JP2022002034A - Extraction method, extraction program, and extraction device - Google Patents

Abstract

To provide an extraction method, an extraction program, and an extraction device that appropriately recognize correspondence between sets detected from text.SOLUTION: When detecting a first set and a second set from text, a server 2 of a knowledge extraction system refers to first order information, specifies a plurality of first elements included in the first set having at least one part not described in the text, and extracts correspondence in which each of the plurality of first elements is associated with each of a plurality of second elements included in the second set, according to the order in the plurality of first elements.SELECTED DRAWING: Figure 3

Description

The present invention relates to an extraction method, an extraction program, and an extraction device.

Extracting Knowledge from Text In a knowledge extraction process, an extraction device such as a computer, for example, a server, may extract knowledge from a text having a plurality of elements such as parallel phrases.

For example, in the text (text) "domestic or foreign, as stipulated in Article 2, Paragraph 1, Item 1 or 2 of the Income Tax Act, respectively", "domestic" and "foreign" are defined. Is a parallel clause.

In order to correctly extract knowledge from this sentence, recognition of the structure of parallel phrases, parallel mapping, recognition of meta expressions as parallel expressions that are unnecessary when expanding, etc. are performed.

For example, in the knowledge extraction process, the extraction device expands the above sentence into the following two sentences, and then performs an appropriate knowledge representation process.
"The country specified in Article 2, Paragraph 1, Item 1 of the Domestic Income Tax Act."
"Foreign countries referred to in Article 2, Paragraph 1, Item 2 of the Foreign Income Tax Act."

Japanese Unexamined Patent Publication No. 2019-079321 Japanese Unexamined Patent Publication No. 2001-084250 JP-A-2009-09503

Hiroki Matsuyama, Kiyoaki Shirai, Akira Shimazu, "Parallel Structure Analysis for Legal Documents", Proceedings of the 18th Annual Meeting of the Association for Natural Language Processing, pp. 975-978, 2012

However, if the text contains a range expression, it may be difficult to recognize the correspondence due to the expansion of the above-mentioned items.

For example, in the following example sentence, there are 14 parallel elements as in the following "corporate taxation trust, ... securities". On the other hand, the articles corresponding to these are "Income Tax Law, Article 2, Paragraph 1, Items 8-3 to 13, 15 to 15-5, or 17" depending on the scope expression. There are three parallel elements.
"Corporate taxation trusts, permanent facilities, public and corporate bonds, deposits and savings, joint investment trusts, loan trusts, investment trusts, securities investment trusts, public and corporate bond investment trusts, public and corporate bond investment trusts, publicly offered public and corporate bond investment trusts, special purpose trusts, specific beneficiaries Securities issuance trust or securities Corporate taxation trust prescribed in Article 2, Paragraph 1, Items 8-3 to 13, 15-5-5 or 17 of the Income Tax Law, respectively. , Permanent facilities, public and corporate bonds, deposits and savings, joint investment trusts, loan trusts, investment trusts, securities investment trusts, public and corporate bond investment trusts, public and corporate bond investment trusts, publicly offered public and corporate bonds, investment trusts, special purpose trusts, specified beneficiary certificate issuance trusts Refers to securities. "

In this way, when the extraction device recognizes the correspondence between the sets detected from the text, if at least one of the elements in the set is not described in the text by a range expression or the like, the number of elements between the sets matches. However, it may be difficult to recognize the correspondence.

In one aspect, one of the objects of the present invention is to be able to appropriately recognize the correspondence between the sets detected from the text.

In one aspect, in the extraction method, the computer may perform the following processing. When the process detects the first set and the second set from the text, the process refers to the first order information, and at least one part is included in the first set not described in the text. A plurality of elements of 1 may be specified. Further, in the process, each of the first plurality of elements is associated with each of the second plurality of elements included in the second set according to the order in the first plurality of elements. The correspondence may be extracted.

In one aspect, the invention can adequately recognize the correspondence between sets detected from the text.

It is a block diagram which shows the functional structure example of the knowledge extraction system which concerns on one Embodiment. It is a block diagram which shows the hardware configuration example of the computer which realizes the function of a server. It is a figure for demonstrating an example of a server process. It is a flowchart explaining the operation example of the element expansion processing control part. It is a figure which shows an example of the order knowledge DB. It is a figure for demonstrating an example of expansion processing of a range element based on order knowledge. It is a figure which shows an example of the order knowledge including a demonstrative word.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the embodiments described below are merely examples, and there is no intention of excluding the application of various modifications or techniques not specified below. For example, the present embodiment can be variously modified and implemented without departing from the spirit of the present embodiment. In the drawings used in the following description, the parts with the same reference numerals represent the same or similar parts unless otherwise specified.

[1] Embodiment [1-1] Description of Knowledge Extraction System First, the knowledge extraction system will be briefly described. The knowledge extraction system is, for example, a system that analyzes sentences and extracts knowledge from the analysis results. The text analysis may include the parallel structure analysis described above. The extracted knowledge may be accumulated in a knowledge base (KB) such as a knowledge graph (KG).

(Comparative example)
Here, a comparative example of parallel structural analysis for legal texts will be described. In the comparative example, the property that the phrases in parallel relation are similar to each other is used as a clue, and the similarity of the phrase sequence and the probability model are used. As a comparative example, for example, the method described in Non-Patent Document 1 described above may be adopted.

In the comparative example, regarding the hierarchical parallel structure, the extraction device prioritizes the parallel structure lower than the upper parallel structure to the parallel key, and sequentially detects the parallel structure in the order of priority. Further, when the extraction device determines the range of the parallel structure, the boundary of the parallel structure is not the boundary of the parallel structure analyzed preferentially, and the boundary of the parallel structure is the boundary of the candidate of the posterior phrase and the anterior phrase. We use the heuristic that the parallel structure analyzed when calculating the degree is used.

In the above-mentioned comparative example, language processing (analysis) of parallel sentences is mainly performed, and knowledge extraction is not performed. Further, in the above-mentioned comparative example, knowledge of typical order such as numbers and alphabets is hard-coded as a recognition method of range expression.

Next, the extraction device according to the comparative example responds to the text (sentence) that "means domestic or foreign as stipulated in Article 2, Paragraph 1, Item 1 or 2 of the Income Tax Act, respectively, domestically or internationally." A case of performing parallel structure analysis will be described.

For example, the extractor independently recognizes each parallel phrase. In the above example, as in the following two definition provisions, the definition texts of "domestic" and "overseas" are "domestic as stipulated in Article 2, Paragraph 1, Item 1 or 2 of the Income Tax Act, respectively. Or it means foreign countries. "
"Domestic or foreign countries as stipulated in Article 2, Paragraph 1, Item 1 or 2 of the Income Tax Act, respectively."
"Foreign countries means domestic or foreign countries as stipulated in Article 2, Paragraph 1, Item 1 or 2 of the Income Tax Act, respectively."

The extractor then associates the components of the individual parallel phrases in sequence. For example, the extraction device extracts the nth element of the parallel phrase (n is an integer of 1 or more and N or less. N is the number of parallel phrases; in this case, N = 2) as follows.
1st: "Domestic, respectively, refers to the domestic stipulated in Article 2, Paragraph 1, Item 1 of the Income Tax Act."
Second: "Foreign countries refer to the foreign countries stipulated in Article 2, Paragraph 1, Item 2 of the Income Tax Act."

