts of speech types, A concept for storing a concept dictionary having a part-of-speech type configuration pattern representing a configuration of a part-of-speech type of a word string including an attribute value to be extracted, and output format information for structuring and outputting the attribute value and the corresponding tag. The dictionary storage unit compares the part-of-speech type configuration pattern of the word string of the phrase-analyzed document with the part-of-speech type configuration pattern of the concept dictionary stored in the concept dictionary storage unit. Attribute value extracting means for extracting an attribute value from a word string of the corresponding word-parsed document; attribute values extracted by the attribute value extracting means; and a concept word stored in the concept dictionary storage means. Based of the output format information, and a, a document generation means for generating a structured document.
[0019]
The invention according to claim 7 is configured by installing the invention according to claim 9 on a computer.
[0020]
According to a ninth aspect of the present invention, there is provided a phrase analysis that executes a phrase analysis on a document including a natural sentence and outputs a phrase-analyzed document in which words constituting the document are associated with parts of speech. And comparing the part-of-speech type composition pattern of the word string of the word-parsed document with the part-of-speech type composition pattern representing the configuration of the word part type of the word string including the attribute value to be extracted. An attribute value extracting step of extracting an attribute value from a word string of the phrase-analyzed document corresponding to a configuration pattern, and output format information for structuring and outputting the extracted attribute value and a tag corresponding to the attribute value And a document generation step of generating a structured document based on the above.
[0021]
The phrase analysis unit performs phrase analysis on a document including a natural sentence. Here, the document containing a natural sentence may be a document consisting of only a natural sentence or a document of a predetermined format having a natural sentence and a predetermined tag. Then, the phrase analyzing unit associates each word constituting the document with the part of speech type, and outputs a word / phrase-analyzed document in which each word is associated with each part of speech type.
[0022]
The concept dictionary storage unit stores a concept dictionary having a part of speech type configuration pattern and output format information. Here, the part-of-speech type configuration pattern represents the configuration of the part-of-speech type of a word string including an attribute value to be extracted when an attribute value is to be extracted from a document. The output format information indicates the output format of the document, and is information that outputs the attribute values and the corresponding tags in a structured manner.
[0023]
The attribute value extraction unit compares the part-of-speech type configuration pattern of the word string of the phrase-analyzed document with the part-of-speech type configuration pattern of the concept dictionary, and extracts a word string corresponding to the part-of-speech type configuration pattern from the phrase-analyzed document. Find out. Then, the attribute value is extracted from the searched word string.
The document generating means generates a structured document in a predetermined format based on the extracted attribute values and the output format information of the concept dictionary. This structured document is a document having a format including an attribute value and a tag even if it is a natural sentence at first.
[0024]
Therefore, according to the seventh and ninth aspects of the invention, the part-of-speech type configuration pattern of the word string of the word-parsed document and the part-of-speech type representation of the word string including the attribute value to be extracted are included. By extracting attribute values by comparing with a configuration pattern, and structuring the extracted attribute values based on output format information to generate a structured document, a structured document in which meaning is extracted from a natural sentence is obtained. be able to.
[0025]
According to an eighth aspect of the present invention, in the invention according to the seventh aspect, the concept dictionary storage unit stores a concept dictionary provided for each concept, and the attribute value extracting unit includes the concept dictionary storage unit. From a plurality of concept dictionaries stored in the dictionary, a concept dictionary having a part-of-speech type configuration pattern that matches the part-of-speech type configuration pattern of the word-parsed document is extracted, and the attribute value is extracted using the selected concept dictionary. I do.
[0026]
The invention described in claim 8 is configured by installing the invention described in claim 10 on a computer.
[0027]
According to a tenth aspect of the present invention, in the ninth aspect of the present invention, the attribute value extracting step comprises: setting a part-of-speech type of the word-parsed document from a plurality of concept dictionaries provided for each concept. A concept dictionary having a part-of-speech type configuration pattern that matches the pattern is selected, and attribute values are extracted using the selected concept dictionary.
