JP2014238628A - Legal document analysis system and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technology for optimizing a legal document in advance so as to correctly interpret semantic contents in subsequent computer processing.SOLUTION: A legal document analysis system 10, provided with an analysis rule storage part 24 for defining a relation between specific elements constituting a document and contents of conversion processing to each element as an analysis rule unique to a legal document in advance to execute morphological analysis processing for decomposing inputted legal document data into morpheme units and syntax analysis processing for specifying a modification structure between respective morphemes, refers to the analysis rule storage part 24 during syntax analysis processing, and generates a sentence subjected to conversion processing defined by an analysis rule in the case that a relation between elements corresponding to the analysis rule is included in the legal document data.

Description

  The present invention relates to a legal document analysis system and program, and more particularly to a preprocessing technique for converting a legal document such as a contract or a clause into a format suitable for subsequent computer processing.

With the progress of text mining technology, computer processing such as search and translation has been applied to natural sentences written by humans.
Legal Transer 2010 Internet URL: http://www.crosslanguage.co.jp/products/legal2010/ Search date: April 23, 2013

Certainly, a document written with an emphasis on readability from the beginning, such as general news articles and blog articles, interprets natural sentences with a certain degree of correct meaning even with the current document analysis technology. I can say that.
However, in the case of legal documents that are described with emphasis on accuracy and rigor rather than readability, it is not sufficient to analyze only in accordance with general grammatical rules.

  Of course, some translation software products that are optimized for legal documents exist (Non-Patent Document 1), but these are only word-level correspondences such as enhancing specialized legal terminology dictionaries. In addition, the grammar analysis is based on English, and the grammar rules and interpretation rules peculiar to the Japanese legal field cannot be handled. For this reason, there is a risk that a translated sentence with an unknown meaning in which words are enumerated is output, or a translated sentence having a wrong meaning is output.

  The present invention has been devised in view of such a current situation, and an object of the present invention is to provide a technique for optimizing a legal document in advance so that the semantic content is correctly interpreted in subsequent computer processing. It is said.

  In order to achieve the above object, the legal document analysis system according to claim 1 preliminarily describes the relationship between specific elements constituting a sentence and the contents of conversion processing for each element as analysis rules specific to the legal document. The system includes analysis rule storage means to be defined, and executes a morphological analysis process that decomposes input legal document data into morpheme units and a parsing process that specifies a dependency structure between morphemes. When the syntax analysis processing refers to the analysis rule storage means, and the legal document data includes a sentence having a relationship between elements corresponding to the analysis rule, the analysis rule defines It is characterized in that a sentence subjected to the converted processing is generated.

  The legal document analysis system according to claim 2 is the system according to claim 1, wherein the analysis rule defines a parallel relationship between elements, and corresponds to the analysis rule in the parsing process. A feature is that a sentence subjected to a corresponding conversion process is generated.

  The legal document analysis system according to claim 3 is the system according to claim 1 or 2, wherein the analysis rule defines an example relation between elements, and in the parsing process, A feature is that a sentence subjected to corresponding conversion processing is generated for the corresponding sentence.

  The legal document analysis program according to claim 4 is an analysis in which a computer defines in advance the relation between specific elements constituting a sentence and the contents of conversion processing for each element as analysis rules specific to the legal document. A rule storage means, a morpheme analysis means for decomposing input legal document data into morpheme units, and a syntax analysis means for specifying a dependency structure between morphemes, wherein the syntax analysis means includes the analysis rule Referring to storage means, if the legal document data contains a sentence with a relationship between the elements corresponding to the analysis rule, a sentence subjected to the conversion process defined in the analysis rule is generated. It is characterized by doing.

  In the case of the legal document analysis system and program according to the present invention, the sentence conversion processing is executed in accordance with the analysis rule stored in the analysis rule storage means, so that the analysis rule specific to the legal document is defined in advance in the analysis rule storage means By doing so, the structure of the legal document can be optimized, and subsequent translation processing and concept retrieval processing can be made highly accurate.