Then, the extraction device expands the ordered set by hard coding. As an example of such expansion, in the case of an ordered set in which "A to D are 3, 5, 7, 8 respectively", the extraction device expands "A to D" to "A, B, C,". D is 3, 5, 7, 8 respectively. " By such an expansion method, the following two sentences are expanded in the above example.
"The country specified in Article 2, Paragraph 1, Item 1 of the Domestic Income Tax Act."
"Foreign countries referred to in Article 2, Paragraph 1, Item 2 of the Foreign Income Tax Act."

However, in the parallel structure analysis described above, only superficial language processing is performed, and the contents of the components are not taken into consideration. For example, "each issue from a to d" or "October to March". It is not possible to support range expressions that include multiple elements, such as "each month up to". As described above, when the elements in the set are not described in the text by the range expression or the like, the number of elements between the sets may not match and it may be difficult to recognize the correspondence.

Therefore, in one embodiment, when the extraction device recognizes the correspondence between the sets detected from the text, even if at least one of the elements in the set is not described in the text by a range expression or the like, the set is set. A method for appropriately recognizing the correspondence between them will be explained.

For example, in the method according to one embodiment described later, a correspondence relationship between a plurality of expressions is extracted by allocating an entity to a plurality of element expressions using order knowledge.

[1-2] Configuration Example of Knowledge Extraction System FIG. 1 is a block diagram showing a functional configuration example of the knowledge extraction system 1 according to an embodiment. The knowledge extraction system 1 is an example of a system for extracting knowledge from text, and as shown in FIG. 1, an example includes a server 2 and one or more terminals 3 (one in the example of FIG. 1). good. A knowledge base such as a knowledge graph (KG) may be connected to the server 2.

The terminal 3 is an example of an information processing terminal used by the administrator or user of the knowledge extraction system 1, and is a computer such as a PC (Personal Computer) or a server. The terminal 3 may send, for example, a text to be extracted for knowledge, such as a document, to the server 2, or may receive the knowledge extracted from the text from the server 2.

Each of the terminal 3 and the server 2 and between the KG and the server 2 (not shown) may be communicably connected to each other via a network (not shown). The network may include a WAN (Wide Area Network), a LAN (Local Area Network), or a combination thereof. The WAN may include the Internet, and the LAN may include a VPN (Virtual Private Network).

The server 2 is an example of an extraction device, an information processing device, or a computer. For example, in the knowledge extraction system 1, the server 2 extracts the knowledge from the received text and saves the extracted knowledge (including transmission to the terminal 3) in response to the reception of the text from the terminal 3. Perform various processes.

The server 2 may be a virtual server (VM; Virtual Machine) or a physical server. Further, the function of the server 2 may be realized by one computer or may be realized by two or more computers. Further, at least a part of the functions of the server 2 may be realized by using the HW (Hardware) resource and the NW (Network) resource provided by the cloud environment.

(Hardware configuration example)
FIG. 2 is a block diagram showing a hardware (HW) configuration example of the computer 10 that realizes the function of the server 2. When a plurality of computers are used as the HW resource that realizes the function of the server 2, each computer may have the HW configuration illustrated in FIG.

As shown in FIG. 2, as an HW configuration, the computer 10 is exemplified by a processor 10a, a memory 10b, a storage unit 10c, an IF (Interface) unit 10d, an I / O (Input / Output) unit 10e, and a reading unit. It may be provided with 10f.

The processor 10a is an example of an arithmetic processing unit that performs various controls and operations. The processor 10a may be connected to each block in the computer 10 so as to be able to communicate with each other by the bus 10i. The processor 10a may be a multi-processor including a plurality of processors, a multi-core processor having a plurality of processor cores, or a configuration having a plurality of multi-core processors.

Examples of the processor 10a include integrated circuits (ICs) such as CPUs, MPUs, GPUs, APUs, DSPs, ASICs, and FPGAs. As the processor 10a, two or more combinations of these integrated circuits may be used. CPU is an abbreviation for Central Processing Unit, and MPU is an abbreviation for Micro Processing Unit. GPU is an abbreviation for Graphics Processing Unit, and APU is an abbreviation for Accelerated Processing Unit. DSP is an abbreviation for Digital Signal Processor, ASIC is an abbreviation for Application Specific IC, and FPGA is an abbreviation for Field-Programmable Gate Array.

The memory 10b is an example of HW that stores information such as various data and programs. Examples of the memory 10b include one or both of a volatile memory such as DRAM (Dynamic Random Access Memory) and a non-volatile memory such as PM (Persistent Memory).

The storage unit 10c is an example of HW that stores information such as various data and programs. Examples of the storage unit 10c include a magnetic disk device such as an HDD (Hard Disk Drive), a semiconductor drive device such as an SSD (Solid State Drive), and various storage devices such as a non-volatile memory. Examples of the non-volatile memory include flash memory, SCM (Storage Class Memory), ROM (Read Only Memory) and the like.

Further, the storage unit 10c may store a program 10g (extraction program) that realizes all or a part of various functions of the computer 10. For example, the processor 10a of the server 2 can realize the function as the server 2 illustrated in FIG. 1 by expanding and executing the program 10g stored in the storage unit 10c in the memory 10b.

The IF unit 10d is an example of a communication IF that controls connection and communication with a network. For example, the IF unit 10d may include a LAN such as Ethernet (registered trademark) or an adapter compliant with optical communication such as FC (Fibre Channel). The adapter may support one or both wireless and wired communication methods. For example, the server 2 may be communicably connected to each of the terminal 3 and the KG (not shown) via the IF unit 10d. Further, for example, the program 10g may be downloaded from the network to the computer 10 via the communication IF and stored in the storage unit 10c.

The I / O unit 10e may include one or both of an input device and an output device. Examples of the input device include a keyboard, a mouse, a touch panel, and the like. Examples of the output device include a monitor, a projector, a printer and the like.

The reading unit 10f is an example of a reader that reads data and program information recorded on the recording medium 10h. The reading unit 10f may include a connection terminal or device to which the recording medium 10h can be connected or inserted. Examples of the reading unit 10f include an adapter compliant with USB (Universal Serial Bus), a drive device for accessing a recording disk, a card reader for accessing a flash memory such as an SD card, and the like. The program 10g may be stored in the recording medium 10h, or the reading unit 10f may read the program 10g from the recording medium 10h and store it in the storage unit 10c.

Examples of the recording medium 10h include a non-temporary computer-readable recording medium such as a magnetic / optical disk or a flash memory. Examples of the magnetic / optical disk include flexible discs, CDs (Compact Discs), DVDs (Digital Versatile Discs), Blu-ray discs, HVDs (Holographic Versatile Discs), and the like. Examples of the flash memory include semiconductor memories such as USB memory and SD card.

The HW configuration of the computer 10 described above is an example. Therefore, the increase / decrease of HW (for example, addition or deletion of arbitrary blocks), division, integration in any combination, addition or deletion of buses, etc. may be appropriately performed in the computer 10. For example, in the server 2, at least one of the I / O unit 10e and the reading unit 10f may be omitted.