[0028]
The concept dictionary storage means stores each concept dictionary for each of various concepts. The attribute value extracting unit specifies the concept of the word-parsed document by selecting a concept dictionary having a part-of-speech type composition pattern that matches the part-of-speech type composition pattern of the word-parsed document from various concept dictionaries. . Then, the attribute value is extracted by using the selected concept dictionary to obtain the attribute value related to the specified concept.
[0029]
Therefore, according to the eighth and tenth aspects of the present invention, a concept dictionary having a part-of-speech type configuration pattern that matches the part-of-speech type configuration pattern of the word-parsed document is selected and selected from various concept dictionaries. By extracting attribute values using the concept dictionary, attribute values can be reliably extracted regardless of the concept of the document content.
[0030]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.
[0031]
FIG. 1 is a block diagram illustrating a configuration of a structured document generation device according to an embodiment of the present invention.
[0032]
The structured document generation device includes a keyboard 1 for inputting character information by key operation, a mouse 2 as a pointing device, an input / output port 3 for inputting and outputting information to and from the outside, and a structured document generation device. (Hard Disc Drive) 4 for storing an application program and the like for executing the arithmetic processing of the microcomputer 5, a microcomputer 5 for executing an arithmetic processing for creating a structured document, and an arithmetic result of the microcomputer 5, for example, a structured document (Liquid Crystal Display) 6 that displays
[0033]
The microcomputer 5 includes a RAM (Random Access Memory) as a work area for data, a ROM (Read Only Memory) storing a predetermined control program, a CPU (Central Processing Unit) for executing arithmetic processing, and the like. I have. The microcomputer 5 creates a structured document based on a text document input from the keyboard 1 or a text document input from the outside via the input / output port 3. Note that a structured document may be created from a text document already recorded on a recording medium (not shown).
[0034]
FIG. 2 is a block diagram showing a functional configuration of the microcomputer 5. The microcomputer 5 includes an XML document creation unit 10 that creates an XML document from a text document, a word analysis unit 20 that performs word / phrase analysis, and a meaning extraction unit 30 that extracts a meaning and creates a structured document. I have.
[0035]
The XML document creation unit 10 creates an XML document using a text document of a natural sentence and information such as the title and creation date and time of the text document, and supplies the created XML document to the phrase analysis unit 20. The text document may be a document generated by operating the keyboard 1, a document input from the input / output port 3, or a document stored in the HDD 4 or another storage medium in advance.
[0036]
FIG. 3 is a block diagram illustrating a functional configuration of the phrase analyzing unit 20.
[0037]
The phrase analyzing unit 20 includes a phrase dictionary 21 indicating a correspondence relationship between a word and a part of speech type, a content character string separating unit 22 for separating a structure tag of an XML document from a content character string as a natural sentence, , A specific word part-of-speech determining unit 24 that determines the part of speech of a specific word, a morphological analysis unit 25 that performs a morphological analysis of a content character string, and substitute words for specific words And a compound word / phrase generation unit 27 that generates compound words and phrases from a plurality of words, and a part-of-speech tag attaching unit 28 that attaches a part-of-speech tag to words and compound words.
[0038]
FIG. 4 is a diagram showing the configuration of the phrase dictionary 21. The phrase dictionary 21 includes a dictionary relating to words and a dictionary relating to words (compound words and phrases).
[0039]
The phrase dictionary 21 is composed of words, parts of speech, and alternative words with respect to words. The alternative word will be described later in detail, but is used only in the case of a specific word.
[0040]
According to FIG. 4, for example, for the word “amplification”, the part of speech type is “noun-sa transformation connection”. This indicates that “amplification” is a noun and is connected to a variance verb. For the word “problem”, the part of speech type is “noun-nai adjective stem”. This indicates that "problem" is a noun and becomes the stem of the adjective "... no." Further, for the word "time", the part of speech is "noun-suffix-adverb possible". This indicates that "time" is a noun, becomes a suffix of "... time", and can be used as an adverb. For the word "ga", the part of speech type is "particle-case particle-general". This indicates that "ga" is a general case particle. As described above, the part of speech type indicates not only the type of part of speech of the corresponding word but also the attribute of the corresponding word.