FIG. 1 is a block diagram showing a contract analysis system 11 including a legal document analysis system 10 according to the present invention.
The contract analysis system 11 includes a Web server 12 and an AP server 14. The AP server 14 includes a legal document analysis unit 16, a document vector generation unit 18, a related clause specification unit 20, and an analysis screen generation unit 22. And an analysis rule storage unit 24, a legal text storage unit 26, and a related literature storage unit 28.

The legal document analysis unit 16, the document vector generation unit 18, the related clause specification unit 20, and the analysis screen generation unit 22 are realized by the CPU of the AP server 14 that operates according to the OS and a dedicated application program.
The analysis rule storage unit 24, the legal text storage unit 26, and the related document storage unit 28 are provided in the external storage device of the AP server 14.

  The above-mentioned legal text storage section 26 contains the main texts of the main laws and regulations (civil law, commercial law, company law, patent law, trademark law, copyright law, unfair competition prevention law, antitrust law, criminal law, civil law law, etc.). And text data of rules are stored.

The related document storage unit 28 stores various related document information for each main article in the main laws and regulations.
FIG. 2 shows an example of this, and it shows a state in which multiple comments (one-point advice by lawyers such as lawyers), precedents and doctrines are linked as related literature for Article 709 of the Civil Code. .
Each related document includes link information with a document title, an outline, other related documents, related images, related moving images, and the like.
The types of related documents are not limited to the above comments, precedents, and theories, but may be contract examples.

  The Web server 12 is responsible for generating and transmitting various screens (Html files) in response to member user requests, and is connected to the client terminals 34 of many member users 32 via the Internet 30. Yes.

A member user 32 who wishes to use the contract analysis system 11 first accesses the Web server 12 from the client terminal 34 and logs in to the contract analysis service.
As a result, the contract document input screen is transmitted from the Web server 12 to the client terminal 34. FIG. 3A shows a contract input screen 40 displayed on the Web browser of the client terminal 34.

The member user 32 inputs each clause 42 of the contract on the contract input screen 40 using a keyboard or copy and paste, and clicks the send button 44.
As a result, the contract data is transmitted from the client terminal 34 to the Web server 12.

Hereinafter, the processing procedure in the system 11 will be described with reference to the flowchart of FIG.
First, the contract document data (text data) 46 transferred from the Web server 12 to the AP server 14 is subjected to morphological analysis processing in the legal document analysis unit 16 (S10).
Here, “morpheme analysis” refers to a process of decomposing a sentence described in a natural language into morphemes, which are the smallest meaningful language units, and identifying each part of speech.

Next, the legal document analysis unit 16 refers to the analysis rule storage unit 24 and applies analysis rule application processing to the contract data subjected to morphological analysis (S12).
Hereinafter, the analysis rule application process will be described in detail.

[1. Analysis rule application processing in syntax analysis]
In text mining, a syntax analysis process that specifies a dependency relationship between each morpheme constituting a document is indispensable, and the legal document analysis unit 16 naturally executes this syntax analysis process.
At this time, the legal document analysis unit 16 executes a special process for extracting a dependency structure from a viewpoint different from normal syntax analysis for a sentence that matches the rules stored in the analysis rule storage unit 24.

(1) Processing related to parallel relations When there is a sentence “D that is A, B, and C” in the contract document, each element is simply AND in general parsing, such as “A + B + C = D”. It is interpreted as a connected parallel relationship of equality, and is converted into a sentence “D which is A, B, and C”.
On the other hand, as shown in FIG. 5A, the legal document analysis unit 16 extracts the correct syntax according to the rule specific to the legal field of “A + B / C” for “A, B and C”. The following two sentences are generated.
(i) “D that is A and B”
(ii) “D that is C”
For this purpose, the analysis rule storage unit 24 defines the following analysis rules.
■ Connection relationship of “A, B and C” → “A + B / C”

Similarly, the sentence “D that is A, B, or C” is also interpreted as “A + B + C = D” in general parsing and converted to the sentence “D that is A, B, and C”. become.
On the other hand, as shown in FIG. 5 (b), the legal document analysis unit 16 can extract the correct syntax “A / B + C” for the part “A, B, or C”. Generate a statement.
(i) “D that is A”
(ii) “D that is B and C”
For this purpose, the analysis rule storage unit 24 defines the following analysis rules.
■ Connection relationship of “A or B or C” → “A / B + C”

(2) Processing related to example relations In legal documents, the words “other” and “other” are interpreted differently.
For example, the sentence “Do not hold the land, sea and air force“ other ”” means that the land, sea and air force is only an example of other forces, whereas “wage, salary“ other ”. In the case of “when there is equivalent income,” means that income equivalent to this is in parallel with wages and salaries.