The terminal 3, which is an example of the information processing terminal, may be realized by the same HW configuration as the computer 10 described above. For example, the processor 10a of the terminal 3 can realize the function as the terminal 3 shown in FIG. 1 by expanding and executing the program 10g stored in the storage unit 10c in the memory 10b.

(Functional configuration example)
Next, an example of the functional configuration of the server 2 will be described with reference to FIG. The server 2 according to the embodiment performs a process of extracting knowledge from the text. For example, the server 2 detects the first set and the second set from the text, and each of the first plurality of elements included in the first set and the second plurality contained in the second set. Extract the correspondence relationship with each of the elements of.

FIG. 3 is a diagram for explaining an example of processing of the server 2. In the example shown in FIG. 3, the text on the server 2 is "corporate taxation trust, ..., or securities, respectively, as stipulated in Article 2, Paragraph 1, Item 8-3 ... or Item 17 of the Income Tax Act. Corporate taxation trust, ..., or securities. ”Is entered.

As illustrated in FIG. 3, the server 2 has a plurality of elements in "corporate taxation trust, ..., Or securities" which are a set of a plurality of elements (parallel phrases) for such a text. Extract the correspondence with each element in "Income Tax Law Article 2, Paragraph 1, Item 8-3 ... or Item 17", which is a set of elements. Of these two sets, one is the first set and the other is the second set.

For this purpose, the server 2 may include a communication unit 21, an element expansion processing control unit 22, and an order knowledge DB (Database) 23, as illustrated in FIG.

The communication unit 21 communicates with the terminal 3 and a KG (not shown). For example, the terminal 3 may receive a text to be extracted from the terminal 3 from the terminal 3. The communication unit 21 may store the received text in, for example, a storage area included in at least one of the memory 10b and the storage unit 10c shown in FIG.

The order knowledge DB 23 is a DB that stores a plurality of order knowledge 23a (see FIG. 5) regarding a set having an order in advance. The plurality of order knowledge 23a is knowledge in the real world, and is an example of a plurality of order information indicating the order in the plurality of elements included in each of the plurality of sets. For example, the order knowledge DB 23 may store a plurality of order knowledge 23a such as the order knowledge 23a of the symbol order, the order knowledge 23a of common sense, and the order knowledge 23a of each field.

The order knowledge 23a of the symbol order may include alphabetical order (A, B, C, ...), Iroha order (a, b, c, ...) and the like. The common sense order knowledge 23a may include a week (Sunday, Monday, ..., Sat) and a month (January, February, ..., December) and the like. The order knowledge 23a in each field may include railway stations, statutory provisions, and the like.

The element expansion processing control unit 22 is an example of a control unit, and manages the server 2 and controls execution such as calling each functional block (module) described later.

For example, the element expansion processing control unit 22 has, as a functional block, a knowledge range certification unit 22a, a multi-element expression analysis unit 22b, a component resolution unit 22c, an order knowledge selection unit 22d, an order recognition unit 22e, and a correspondence authentication unit 22f. , A meta-expression recognition unit 22g, and a knowledge conversion unit 22h may be provided.

FIG. 4 is a flowchart illustrating an operation example of the element expansion processing control unit 22. Hereinafter, an operation example of the functional blocks 22a to 22h will be described with reference to FIG.

The knowledge conversion range certification unit 22a certifies (determines) the knowledge conversion range on the text (step S1). For example, the knowledge acquisition range recognition unit 22a specifies a text pattern including a multi-element expression which is the above-mentioned set.

The multi-element expression analysis unit 22b detects and analyzes parallel phrases and range expressions from the range certified by the knowledgeable range certification unit 22a (step S2).

The component solving unit 22c recognizes the entity indicated by the multi-element representation of the analysis result (step S3). For example, the component solving unit 22c has an element (what is indicated) specifically indicated by the expression and the element concerned with respect to expressions such as "Income Tax Law Article 2, Paragraph 1" and "the same paragraph". The substance of (what kind of content the element is) is recognized.

Each of the processes of the knowledge acquisition range recognition unit 22a, the multi-element expression analysis unit 22b, and the component resolution unit 22c may be realized by various known methods.

The order knowledge selection unit 22d selects the order knowledge 23a to be used from the order knowledge DB 23 (step S4).

The order recognition unit 22e decomposes the range representation into individual elements by using the order knowledge 23a selected by the order knowledge selection unit 22d (step S5).

When the knowledgeable range certification unit 22a specifies a plurality of multiple element representations, the correspondence certification unit 22f associates the individual elements certified (disassembled) by the component resolution unit 22c and the order recognition unit 22e (the correspondence certification unit 22f). Step S6).

The meta expression recognition unit 22g recognizes information unnecessary for knowledge conversion, for example, meta expressions such as “each” (step S7).

The knowledge conversion unit 22h performs knowledge conversion for each of the parallel elements certified by the correspondence certification unit 22f (step S8), and the process is completed. The processing of the knowledge-making unit 22h may be realized by various known methods.

When the correspondence is extracted by the knowledge conversion by the knowledge conversion unit 22h, the communication unit 21 may transmit (store) the information of the correspondence to, for example, a KG (not shown). Alternatively, the server 2 may act as a KG by accumulating the correspondence-related information and provide the correspondence-related information to the user, another system, or the like.

[1-3] Description of Element Expansion Processing Control Unit Next, an example of processing of each functional block of the element expansion processing control unit 22 will be described.

(Knowledge scope certification unit 22a)
The knowledge conversion range certification unit 22a certifies the range of knowledge conversion from the text, regardless of the presence or absence of the multi-element expression, according to the target of knowledge conversion. For example, the knowledge-making range certification unit 22a searches the text for an expression that matches the pattern description based on a pattern such as a text (or a morpheme), and specifies the expression as a part (knowledge-making range) for knowledge-making. ..

An example of pattern description is described below.
-Definition rule = <definition word> blank <definition body>.
-Definition rule = <definition body> (referred to as "<definition word>" in <definition valid range>)
-Definition rule = <definition word> (meaning <definition text>)

For example, the knowledge conversion range certification unit 22a reads the pattern description, performs appropriate processing (for example, processing according to a rule such as not including a symbol in the definition text or the definition word), and converts it into a regular expression. As an example, the knowledge conversion range certification unit 22a converts the above first pattern description into the following regular expression. In the regular expressions below, "one" is the first Chinese character in unicode (unicode: 4e00) and "\ u9fa5" is the last Chinese character in unicode (unicode: 9fa5). This makes "1- \ u9fa5" a regular expression that matches any Chinese character. In addition, at least one of the third level kanji, the fourth level kanji, the dance character, and the like may be included in the regular expression depending on the text to be analyzed.
r'[I-\ u9fa5 A-N-A-Za-z] + \ s + [I-\ u9fa5 A-N-A-Za-z] +. '

Then, the knowledge conversion range certification unit 22a searches for a portion that matches the regular expression of each pattern with respect to the text to be processed, and specifies the matching portion as a knowledge conversion portion.

(Multi-element expression analysis unit 22b)
The multi-element expression analysis unit 22b independently analyzes each multi-element expression from the knowledge conversion range certified by the knowledge conversion range certification unit 22a. At the time when the processing by the multi-element expression analysis unit 22b is performed, the multi-element expression is not associated with the entity indicated by the multi-element expression. For example, the multi-element expression analysis unit 22b searches the knowledgeable range for an expression that matches the pattern description based on a pattern such as text (or morpheme), and detects the expression as an element of the multi-element expression.