[0041]
Here, the “word” includes not only general words such as nouns and particles, but also specific words indicating words newly defined by the phrase dictionary 21.
[0042]
For example, “2S [AD] [0-9] +” shown in FIG. 4 starts with 2S, is followed by one character from A to D, and ends with one or more characters. Represents a word. The word corresponding to the above condition indicates the product number of the transistor and is not in a general dictionary. Thus, the phrase dictionary 21 defines words that meet the above conditions as specific words.
[0043]
Here, for the word (specific word) “2S [AD] [0-9] +”, the part-of-speech type is “noun-proper noun-identifier-part number-transistor” and the alternative word is “alternative-transistor part number”. ". This means that the specific word is a proper noun, an identifier, and a product number of a transistor. Further, since the specific word has no meaning in itself, it indicates that there is a “transistor part number” as an alternative word.
[0044]
The phrase dictionary 21 is composed of “phrase strings” and “part-of-speech types” in terms of parts related to phrases (compound words and phrases), and also functions as a dictionary for creating compound words and phrases.
[0045]
"Phrase string" represents a pattern to be formed as a compound word or phrase. For example, according to FIG. 4, “(noun and not (noun-suffix-adverb) or symbol) * [2, ∞]” is a noun (except for noun-suffix-adverb) or a symbol This indicates that the pattern is two or more continuous patterns, and the “part of speech” of this pattern is a noun phrase. That is, a pattern of a phrase string corresponding to such a condition is treated as a noun phrase.
[0046]
FIGS. 5 and 6 are diagrams illustrating documents analyzed by the phrase analyzing unit 20. FIG. Hereinafter, a case will be described in which the XML document shown in FIG.
[0047]
When supplied with the XML document shown in FIG. 5A from the XML document creation unit 10, the content character string separation unit 22 separates the structure tag of the XML document from the content character string that is a natural sentence. Then, the content character string separating unit 22 supplies the structure tag of the XML document shown in FIG. 5B to the part-of-speech tag attaching unit 28, and converts the content character string of the XML document shown in FIG. This is supplied to the replacement unit 23.
[0048]
The specific word separation / replacement unit 23 compares the content character string supplied from the content character string separation unit 22 with the phrase dictionary 21, replaces the specific word included in the content character string with an alternative word, and replaces the specific word. To separate. Then, the separated specific word is supplied to the specific word part-of-speech determination unit 24, and the content character string obtained by replacing the specific word with the alternative word is supplied to the morphological analysis unit 25.
[0049]
Specifically, the specific word separation / replacement unit 23 replaces the specific word “2SC7777” included in the content character string with “substitute-transistor part number” based on the phrase dictionary 21. Then, the specific word “2SC7777” shown in FIG. 5D is separated and supplied to the specific word part-of-speech determination unit 24, and the replaced content character string “(alternative-transistor part number) shown in FIG. Is generated during class amplification "to the morphological analysis unit 25.
[0050]
The specific word part-of-speech determination unit 24 determines the part of speech of the specific word based on the word dictionary 21 and supplies the specific word and part-of-speech type information to the alternative word replacement unit 26. Specifically, the part of speech of “2SC7777” is determined, and as shown in FIG. 5 (F), “2SC7777−noun−proper noun−identifier−part number−transistor” which is the specific word and part of speech type information is replaced with a substitute word. To the unit 26.
[0051]
The morphological analysis unit 25 performs morphological analysis on the content character string supplied from the specific word separation / substitution unit 23 while referring to the phrase dictionary 21. Specifically, the content character string is decomposed into individual words, and each decomposed word (including a specific word) is associated with a part of speech type. Then, the morphological analysis unit 25 supplies each word and the corresponding part of speech type shown in FIG. In FIG. 5 (G), the words “time”, “ni”, “fever”, “ga”, and “problem” are omitted, but morphological analysis is similarly performed on each of these words. Then, each word and the corresponding part of speech type are supplied to the alternative word replacement unit 26.