For this reason, the legal document analysis unit 16 converts it into a form that conforms to the meaning of the sentence based on the sentence “Do not hold the land, sea, air force and other forces”. A new sentence is generated and replaced with the original sentence.
In addition, the legal document analysis unit 16 based on the sentence “When there is wage, salary or other income equivalent to this…”, “(1) When there is wage…”, “(2) Salary In some cases, like "..." and "(3) When there is income equivalent to wages / salary ...", in order to correctly express the meaning of the sentence, multiple sentences are generated and replaced with the original sentence.

  On the other hand, when a general parsing engine is used, the difference in the presence or absence of “no” is ignored. As a result, “other” and “other” are processed as words representing the same document structure. For this reason, the sentence “Do not hold the land, naval and air forces and other forces” is input to the subsequent processing as the sentence “Do not hold the land and naval air forces and other forces”. Similarly, the sentence “When there is wage, salary or other income equivalent to this ...” is put into the subsequent processing as the sentence “When there is wage, salary or other income equivalent to this ...”. The

For example, the following analysis rules are defined in the analysis rule storage unit 24 so that the legal document analysis unit 16 can generate a correct sentence in accordance with the above-described legal practice.
■ “Other” → The word located immediately before is an example of the word located immediately after. → “For example” is added to the head of the word located immediately before, and the whole word is enclosed in parentheses. Move back. (1) “Other” → Parallel relationship between the word immediately before and the word immediately after → Create a sentence using each of them independently.
The legal document analysis unit 16 follows the analysis rules when performing the syntax analysis process, thereby correctly deriving the dependency structure between the elements as shown in FIGS. 6 (a) and 6 (b), and creating a new sentence. Generation can be performed.

(3) Exception handling related to conditional relations Generally, "case" and "time" can be interpreted in the same meaning, but when used in the same sentence in the field of law, "case" ”And“ time ”are interpreted as“ small condition ”.
For example, the sentence “If one of the parties violates this Agreement and the violation is not corrected after 30 days, the other party may terminate this Agreement” In this case, it is required that “30 days have passed” does not meet the condition, and “contract violation” occurs as a major premise.

For this reason, the analysis rule storage unit 24 defines the following analysis rules.
■ When "Case" and "Time" exist in the same sentence → "Precursor of" Case "" becomes a major condition, and "Precedence of" Case "" becomes a minor condition. When performing the parsing process, it is possible to correctly derive the dependency structure between the elements according to the analysis rule.

[2. Application processing of analysis rules by adding annotation information]
In addition to parsing, there are many cases where special consideration is required for specific terms when evaluating legal documents.
For this reason, the legal document analysis unit 16 adds a corresponding annotation information to a term that matches the analysis rule stored in the analysis rule storage unit 24, or executes a process of converting it into a specific character string.

(1) Processing related to numerical values + range specification words In general syntax analysis, "range specification words" such as "more than", "less than", "less than", and "greater than" are added to specific numerical values. For words, the numerical values are not extracted, and the specified range cannot be extracted correctly.
However, this range designation is extremely important when evaluating legal documents.
For this reason, the analysis rule storage unit 24 defines the following analysis rules.
■ "More" → Includes the previous value ■ "Following" → Includes the previous value ■ "Over" → Does not include the previous value ■ "Less than" → Does not include the previous value

In accordance with this analysis rule, the legal document analysis unit 16 assigns annotation information “includes 18 years old” to the part of the document to be analyzed that includes the word “18 years or older” as follows: .
● “… Members over the age of 18 do not require parental consent.”
<Including 18 years old>

Alternatively, when the word “under the age of 18” is included in the document to be analyzed, the legal document analysis unit 16 adds annotation information “not including the age of 18” to the portion as follows.
● “However, this does not apply if the relative living together is under the age of 18.”
<Excluding 18 years old>

  Clarify that the interpretation rules for this "number + range specification word" are not specific to the legal field and are consistent with common sense judgment, but the legal interpretation is not different from general common sense. The significance is recognized.