An example of pattern description of multi-element expression is described below.
-Multiple element representation = (<phrase>,) * <phrase> (and | and | or | or | other? | (Etc. | etc.)) <Phrase>
・ Multi-element expression = from <phrase> to <phrase>

For example, the multi-element expression analysis unit 22b reads the pattern description, performs appropriate processing, and converts it into a regular expression.

Then, the multi-element expression analysis unit 22b searches for a portion that matches the regular expression of each pattern for the text to be processed, and certifies (identifies) the matching portion as an element of the multi-element expression.

In addition, there are cases where a multi-element expression is included in the multi-element expression (nested structure), and a case where the pattern is in the form of a multi-element expression but not a parallel element (for example, when it is an adverbial phrase). exist. Therefore, the multi-element expression analysis unit 22b may determine whether the element matched by the search is an element of the multi-element expression to be specified.

Further, the multi-element expression analysis unit 22b may divide the above-mentioned pattern description into a plurality of parts and perform analysis for each divided pattern description.

(Component solving unit 22c)
The component solving unit 22c recognizes (identifies) the entity pointed to by the multi-element expression. For example, the component solving unit 22c may perform at least one of the following processes (i) to (iii).

(I) Reference expression resolution process The component resolution unit 22c specifies a specific indication for expressions or omissions such as “same”, “previous”, and “next”. For example, in the component resolution unit 22c, among the multi-element expressions of "Income Tax Law Article 2, Paragraph 1, Item 8-3 to 13", "No. 13" is "Income Tax Law Article 2". It may be specified as "paragraph 1, item 13".

(Ii) Entity Reference Processing The Component Resolution Unit 22c responds to expressions that refer to an entity, such as "Income Tax Law, Article 2, Paragraph 1, Item 8-3," in response to "Income Tax Law, Article 2, Item 2" in KG, etc. Identify (identify) the identifier (ID; Identifier) of the entity (entity) of "1. No. 8-3". The ID may be, for example, a URI (Uniform Resource Locator) in KG. As an example, in a KG such as Linked Data, information related to the item can be obtained by referring to the URI. The component solving unit 22c may specify such a URI such as KG.

(Iii) Character string reference processing The component solving unit 22c processes elements such as definition words that do not require substance reference according to (ii) above as literals (constants). For example, the component solving unit 22c may process "domestic" and "foreign" of "meaning domestic or foreign specified in ..." as literals, respectively.

(Order knowledge selection unit 22d)
The order knowledge selection unit 22d selects the order knowledge 23a to be used from the order knowledge DB 23. For example, the order knowledge selection unit 22d determines which order knowledge 23a is to be used when a plurality of order knowledge 23a exists in the order knowledge DB 23.

FIG. 5 is a diagram showing an example of the order knowledge DB 23. The order knowledge DB 23 may include an order set (order set), each element belonging to the order set, an index indicating the order of each element, and the like as the order knowledge 23a.

In the example of FIG. 5, the “a” type “: Week” defined as “ex: OrderSet” in the first line is the order set. Each of ": Sunday" and ": Monday" (others omitted) defined as "memberOf: Week" in the 6th and 8th lines belong to the order set ": Week" (member element). The order of the elements is indicated by "ex: index". In FIG. 5, “#” is a symbol indicating a comment out in the line.

As described above, FIG. 5 shows an example in which the order knowledge DB 23 stores one order knowledge 23a relating to the order set “: Week”, but the present invention is not limited to this, and the order knowledge DB 23 has a plurality of orders. Knowledge 23a may be stored. Other examples of sequence knowledge 23a include sequence knowledge 23a relating to statutory provisions, as well as sequence sets "note names" and elements "A", "A #", "B", "C", "C #", ... and so on.

In the order knowledge 23a, each element is set with one or more phenotypes expressing the element. In the example of FIG. 5, the element ": Sunday" is set to "Sunday" and "Sun" in Japanese "@ja" and "Star period heaven" in Chinese "@zn" as the phenotype "rdfs: label". Has been done. As another example, in the element "A" of the order set "note name", the phenotype "rdfs: label" is "A", "i", "la", "La", "la", ...・ ・ Etc. may be set.

Further, each sequence knowledge 23a may include metadata. The metadata may include attributes indicating the nature of the order set, as well as information such as descriptions (or keywords thereof) of fields of use and the like.

The attribute information in the metadata may include, for example, information such as the directionality of the order of the elements in the sequence set, the circularity of the elements, and the like.

In the example of FIG. 5, when the value of the presence / absence of circulation defined as “ex: cyclicity” in the third line is “True”, it means that there is circulation, and when it is “False”, it means circulation. It means that there is no sex. For example, in the case of the sequence set ": Week", "Saturday" (not shown) indicated by the element ": Saturday" and the index "6" is followed by "Sunday" indicated by the element ": Sunday" and the index "0". (See the 6th line), so the value of the cyclicity "ex: cyclicity" is "True". The number of elements in the order set is indicated by the value of "ex: numberOfElements" in the fourth row ("7" in the example of FIG. 5).

Examples of a cyclical order set include, in addition to ": Week" (day of the week), a month name, a station name of a circular line (for example, the Yamanote line), and the like.

In addition, the circular order set has a clear direction as a standard directional order set (in the case of a day of the week, "Sunday" is followed by "Monday", etc.), such as the day of the week and the month name. There is an order set that is not. An example of an order set in which the direction is not clear is the station name of a circulation line (for example, the Yamanote line). For an order set in which the directionality is not clear, expressions indicating the directionality (for example, "inner loop", "outer loop") and the like may be set as metadata.

Information on the description (or its keyword) of the field of use or the like in the metadata includes, for example, an explanation of the order set, a related keyword of the order set, and the like.

In the example of FIG. 5, "week. A unit of time with 7 days as one cycle." (Japanese "@ja") defined as "ex: description" in the second line is an explanation of the order set. In addition, "week", "day of the week", "weekday", and "holiday" (all in Japanese "@ja") defined as "ex: topic" in the fifth line are related keywords. As another example, "absolute pitch", "scale", "octave", ..., Etc. may be set as related keywords of the order set "note name".

The order knowledge DB 23 may store information including member elements and indexes for each of the plurality of order knowledge 23a in the above-mentioned format. Further, the sequence knowledge DB 23 may hold the information of the sequence knowledge 23a in a table format, or may hold the information of the sequence knowledge 23a in a table format or, for example, in an RDF (Resource Description Framework) format as KG. The order knowledge DB 23 may be realized, for example, by a storage area possessed by at least one of the memory 10b and the storage unit 10c shown in FIG. In the order knowledge DB 23, for example, one or more order knowledge 23a may be stored before the execution timing of the knowledge extraction process.

For example, the order knowledge selection unit 22d extracts from the order knowledge DB 23 the order knowledge 23a having each of the range expressions specified by the component solving unit 22c in the notation of the member elements.

Here, the order knowledge 23a may not be uniquely determined. For example, when the range expression is "ha to ho", "ha" and "ho" are the order knowledge 23a of the iroha order (a, ro, ha, ni, ho) of the order knowledge 23a, and It corresponds to both of the order knowledge 23a in the order of the Japanese syllabary (ha, hi, fu, he, e).