[0052]
The alternative word replacement unit 26 selects an alternative word from each morphologically analyzed word based on the information supplied from the specific word part of speech determining unit 24 and the morphological analysis unit 25, and specifies the selected alternative word as an original. Replace with word. Specifically, “substitute-transistor” is replaced with “2SC7777”. Then, each word and the corresponding part of speech type shown in FIG. 6A are supplied to the compound word / phrase generation unit 27. In FIG. 6A, the words “hour”, “ni”, “fever”, “ga”, and “problem” are not shown. These words are also omitted in FIG. 6B described later.
[0053]
The compound word / phrase generation unit 27 determines whether there is a substring of the word corresponding to the compound word / phrase in the phrase dictionary 21 in the continuous words supplied from the alternative word replacement unit 26, and If there is a subsequence, all the corresponding subsequences are replaced with compound words / phrases.
[0054]
Specifically, the partial strings “A”, “class”, and “amplification” of the word correspond to the phrase string pattern of the noun phrase in the phrase dictionary 21 shown in FIG. Therefore, the compound word / phrase generation unit 27 generates a noun phrase “A class amplification” from the word substrings “A”, “class”, and “amplification”, and generates each word shown in FIG. And the corresponding part of speech type are supplied to the part of speech tag assigning unit 28. When there is no substring of the word corresponding to the compound word / phrase in the word / phrase dictionary 21, the compound word / phrase generation unit 27 supplies the information supplied from the substitute word replacement unit 26 to the part-of-speech tag attaching unit 28 as it is. I do.
[0055]
The part-of-speech tag assigning section 28 assigns a part-of-speech tag indicating each part-of-speech type to each word (including compound words / phrases) supplied from the compound word / phrase generation section 27. Then, each word and each part of speech tag are embedded as elements of the structure tag <content> supplied from the content character string separating unit 22. As a result, as shown in FIG. 6C, the part-of-speech tag assigning unit 28 generates and outputs an XML document that has undergone term analysis.
[0056]
As described above, the phrase analyzing unit 20 separates the structure tag and the content character string (natural sentence) from the general XML document, performs a morphological analysis on the content character string, and adds a part of speech to each word of the content character string. A process of adding a tag and embedding each word and the part of speech tag in the original XML document is performed. That is, the phrase analysis unit 20 performs a morphological analysis on a natural sentence and adds a part-of-speech tag to each word constituting the natural sentence, thereby clarifying the document structure of the natural sentence and extracting the meaning of the document. Can be output.
[0057]
In addition, the phrase analyzing unit 20 registers the newly defined specific word and its alternative word in the phrase dictionary 21, so that the morphological analysis uses the alternative word instead of the specific word, and after the morphological analysis, By replacing a substitute word with the original specific word, morphological analysis can be accurately performed even for technical terms and technical terms that are not in a general dictionary.
[0058]
It should be noted that the phrase dictionary 21 is not limited to the configuration shown in FIG. 4, and can store other words and part of speech. Further, the phrase dictionary 21 can store not only a pattern for creating a noun phrase from a plurality of nouns, but also a pattern for generating other compound words / phrases such as adjective phrases and adverb phrases. Of course.
[0059]
FIG. 7 is a block diagram illustrating a functional configuration of the meaning extracting unit 30. The meaning extracting unit 30 extracts a meaning from the XML document that has been subjected to the phrase analysis, and generates an XML document in which the extracted meaning is structured.
[0060]
The meaning extracting unit 30 selects an optimal item concept dictionary from a plurality of item concept dictionaries by performing pattern matching with a concept dictionary 31 storing an attribute value extraction pattern and an attribute value output format for each of various concepts. A pattern matching unit 32, an attribute value extracting unit 33 for extracting attribute values using the selected item concept dictionary, and a shaping unit 34 for shaping the extracted attribute values and corresponding tags into an XML document and outputting the XML document. ing.
[0061]
FIG. 8 is a block diagram showing the configuration of the concept dictionary 31. The concept dictionary is composed of a plurality of item concept dictionaries configured for each of various concepts (items) such as “current problem” and “solution”. Here, a description will be given using the “current problem” as an example.