(2) Processing related to revocation of effect In a general parsing engine, “cancel” and “cancel” are interpreted as words having the same meaning as “cancel”. Differences occur. In other words, “cancellation” has the effect of cancellation going back to the beginning of the contract, while “cancellation” has the effect of cancellation from the time of cancellation to the future.
In addition, “cancel” can be said to be synonymous with the expression “deemed not to be new” in terms of its meaning.

Based on the above, the analysis rule storage unit 24 defines the following analysis rule.
■ "Release" → "Returning effectiveness"
■ "Cancellation" → "Effectiveness toward the future"
■ “Consider that it was not the beginning” → “= Cancel”

In accordance with this analysis rule, the legal document analysis unit 16 assigns annotation information “retroactively lost” to the portion of the analysis target document that includes the word “cancel”.
● “… Can cancel the contract.”
<Retroactive loss of effectiveness>

Further, when the word “cancel” is included in the analysis target document, the legal document analysis unit 16 gives the annotation information “loss of effectiveness toward the future” as follows.
● "... within 3 months after Party A or Party B has indicated the intention to cancel ..."
<Loss of efficacy toward the future>

  Furthermore, if the analysis target document includes the expression “consider that it was not the beginning”, the legal document analysis unit 16 replaces it with “cancel”.

(3) Handling of urgent expressions In legal documents, there are multiple terms that indicate temporal urgency such as “immediately”, “promptly”, and “without delay”. Is present.
Specifically, “immediately” is the most urgent, “promptly” is less impulsive, and in the case of “without delay,” there is a nuance that allows some delay due to the existence of a reasonable reason. include.

For this reason, the analysis rule storage unit 24 defines the following analysis rules.
■ "Immediately" → "Urgentness: High"
■ “Promptly” → “Emergency: Medium”
■ “No delay” → “Urgentness: Low”

In accordance with this analysis rule, if the word “immediately” is included in the document to be analyzed as follows, the legal document analysis unit 16 assigns annotation information of “urgency: high” to that portion.
● “In the case of…, Party A shall immediately notify to that effect”
<Emergency: High>

In addition, when the word “promptly” is included in the analysis target document, the legal document analysis unit 16 assigns annotation information of “emergency level: medium” as follows.
● “... I promise to evacuate quickly”
<Emergency: Medium>

Further, when the word “without delay” is included in the analysis target document, the legal document analysis unit 16 assigns the annotation information “urgent: low” as follows.
● “In case of…, Party A must contact you without delay.”
<Urgent level: Low>

(4) Processing related to prohibited expressions In legal documents, even if it means the same prohibition, the expression "Do not do A" and the expression "Cannot do A" indicate the degree of prohibition. Are interpreted differently.
For this reason, the analysis rule storage unit 24 defines the following analysis rules.
■ "Do not do" → "strong prohibition"
■ "I can't do" → "Weak ban"

In accordance with this analysis rule, the legal document analysis unit 16 assigns “strongly prohibited” annotation information to the portion of the analysis target document that includes “not to be”.
● “... Do not leak trade secrets of the other party to each other.”
<Strong prohibition>

Further, when the expression “cannot be” is included in the analysis target document, the legal document analysis unit 16 assigns “weak prohibition” annotation information as follows.
● “… Cannot be disclosed to a third party unless consent is obtained.”
<Weak ban>

(5) Processing related to business-related terms Generally, “business” and “sales” are interpreted as synonyms, and in the parsing engine, they are unified into either one as synonyms. However, when used in legal documents, the former means “a general term for actions that are repeatedly performed for a certain purpose”, whereas the latter is interpreted as meaning “a business for profit”. There is a clear difference between the two.
In addition, the expression “as business” is understood as synonymous with “as business” in a specific legal field (for example, patent law).