In such a case, the order knowledge selection unit 22d obtains the degree of matching between the expression (for example, a keyword or the like) and the member element of the order set based on the expression included in the text (extraction range), and uses it as an evaluation function. Based on this, the order knowledge 23a may be determined. The expressions included in the text (extraction range) include, for example, other range expressions (including, for example, “variant”) specified by the component solving unit 22c, and the range in the text (extraction range). Keywords (for example, "octave") appearing in expressions other than expressions (general expressions) can be mentioned.

The order knowledge selection unit 22d may determine, for example, one order knowledge 23a having the maximum evaluation value based on the evaluation function, or have n (n is an integer of 2 or more) higher evaluation values. The sequence knowledge 23a (as a candidate) may be determined.

(Order recognition unit 22e)
The order recognition unit 22e may execute the following processes (I) to (VI). Further, the order recognition unit 22e (or the multi-element expression analysis unit 22b) optionally executes the following processes (VII) and (VIII) in connection with the following processes (I) to (VI). May be good.

(I) Identification of substance The order recognition unit 22e searches for an expression that matches the label (rdfs: label) of the order knowledge 23a for the start element expression s and the end element expression e of the range expression, and the start element. Recognize each entity of the expression s and the terminal element expression e.

The starting element representation s is an example of the starting element (first appearing) among the plurality of elements described in the text of the first plurality of elements. The terminal element representation e is an example of the (last appearing) terminal element among the plurality of elements described in the text of the first plurality of elements. The first plurality of elements are, in the example of FIG. 3, a plurality of provisions of the Income Tax Act.

As an example, when the range expression is "from month to day", the order recognition unit 22e specifies "month" as the start element expression s and "day" as the end element expression e.

The order recognition unit 22e refers to the member elements of the order knowledge 23a, and sets “: Monday” as the entity S of the start element expression s “month” and “: Sunday” as the entity E of the end element expression e “day”, respectively. Identify.

As the entity S and the entity E, a plurality of candidates may be specified for each. Hereinafter, in order to include the case where a plurality of candidates are specified, the substance S and the substance E may be referred to as the substance S_i and the substance E_j, respectively. i and j may be Japanese syllabary, pitch, or the like, respectively, or may be an integer of 1 or more. In other words, there may be a plurality of entity candidates S_1, S2, ..., E_1, E_2, ... For the start element expression s or the end element expression e. For example, the entity "i_gojyuon", "i_pitch", and the like may exist in "i", and these entities (candidates) may be expressed as S_i and E_j.

(II) Certification of order set When the entity associated with at least one of the start element expression s or the end element expression e has a plurality of candidates S_i and E_j, the order recognition unit 22e is assigned to each of the entities S_i and E_j. On the other hand, the order set O to which the entity belongs is obtained, and the order set common between the entities S_i and the entity E_j is obtained in the order set O.

For example, the order recognition unit 22e obtains “: Week” (O) to which “: Monday” (S_i) belongs and “: Week” (O) to which “: Sunday” (E_j) belongs, and the entity S_i And, as an order set common between the entities E_j, ": Week" is obtained.

In this way, the order recognition unit 22e specifies the start element expression s and the end element expression e, and from the plurality of order knowledge 23a, each of the start element expression s and the end element expression e is in the plurality of ordered elements. The first order knowledge 23a included in is specified. Thereby, it is possible to specify an appropriate order knowledge 23a to be used for the first set which is a range representation.

(III) Order set determination When the order recognition unit 22e obtains a plurality of order sets as candidates in the process of (II) above, in other words, the start element expression s and the end element are contained in the ordered plurality of elements. When there are a plurality of candidates of the first order knowledge 23a including the expression e, one order set is determined from the candidate order sets.

For example, it is assumed that the entity S_i or the entity E_j is an entity associated with the character string "a" as shown below.
: I @ Japanese syllabary: memberOf: Japanese syllabary; ex: index 1; rdfs: label “I” @ja
: I @ Iroha: memberOf: Iroha; ex: index 0; rdfs: label “I” @ja

As in the above example, the entity associated with the character string "i" is both an element of the order set ": Japanese syllabary" and an element of the order set ": Iroha". Similarly, the entity associated with the character string "d" is also an element of both.

In the above case, the order recognition unit 22e determines the order set based on the number of elements or heuristics.

For example, in the determination of the order set based on the number of elements, the order recognition unit 22e has, for example, the entity "I_Iroha" and the entity "Ha_Iroha" having three elements, but the entity "I_50". Since the number of elements is 25 between the substance "sound" and the entity "ha_five sounds", an order set that can be consistent with the number of other multiple elements in the knowledge extraction range, for example, the number of elements from the entity S_i to the entity E_j. You may find an order set that matches the number of other elements in the knowledge extraction range.

In other words, in the order knowledge 23a, the number of elements up to the start element expression s and the end element expression e according to the order indicated by each of the order knowledge 23a of the plurality of candidates matches the number of elements of the second plurality of elements. The candidate to be used is specified as the first order knowledge 23a. The second plurality of elements are definition phrases (corporate taxation trust, etc.) in the example of FIG.

Further, for example, the order recognition unit 22e may determine an order set suitable for heuristics such as "the index of the starting element expression s is often" 0 "" in the determination of the order set based on the heuristics.

As an example, in the order recognition unit 22e, if the starting element expression s is "i", the entity S_i may be an order set ": Iroha" rather than an order set ": Japanese syllabary". The order set ": Iroha" may be selected based on heuristics such as higher. The order recognition unit 22e may hold, for example, each of one or more heuristics as a linear expression and use it.

In other words, the order knowledge 23a is the first from the plurality of candidates depending on the heuristics based on the order indicated by each of the order knowledge 23a of the plurality of candidates and at least one of the start element expression s and the end element expression e. Specify the sequence knowledge 23a.

In this way, the order knowledge 23a can determine one appropriate order set from a plurality of candidate order sets based on the number of elements or heuristics.

In addition, in the process of (III), the order recognition unit 22e does not determine one order set from the plurality of order sets as candidates, but performs the processing after (IV) for each of the plurality of order sets as candidates. One set of order may be determined in any of the processes (IV) to (VIII).

(IV) Specifying the index The order recognition unit 22e obtains the indexes I_S and I_E for the substance S of the specified start element expression s and the substance E of the end element expression e, respectively.

For example, assume that the order set “: Week”, the entity S “: Monday”, and the entity E “: Sunday” are specified, respectively. In this case, the order recognition unit 22e specifies "1" as the index I_S of the entity S ": Monday" based on the order knowledge 23a of the order set ": Week", and sets "1" as the index I_E of the entity E ": Sunday". Specify "0" (see Fig. 5).

(V) Directional recognition The order recognition unit 22e determines the presence or absence of circulation in the order set. For example, the order recognition unit 22e refers to the order knowledge 23a of the order set and determines whether or not the value of the metadata circulation “ex: cyclicity” is “True”.

When the value of the cyclicity “ex: cyclicity” is “False” (no cyclicity), the order recognition unit 22e determines that the direction is “+” (plus) when I_E> I_S. When I_E <I_S, it is determined that the direction is "-" (minus).