[0062]
The item concept dictionary includes “items” representing the items of the item concept dictionary, “synonyms” representing synonyms of the “items”, “attributes describing the items”, and “XML output” representing the XML output format of the extracted attribute values. Format ".
[0063]
The “attribute that describes the item” includes an “attribute name” and a “tag pattern”. The “attribute name” indicates what attribute value is to be extracted. The “tag pattern” indicates a configuration pattern of each part-of-speech tag of the word string including the attribute value, and a position of the attribute value included in the configuration pattern.
[0064]
Here, “noun *” indicates an arbitrary noun or a noun phrase. Further, a pair of “@” and an attribute name following the symbol indicates that the character face of the word corresponding to the position of the tag pattern becomes the value of the attribute.
[0065]
For example, in FIG. 8, “question” indicates that an attribute value that describes what kind of “problem” is extracted from the XML document that has undergone phrase analysis. The “tag pattern” of “problem” indicates a part-of-speech tag configuration pattern of a word string including an attribute value that describes “problem”.
[0066]
Note that the “tag pattern” of the “problem” means that the part of speech tag of the first word is a noun or a noun phrase, and the part of speech tag of the next word may be arbitrary, but that the next word is “ga”. Is shown. Here, since the element of the first part-of-speech tag includes “＄ problem”, the word corresponding to the first part-of-speech tag becomes the value (attribute value) of the attribute name “problem”.
[0067]
“Target” indicates that an attribute value that describes what “target” is extracted from the XML document that has undergone the phrase analysis. In the “tag pattern” of “target”, the part of speech tag of the first word is a noun or a noun phrase, and the part of speech tag of the next word may be arbitrary, but the next word (word string) is “in”. It is shown that. Also, since the element of the first part of speech tag includes “＄ target”, the word corresponding to the first part of speech tag becomes the value (attribute value) of the attribute name “target”.
[0068]
Further, “problem situation” indicates that an attribute value describing what “problem situation” is extracted from the XML document that has been subjected to the phrase analysis. In the “tag pattern” of the “problem situation”, the part-of-speech tag of the first word is a noun or a noun phrase, and the part-of-speech tag of the next word is “noun-suffix-adverbable” (a suffix that can be used as an adverb). And the part of speech tag of the last word may be arbitrary, but indicates that the last word is “ni”. In addition, since the element of the first part of speech tag includes “＄ problem occurrence situation”, the word corresponding to the first part of speech tag becomes the value (attribute value) of the attribute name “problem occurrence situation”.
[0069]
The “XML output format” indicates that an XML document in which each of the extracted attribute values has been replaced with “@attribute name” is output.
[0070]
For example, the first line of the “XML output format” in FIG. 8 indicates that the output is performed by replacing the attribute value of “problem”. The <target> and <when a problem occurs> incorporated as elements of the <current problem> similarly indicate that corresponding elements are formed and output together with the structure tag.
[0071]
Then, the pattern matching unit 32, the attribute value extraction unit 33, and the shaping unit 34 execute the following processing using the concept dictionary 31 configured as described above.
[0072]
The pattern matching unit 32 compares the XML document supplied with the phrase analysis supplied from the phrase analysis unit 20 with each item concept dictionary of the concept dictionary 31, and selects an item concept dictionary corresponding to the XML document. Specifically, from each item concept dictionary, an item concept dictionary in which all tag patterns completely match the configuration patterns of the part-of-speech tags of the document is selected. Then, the pattern matching unit 32 supplies the “items” of the selected item concept dictionary to the attribute value extraction unit 33 and supplies the XML document of the part matched with the item concept dictionary to the attribute value extraction unit 33.
[0073]
For example, here, the configuration pattern of the part-of-speech tag of the document shown in FIG. 6 (C) matches all the “tag patterns” of the item concept dictionary “current problem” shown in FIG. Therefore, the pattern matching unit 32 supplies the item name “current problem” and the element of <content> in FIG. 6C to the attribute value extraction unit 33.
[0074]
If there are a plurality of item concept dictionaries in which the configuration patterns of the part-of-speech tags of the document supplied from the phrase analysis unit 20 and all the tag patterns completely match, the condition of the tag pattern of the item concept dictionary must be further restricted. Just fine. For example, instead of “noun *” representing an arbitrary noun or a noun phrase, “noun-proper noun” representing a proper noun may be used, and other conditions may be limited.
[0075]
The attribute value extracting unit 33 reads out the item concept dictionary indicated by the “item” supplied from the pattern matching unit 32 from the concept dictionary 31 and converts the document supplied from the pattern matching unit 32 into a tag pattern of the item concept dictionary. Find the word string. Then, a word corresponding to the "@attribute value" of the tag pattern is extracted from the word string as an attribute value.
[0076]
Specifically, the attribute value extracting unit 33 searches for the word strings respectively corresponding to the three tag patterns shown in FIG. 8 from the elements of <contents> in FIG. 6C. Then, the words “fever”, “2SC7777”, and “class A amplification” corresponding to the “$ attribute value” of each tag pattern are extracted from each of the searched word strings, and these are extracted as attribute values by the shaping unit 34. Supply.
[0077]
The shaping section 34 shapes each attribute value supplied from the attribute value extracting section 33 in accordance with the output format of the concept dictionary 31, generates an XML document, and outputs this XML document to the outside.
[0078]
Specifically, the attribute value “fever” describing “problem”, the attribute value “2SC7777” describing “object”, and the attribute value “A-class amplification” describing “problem situation” are output as “XML output”. The element is formed by substituting the corresponding "@attribute value" of the "format", and the formed element and the structure tag are output in the XML document format.
[0079]
FIG. 9 is a diagram illustrating a meaning-extracted XML document output from the meaning extracting unit 30. Conventionally, a structured document can be generated from a predetermined document. However, when a natural sentence is included in the predetermined document, the natural sentence is output as it is. That is, the conventional structured document includes a natural sentence in the element of the structure tag. On the other hand, the meaning-extracted XML document shown in FIG. 9 is a document in which the meaning of the natural sentence included in the element of the structure tag is extracted, and the extracted meaning is structured. .
[0080]
As described above, the structured document generation device according to the present embodiment performs a morphological analysis on a natural sentence included in an XML document to assign a part of speech type to each word, The attribute value is extracted from the matching word string by comparing the word string including the value with the tag pattern indicating the part of speech type configuration. Then, by outputting the extracted attribute values in accordance with a predetermined XML output format, it is possible to generate an XML document in which a natural sentence is structured.
[0081]
In particular, the structured document generation apparatus prepares in advance a tag pattern for extracting each attribute value for each of various concepts (items). The concept dictionary can be specified, and further, attribute values for explaining the concept can be extracted from the natural sentence.
[0082]
In addition, the structured document generation device prepares, in advance, for each of various concepts (items), an XML output format of an attribute value necessary to explain the concept and a structure tag. An XML document in which the extracted attribute values are structured can be easily generated simply by outputting in accordance with the XML output format.
[0083]
As shown in FIG. 9, the XML document generated by the structured document generation device is a structured document in which the concept of a natural sentence is described by a structured tag and an attribute value. It is a document that is easy to respond to inquiries. Hereinafter, an application example of the XML document will be described.
[0084]
(Application Example 1)
A case will be described in which a database in which "product weight is reduced" is searched from a database in which a large number of documents are stored.
[0085]
10A is a diagram showing a conventional natural sentence document stored in the database, FIG. 10B is a diagram showing a conventional search result, and FIG. 10C is an XML document of the present application stored in the database. FIG. The XML document of the present application refers to an XML document generated by the above-described structured document generation device from the conventional natural sentence document shown in FIG. That is, the documents of (A) and (C) have the same content, suggesting that the weight of the product increases.
[0086]
Here, when a document describing that "product weight is reduced" is searched by a general method, "product", "weight", and "reduction" are usually used as search keywords. When conventional documents are stored in the database, conventional documents containing the words “product”, “weight”, and “reduction” are erroneously searched as shown in FIG.