When the value of the cyclicity “ex: cyclicity” is “True” (there is a circularity), the order recognition unit 22e indicates the direction when the direction is clear (including the case where there is no direction indication expression in the metadata). Is determined to be "+" (plus). In the example of FIG. 5, since the value of “ex: cyclicity” is “True” and the directionality is clear, the order recognition unit 22e determines that the direction is “+”.

On the other hand, when the value of the cyclicity “ex: cyclicity” is “True” (there is circularity) and the directionality is unclear, the order recognition unit 22e is directed from the order knowledge 23a of the order set. The instructional expression is extracted, and the direction (“+” or “-” of the index) is determined according to the expression. The direction indication expression will be described later.

(VI) Certification of range elements The order recognition unit 22e refers to the order knowledge 23a and sets the index from I_S to I_E according to the direction by "+1" (increment) or "-1" (decrement), respectively. Extract the elements of.

At this time, if the order set has circularity, the order recognition unit 22e may calculate, for example, the remainder (mod) of the number of elements in the set. For example, when the number of elements is “7” and the direction is “+”, the index “6” is followed by “(6 + 1) (mod 7) = 0”, which results in “0”.

As an example, when the index "I_S" of the entity S ": Monday" is "1", the index "I_E" of the entity E ": Sunday" is "0", and the direction is "+", the order is recognized. Part 22e determines and extracts each of the index and the element as follows.
・ Index: “1”, “2”, “3”, “4”, “5”, “6”, “0”
-Elements: “: Monday”, “: Tuesday”, ..., “: Sunday”

Hereinafter, as another example, the order recognition unit 22e will explain the process of extracting an element from the range expression “from income tax law Article 2, paragraph 1, item 8-3 to item 13”.

FIG. 6 is a diagram for explaining an example of the expansion process of the range element based on the order knowledge 23a. FIG. 6 illustrates the order set “Income Tax Act” and its parameters, as well as some of the member elements of the order set. In FIG. 6, the index may be represented by a scalar of one layer or a vector of a plurality of layers in the index of the member element in the whole of the order set “Income Tax Act”. Examples of expression of indexes in multi-level vectors are Article 1 (1,), Article 1 Paragraph 1 (1, 1), Article 2 (2,), Article 2 Paragraph 1 (2,,). 1) etc. may be used.

The order recognition unit 22e can develop the element (range element) of the range expression as follows by referring to the order knowledge 23a illustrated in FIG.
"Income Tax Law, Article 2, Paragraph 1, Item 8-3", "Income Tax Law, Article 2, Paragraph 1, Item 8-4", "Income Tax Law, Article 2, Paragraph 1, Item 9", "Income Tax Law, No. 2" Article 1, Paragraph 1, Item 10, "Income Tax Law, Article 2, Paragraph 1, Item 11", "Income Tax Law, Article 2, Paragraph 1, Item 12", "Income Tax Law, Article 2, Paragraph 1, Item 12-2" , "Income Tax Law, Article 2, Paragraph 1, Item 13"

As described above, the order recognition unit 22e is based on the presence or absence of the circulation of the order indicated by the first order knowledge 23a and the direction of the order specified based on the presence or absence of the circulation, from the first order knowledge 23a. Each element from the start element expression s to the end element expression e is extracted as the first plurality of elements. As a result, even if there are elements not described in the text, it is possible to take out the first plurality of elements that are appropriate for the range specification in the first set.

(VII) Direction instruction processing (optional)
Hereinafter, the direction instruction expression in the process of (V) above will be described. An ordinal set with unclear directionality may retain directional directives given for each direction as metadata for ordinal knowledge 23a.

FIG. 7 is a diagram showing an example of order knowledge 23a including direction indicating words. As illustrated in FIG. 7, ascending and descending labels (direction directives) may be set as parameters in the order knowledge 23a including each stop station of the order set “: Yamanote Line” as a member element.

For example, the ascending label “ex: ascendingOrderLabel” is set to “outer loop” (Japanese “@ja”), and the descending label “ex: descendingOrderLabel” is set to “inner loop” (Japanese “@ja”). May be set.

When the multi-element expression analysis unit 22b extracts the parallel phrase and the range expression from the range certified by the knowledgeable range certification unit 22a, the multi-element expression analysis unit 22b may also extract the directional demonstrative words as illustrated below.
r “In order of <direction indicator> from <starting element> to <ending element>”

For example, when the multi-element expression analysis unit 22b extracts a demonstrative word, the order recognition unit 22e indicates a direction that matches (matches) the extracted demonstrative word in the label of the demonstrative word in the order set. You may determine the direction by searching for a word. The determination of the direction may be executed by the multi-element expression analysis unit 22b.

(VIII) Exclusion expression processing and combined processing (optional)
The multi-element expression analysis unit 22b may extract the enumeration expression and the exclusion expression. As an example of the exclusion expression, it is the expression "excluding <exclusion range>" described in parentheses "()" as shown below.
r "From <starting element> to <ending element> (excluding <exclusion range>)"

For example, the order recognition unit 22e may specify a specific entity with reference to the order knowledge 23a for the counting expression and the exclusion expression. As an example, the order recognition unit 22e repeatedly calls a plurality of range elements for the enumeration expression, and connects and extracts the elements. As another example, the order recognition unit 22e removes the element specified by the exclusion expression from the extracted elements for the exclusion expression.

As described above, the order recognition unit 22e includes, from the plurality of order knowledge 23a, each of the plurality of elements described in the text among the first plurality of elements in the ordered plurality of elements. The order knowledge 23a of is specified. Similarly, the order recognition unit 22e includes, from the plurality of order knowledge 23a, each of the plurality of elements described in the text among the second plurality of elements in the ordered plurality of elements. The order knowledge 23a of 2 is specified.

As a result, the order recognition unit 22e can recognize the entity pointed to by each expression of the first plurality of elements included in the first set, instead of the superficial expression, and can specify the range in the first set, etc. It is possible to take out appropriate elements in.

Further, by deleting the exclusion expression from the first plurality of elements, the order recognition unit 22e can extract an appropriate element even when the text includes the exclusion expression.

As described above, the knowledge acquisition range recognition unit 22a, the multi-element expression analysis unit 22b, the component resolution unit 22c, the order knowledge selection unit 22d, and the order recognition unit 22e as an example (partial function) of the control unit According to this, at least the first plurality of elements can be specified as follows. That is, when the control unit detects the first set and the second set from the text, it refers to the first order knowledge 23a, and at least one part is included in the first set not described in the text. The first plurality of elements can be identified.

(Correspondence Relationship Certification Department 22f)
When the correspondence recognition unit 22f has multiple element expressions in a plurality of places in the text to be knowledgeed, the correspondence recognition unit 22f can be used between the multiple element expressions for each element expression developed from the multiple element expressions by the order recognition unit 22e. Correspond with.