[0087]
On the other hand, when the XML document of the present application is stored in the database, the structure tags may be sequentially followed. Here, a structure tag related to “product weight” is searched first, and a structure tag indicating whether the weight of the product has been increased or decreased is searched from the elements of the structure tag. Specifically, first, <product weight> is searched, and <direction> representing a change in weight is searched from among the elements. In the XML document of the present application, since the element of <direction> is “increased”, it is not erroneously searched.
[0088]
Therefore, the structured document generation device can generate an XML document that does not cause an erroneous search from a document that is erroneously searched in a normal keyword search.
[0089]
(Application 2)
A case will be described in which an inquiry "What is the current countermeasure for processing defects?"
[0090]
FIG. 11A is a diagram showing a document of a conventional natural sentence stored in a database, and FIG. 11B is a diagram showing an XML document of the present application stored in the database. The documents in (A) and (B) have the same contents.
[0091]
If a conventional document is stored in the database, no response can be made even if an inquiry "What is the current countermeasure for processing defects?"
[0092]
On the other hand, when the XML document of the present application is stored in the database, the structure tag related to the inquiry may be followed. Here, first, a structure tag related to “measures” is searched, and elements of the structure tag (lower-order structure tags and their elements) are extracted. Then, by combining the extracted structure tags and elements, it is possible to reply to the above inquiry that “the thickness has been increased by 2 mm”.
[0093]
Therefore, the structured document generation device can generate an XML document that can easily respond to an inquiry by extracting and structuring the meaning of a natural sentence.
[0094]
The present invention is not limited to the above-described embodiment, and various design changes can be made within the scope described in the claims.
[0095]
For example, the concept dictionary 31 is not limited to the configuration shown in FIG. In the present embodiment, a case where one item concept dictionary has three tag patterns has been described as an example. However, the same number of tag patterns as the number of attribute values to be extracted can be provided.
[0096]
Further, in the present embodiment, an XML document has been described as an example, but an SGML document may be used, for example. At this time, the “XML output format” of the item concept dictionary may be changed to the “SGML output format”. The document format is not particularly limited in consideration of creating a structured document from natural sentences.
[0097]
【The invention's effect】
The structured document generation device and the structured document generation program according to the present invention can automatically generate a structured document having a semantic structure of a sentence based on a sentence including a natural sentence.
[0098]
In addition, the structured document generation device and the structured document generation program according to the present invention provide a configuration of a part-of-speech type of a word string of a word-parsed document and a configuration of a part-of-speech type of a word string including an attribute value to be extracted. By extracting attribute values by comparing them with the part-of-speech type composition pattern representing the structure, and generating a structured document by structuring the extracted attribute values based on the output format information, a structure in which the meaning is extracted from the natural sentence Document can be obtained.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of a structured document generation device according to an embodiment of the present invention.
FIG. 2 is a block diagram showing a functional configuration of a microcomputer.
FIG. 3 is a block diagram showing a functional configuration of a phrase analyzing unit.
FIG. 4 is a diagram showing a configuration of a phrase dictionary.
FIG. 5 is a diagram illustrating a document being analyzed by a phrase analysis unit.
FIG. 6 is a diagram illustrating a document that has been analyzed by a phrase analysis unit.
FIG. 7 is a block diagram illustrating a functional configuration of a meaning extracting unit.
FIG. 8 is a block diagram showing a configuration of a concept dictionary.
FIG. 9 is a diagram illustrating a meaning-extracted XML document output from a meaning extraction unit.
10A is a diagram showing a conventional document of a natural sentence stored in a database, FIG. 10B is a diagram showing a conventional search result, and FIG. 10C is an XML document of the present application stored in the database FIG.
11A is a diagram showing a document of a conventional natural sentence stored in a database, and FIG. 11B is a diagram showing an XML document of the present application stored in the database.