As an example, when the text shown in FIG. 5 is input to the server 2, from the processing results by the above-mentioned knowledge acquisition range recognition unit 22a to order recognition unit 22e, the three shown in parentheses “()” are as follows. Multi-element representation is recognized.
"(Corporate taxation trust, permanent facility, ... securities) (Income Tax Law, Article 2, Paragraph 1, Item 8-3, Income Tax Law, Article 2, Paragraph 1, Item 8-4, Income Tax Law, Second) Article 1, item 9, ..., Income Tax Act, Article 2, paragraph 1, item 17) (corporate taxation trust, permanent facility, ... securities). "

In the case of the above example, the correspondence relationship recognition unit 22f extracts the correspondence relationship in which the elements are associated between the plurality of element expressions as follows.
(Corporate taxation trust, Income Tax Act, Article 2, Paragraph 1, Item 8-3, Corporate taxation trust),
(Permanent establishment, Income Tax Act, Article 2, Paragraph 1, Item 8-4, Permanent establishment),
…
(Securities, Income Tax Act, Article 2, Paragraph 1, Item 17, Securities)

In this way, the correspondence certification unit 22f determines the order in the first plurality of elements indicated by the first order knowledge 23a and the order in the second plurality of elements indicated by the second order knowledge 23a. Based on this, the correspondence relationship in which each of the first plurality of elements and each of the second plurality of elements are associated with each other is extracted. As a result, an appropriate correspondence can be obtained.

(Meta expression recognition unit 22g)
The meta expression recognition unit 22g identifies and identifies meta expressions such as "each", "each", "and", and "or" by referring to a preset dictionary. Delete the expression. The deletion may include, for example, the deletion of the meta expression from the extraction result, or may include ignoring (managing as an invalid expression) so that the expression is not the expression to be processed in the processing by the knowledge acquisition unit 22h. good.

(Knowledge Department 22h)
The knowledge conversion unit 22h performs knowledge conversion for each element extracted as a correspondence relationship by the correspondence relationship certification unit 22f. For example, the knowledge conversion unit 22h may hold the extracted correspondence as KG in a DB (not shown) in RDF format. The processing by the knowledge-making unit 22h may be realized by various known methods.

As described above, according to the correspondence recognition unit 22f (and the meta expression recognition unit 22g and the knowledge acquisition unit 22h) as an example (partial function) of the control unit, the correspondence is as follows. Can be extracted. That is, the control unit has a correspondence relationship in which each of the first plurality of elements and each of the second plurality of elements included in the second set are associated with each other according to the order in the first plurality of elements. To extract.

As described above, according to the knowledge extraction system 1 according to the embodiment, it is possible to specify a plurality of elements included in the range expression even for a range expression in which at least a part of the elements is not described in the text. can. Therefore, the correspondence between the sets detected from the text can be appropriately recognized.

For example, when the text shown in FIG. 5 is input, the server 2 uses the definition words such as "corporate taxation trust, permanent establishment, ... securities" and the provisions of the Income Tax Act, as shown in FIG. It is possible to appropriately extract the correspondence relationship associated with.

[2] Others The technique according to the above-described embodiment can be modified or modified as follows.

For example, the communication unit 21, the element expansion processing control unit 22, and the order knowledge DB 23 included in the server 2 shown in FIG. 1 may be merged or divided in any combination. Further, the functional blocks 22a to 22h included in the element expansion processing control unit 22 may be merged or divided in any combination.

Further, the server 2 shown in FIG. 1 may have a configuration in which a plurality of devices cooperate with each other via a network to realize each processing function. As an example, the communication unit 21 may be a Web server, the element expansion processing control unit 22 may be an application server, the order knowledge DB 23 may be a DB server, and the like. In this case, the Web server, the application server, and the DB server may cooperate with each other via the network to realize each processing function as the server 2.

Further, in the example shown in FIG. 5, at least one of the elements of the income tax law clause, that is, the multi-element representation of the income tax law clause as the first set and the multi-element representation of the definition word as the second set. One part is not described in the text.

However, the present invention is not limited to this, and when the multi-element expression of the definition word is a range expression, the definition word is set as the first set, and at least one part of the elements of the multi-element expression of the definition word is in the text. It does not have to be described. Even in this case, the knowledge extraction system 1 has a correspondence relationship between a plurality of elements included in the definition word as the first set and a plurality of elements included in the provisions of the Income Tax Act as the second set. Can be properly extracted.

Further, in both the multi-element expression of the definition word and the multi-element expression of the provisions of the Income Tax Act, at least one part of the element may not be described in the text. Even in this case, the knowledge extraction system 1 can appropriately extract the correspondence between the plurality of elements included in the definition word and the plurality of elements included in the provisions of the Income Tax Act.

[3] Additional notes The following additional notes will be further disclosed with respect to the above embodiments.

(Appendix 1)
When the first set and the second set are detected from the text, the first plurality of first sets are referred to and at least one part thereof is included in the first set not described in the text. Identify the element and
According to the order in the first plurality of elements, a correspondence relationship in which each of the first plurality of elements is associated with each of the second plurality of elements included in the second set is extracted. ,
An extraction method in which a computer performs processing.

(Appendix 2)
The process of identifying the first plurality of elements is described in the text of the first plurality of elements from the plurality of order information indicating the order in the plurality of elements included in each of the plurality of sets. Each of the plurality of elements is included in the plurality of ordered elements, including specifying the first order information.
The extraction method according to Appendix 1.

(Appendix 3)
The process of specifying the first order information is
The start element and the end element of the plurality of elements described in the text among the first plurality of elements are specified.
From the plurality of order information, the first order information including each of the start element and the end element in the ordered elements is specified.
The extraction method according to Appendix 2.

(Appendix 4)
In the process of specifying the first order information, when there are a plurality of candidates for the first order information including the start element and the end element in the plurality of ordered elements, each of the plurality of candidates is used. The candidate including the candidate whose number of elements from the start element to the end element according to the indicated order matches the number of elements of the second plurality of elements is specified as the first order information.
The extraction method according to Appendix 3.

(Appendix 5)
In the process of specifying the first order information, when there are a plurality of candidates for the first order information including the start element and the end element in the plurality of ordered elements, each of the plurality of candidates is used. The first order information is specified from the plurality of candidates according to the heuristics based on the order shown and at least one of the start element and the end element.
The extraction method according to Appendix 3.

(Appendix 6)
Each of the plurality of order information includes the presence or absence of circulation of the order within the plurality of elements.
The process for extracting the correspondence is the first order information based on the presence or absence of the circulation in the order indicated by the first order information and the direction of the order specified based on the presence or absence of the circulation. Includes extracting each element from the start element to the end element as the first plurality of elements.
The extraction method according to any one of Supplementary note 3 to Supplementary note 5.

(Appendix 7)
The process of extracting the correspondence includes deleting the exclusion expression included in the first set from the first plurality of elements.
The extraction method according to any one of Supplementary note 1 to Supplementary note 6.

(Appendix 8)
From the plurality of order information indicating the order in the plurality of elements contained in each of the plurality of sets, each of the plurality of elements described in the text among the second plurality of elements is ordered. Identify the second order information contained within the element of
The computer executes the process,
The process of extracting the correspondence is divided into the order in the first plurality of elements indicated by the first order information and the order in the second plurality of elements indicated by the second order information. Including extracting the correspondence based on
The extraction method according to any one of Supplementary note 1 to Supplementary note 7.

(Appendix 9)
When the first set and the second set are detected from the text, the first plurality of first sets are referred to and at least one part thereof is included in the first set not described in the text. Identify the element and
According to the order in the first plurality of elements, a correspondence relationship in which each of the first plurality of elements is associated with each of the second plurality of elements included in the second set is extracted. ,
An extraction program that lets a computer perform processing.