[Explanation of symbols]
20 Phrase analyzer
21 Word Dictionary
22 Content string separator
23 Specific word separation / replacement unit
24 Specific word part of speech decision unit
25 Morphological analyzer
26 Alternative word substitution unit
27 Compound word / phrase generator
28 Part-of-speech tag assigning unit
30 Meaning extractor
31 Concept Dictionary
32 pattern matching unit
33 Attribute value extractor
34 Shaping Department

Claims (10)

A sentence input means for inputting a sentence including a natural sentence,
Structured document generating means for generating a structured document constituted by the semantic structure of the text input by the text input means,
Structured document generation device provided with. The structured document generating unit is configured to have a semantic structure including a phrase representing the concept and an attribute value describing the phrase based on the concept of the sentence input by the text input unit. 2. The structured document generation device according to claim 1, wherein the structured document generation device generates a sentence. A concept dictionary storing means for storing a concept dictionary representing a phrase representing a concept, an attribute value explaining the phrase, and a pattern in which the phrase and the attribute value are described in a sentence,
The structured document generation unit generates a structured document constituted by the semantic structure by comparing a sentence input by the sentence input unit with a concept dictionary stored in the concept dictionary storage unit. The structured document generation device according to claim 2. On the computer,
A sentence input process of inputting a sentence including a natural sentence,
A structured document generating step of generating a structured document constituted by the semantic structure of the sentence input in the sentence input step,
A structured document generation program for executing a process including The structured document generation step is a structuring that is configured by a semantic structure having a phrase representing the concept and an attribute value that describes the phrase based on the concept of the text input in the text input step. The structured document generation program according to claim 4, which generates a sentence. The structured document generating step includes:
A sentence input in the sentence input step,
A concept dictionary representing a phrase representing a concept, an attribute value explaining the phrase, and a pattern in which the phrase and the attribute value are described in a sentence,
The structured document generation program according to claim 5, wherein a structured document configured by the semantic structure is generated by collating the structured document. A phrase analysis unit that performs phrase analysis on a document including a natural sentence and outputs a phrase-analyzed document in which the words constituting the document and the parts of speech are associated with each other;
A concept for storing a concept dictionary having a part-of-speech type configuration pattern representing a configuration of a part-of-speech type of a word string including an attribute value to be extracted, and output format information for structuring and outputting the attribute value and the corresponding tag. Dictionary storage means;
The phrase analysis corresponding to the part-of-speech type configuration pattern is performed by comparing the configuration pattern of the part-of-speech type of the word string of the phrase-analyzed document with the part-of-speech type configuration pattern of the concept dictionary stored in the concept dictionary storage unit. Attribute value extracting means for extracting an attribute value from a word string of a completed document;
Document generation means for generating a structured document based on the attribute value extracted by the attribute value extraction means and the output format information of the concept dictionary stored in the concept dictionary storage means,
Structured document generation device provided with. The concept dictionary storage means stores a concept dictionary provided for each concept,
The attribute value extracting unit selects a concept dictionary having a part-of-speech type configuration pattern that matches a part-of-speech type configuration pattern of the word-parsed document from a plurality of concept dictionaries stored in the concept dictionary storage unit. 8. The structured document generation device according to claim 7, wherein attribute values are extracted using the selected concept dictionary. On the computer,
A phrase analysis step of performing phrase analysis on a document including a natural sentence, and outputting a phrase-analyzed document in which words constituting the document are associated with part-of-speech types;
The part-of-speech type configuration pattern of the word string of the phrase-analyzed document is compared with the part-of-speech type configuration pattern representing the configuration of the part-of-speech type of the word string including the attribute value to be extracted. An attribute value extracting step of extracting an attribute value from a word string of the corresponding word-parsed document;
A document generation step of generating a structured document based on the extracted attribute values and output format information for structuring and outputting the tags corresponding to the attribute values,
A structured document generation program for executing a process including The attribute value extracting step selects and selects, from a plurality of concept dictionaries provided for each concept, a concept dictionary having a part-of-speech type configuration pattern that matches the part-of-speech type configuration pattern of the word-parsed document. The structured document generation program according to claim 9, wherein attribute values are extracted using a concept dictionary.

Images