(Appendix 10)
The process of identifying the first plurality of elements is described in the text of the first plurality of elements from the plurality of order information indicating the order in the plurality of elements included in each of the plurality of sets. Each of the plurality of elements is included in the plurality of ordered elements, including specifying the first order information.
The extraction program according to Appendix 9.

(Appendix 11)
The process of specifying the first order information is
The start element and the end element of the plurality of elements described in the text among the first plurality of elements are specified.
From the plurality of order information, the first order information including each of the start element and the end element in the ordered elements is specified.
The extraction program according to Appendix 10.

(Appendix 12)
In the process of specifying the first order information, when there are a plurality of candidates for the first order information including the start element and the end element in the plurality of ordered elements, each of the plurality of candidates is used. The candidate including the candidate whose number of elements from the start element to the end element according to the indicated order matches the number of elements of the second plurality of elements is specified as the first order information.
The extraction program according to Appendix 11.

(Appendix 13)
In the process of specifying the first order information, when there are a plurality of candidates for the first order information including the start element and the end element in the plurality of ordered elements, each of the plurality of candidates is used. The first order information is specified from the plurality of candidates according to the heuristics based on the order shown and at least one of the start element and the end element.
The extraction program according to Appendix 11.

(Appendix 14)
Each of the plurality of order information includes the presence or absence of circulation of the order within the plurality of elements.
The process for extracting the correspondence is the first order information based on the presence or absence of the circulation in the order indicated by the first order information and the direction of the order specified based on the presence or absence of the circulation. Includes extracting each element from the start element to the end element as the first plurality of elements.
The extraction program according to any one of Supplementary note 11 to Supplementary note 13.

(Appendix 15)
The process of extracting the correspondence includes deleting the exclusion expression included in the first set from the first plurality of elements.
The extraction program according to any one of Supplementary note 9 to Supplementary note 14.

(Appendix 16)
From the plurality of order information indicating the order in the plurality of elements contained in each of the plurality of sets, each of the plurality of elements described in the text among the second plurality of elements is ordered. Identify the second order information contained within the element of
Let the computer perform the process
The process of extracting the correspondence is divided into the order in the first plurality of elements indicated by the first order information and the order in the second plurality of elements indicated by the second order information. Including extracting the correspondence based on
The extraction program according to any one of Supplementary note 9 to Supplementary note 15.

(Appendix 17)
When the first set and the second set are detected from the text, the first plurality of first sets are referred to and at least one part thereof is included in the first set not described in the text. Identify the element and
According to the order in the first plurality of elements, a correspondence relationship in which each of the first plurality of elements is associated with each of the second plurality of elements included in the second set is extracted. ,
An extraction device equipped with a control unit.

(Appendix 18)
The control unit selects each of the plurality of elements described in the text among the first plurality of elements from the plurality of order information indicating the order in the plurality of elements included in each of the plurality of sets. , Identifying the first order information contained within a plurality of ordered elements,
The extraction device according to Appendix 17.

(Appendix 19)
The control unit
The start element and the end element of the plurality of elements described in the text among the first plurality of elements are specified.
From the plurality of order information, the first order information including each of the start element and the end element in the ordered plurality of elements is specified.
The extraction device according to Appendix 18.

(Appendix 20)
The control unit
From the plurality of order information indicating the order in the plurality of elements contained in each of the plurality of sets, each of the plurality of elements described in the text among the second plurality of elements is ordered. Identify the second order information contained within the element of
The correspondence is extracted based on the order in the first plurality of elements indicated by the first order information and the order in the second plurality of elements indicated by the second order information. ,
The extraction device according to any one of Supplementary note 17 to Supplementary note 19.

1 Knowledge extraction system 10 Computer 2 Server 21 Communication unit 22 Element expansion processing control unit 22a Knowledge range certification unit 22b Multi-element expression analysis unit 22c Component resolution unit 22d Order knowledge selection unit 22e Order recognition unit 22f Correspondence relationship certification unit 22g Meta Expression recognition department 22h Knowledge conversion department 23 Order knowledge DB
23a Order knowledge 3 Terminal

Claims (10)

When the first set and the second set are detected from the text, the first plurality of first sets are referred to and at least one part thereof is included in the first set not described in the text. Identify the element and
According to the order in the first plurality of elements, a correspondence relationship in which each of the first plurality of elements is associated with each of the second plurality of elements included in the second set is extracted. ,
An extraction method in which a computer performs processing. The process of identifying the first plurality of elements is described in the text of the first plurality of elements from the plurality of order information indicating the order in the plurality of elements included in each of the plurality of sets. Each of the plurality of elements is included in the plurality of ordered elements, including specifying the first order information.
The extraction method according to claim 1. The process of specifying the first order information is
The start element and the end element of the plurality of elements described in the text among the first plurality of elements are specified.
From the plurality of order information, the first order information including each of the start element and the end element in the ordered elements is specified.
The extraction method according to claim 2. In the process of specifying the first order information, when there are a plurality of candidates for the first order information including the start element and the end element in the plurality of ordered elements, each of the plurality of candidates is used. The candidate including the candidate whose number of elements from the start element to the end element according to the indicated order matches the number of elements of the second plurality of elements is specified as the first order information.
The extraction method according to claim 3. In the process of specifying the first order information, when there are a plurality of candidates for the first order information including the start element and the end element in the plurality of ordered elements, each of the plurality of candidates is used. The first order information is specified from the plurality of candidates according to the heuristics based on the order shown and at least one of the start element and the end element.
The extraction method according to claim 3. Each of the plurality of order information includes the presence or absence of circulation of the order within the plurality of elements.
The process for extracting the correspondence is the first order information based on the presence or absence of the circulation in the order indicated by the first order information and the direction of the order specified based on the presence or absence of the circulation. Includes extracting each element from the start element to the end element as the first plurality of elements.
The extraction method according to any one of claims 3 to 5. The process of extracting the correspondence includes deleting the exclusion expression included in the first set from the first plurality of elements.
The extraction method according to any one of claims 1 to 6. From the plurality of order information indicating the order in the plurality of elements contained in each of the plurality of sets, each of the plurality of elements described in the text among the second plurality of elements is ordered. Identify the second order information contained within the element of
The computer executes the process,
The process of extracting the correspondence is divided into the order in the first plurality of elements indicated by the first order information and the order in the second plurality of elements indicated by the second order information. Including extracting the correspondence based on
The extraction method according to any one of claims 1 to 7. When the first set and the second set are detected from the text, the first plurality of first sets are referred to and at least one part thereof is included in the first set not described in the text. Identify the element and
According to the order in the first plurality of elements, a correspondence relationship in which each of the first plurality of elements is associated with each of the second plurality of elements included in the second set is extracted. ,
An extraction program that lets a computer perform processing. When the first set and the second set are detected from the text, the first plurality of first sets are referred to and at least one part thereof is included in the first set not described in the text. Identify the element and
According to the order in the first plurality of elements, a correspondence relationship in which each of the first plurality of elements is associated with each of the second plurality of elements included in the second set is extracted. ,
An extraction device equipped with a control unit.

Images