Based on the above, the analysis rule storage unit 24 defines the following analysis rule.
■ “Business” ≠ “Sales”
■ “As a business” → “As a business”

  In accordance with this rule, the legal document analysis unit 16 refrains from replacing either expression with the other expression if the word “business” or “sales” is included in the document to be analyzed. Maintain expression.

Further, if the word “as business” is included in the document to be analyzed, the legal document analysis unit 16 replaces it as “business” as follows.
● “... I think about the value of patented invention as a business ...”
↓
“… I think about the compensation when the patented invention is implemented as a business…”

The contract document data that has undergone processing such as morphological analysis, syntax analysis, addition of annotation information, and character string conversion in the legal document analysis unit 16 is passed to the document vector generation unit 18 where it is converted into a document vector for each contract clause. (S14).
For example, as shown in FIG. 7, when a contract includes a plurality of clauses, the document vector generation unit 18 generates a plurality of document vectors 47 for each clause.

Here, “document vectorization” is a technique for expressing the characteristics of a document as a vector based on the combination of keywords included in each document and the appearance frequency. By inputting each text into an existing vectorization engine, Calculated.
However, the document vectorization process is limited to the text originally included in the contract and the character string replaced by the legal document analysis unit 16, and the annotation information added by the legal document analysis unit 16 Is excluded.
As described above, sentences that require parsing specific to legal documents have been converted into sentences optimized by the legal document analysis unit 16 in advance, so that the correct document in line with the meaning of the sentence in the document vectorization process A vector is generated.

  The legal document analysis unit 16 performs the same morphological analysis processing (S10) and analysis rule application processing on the legal text stored in the legal text storage unit 26 in parallel with the analysis processing of the contract data 46 described above. (S12) is executed, and the analyzed legal text data is passed to the document vector generation unit 18.

  Receiving this, the document vector generation unit 18 first executes a vectorization process for generating a document vector in units of legal provisions as shown in FIG. 7 (S16). At this time, if the text includes a plurality of clauses, a document vector is generated for each clause.

Next, the document vector generation unit 18 compares the document vectors in units of sentences, collects those whose angles between each other are within a threshold value, and executes a grouping process for merging each article into one document (S18). .
At this time, a grouped document 48 that is collected according to the degree of similarity of the description content is generated purely according to the degree of similarity of the description contents, beyond the boundaries of the jurisdictions such as the “civil law” and the “unfair competition prevention law”.

Next, the document vector generation unit 18 performs the vectorization process again on each grouped document 48 (S20). As a result, a document vector 49 for each grouped document 48 is obtained.
Finally, the document vector generation unit 18 passes the vector data 49 of each grouped document 48 and the vector data 47 of each contract clause to the related clause identification unit 20.

In response to this, the related clause specifying unit 20 matches each contract clause vector 47 with the vector 49 of the legal clause grouped document 48 (S22), and is included in the grouped document 48 having the smallest angle between them. Each legal article to be identified is identified as a related article of the contract clause.
FIG. 7 shows an example in which the civil law article 709 and the unfair competition prevention law article 4 are specified as related articles with respect to article 21 of the contract.
The related clause information 50 is passed to the analysis screen generation unit 22.

The analysis screen generating unit 22 refers to the related document storage unit 28, and extracts related documents (comments, precedents, academics, etc.) related to the related articles (S24).
Next, the analysis screen generator 22 generates a contract analysis screen (S26), and passes the analysis screen data 52 to the Web server 12.

The analysis screen data 52 is transmitted from the Web server 12 to the client terminal 34 and displayed on the Web browser.
FIG. 3B shows a contract analysis screen 60 displayed on the Web browser, and related clauses 62 are listed for each clause 42 in the contract. Each related clause 62 is displayed in descending order of similarity to the document vector of the contract clause.

In each related article 62, buttons (comment button 64, case button 66, and doctrine button 68) indicating the existence of each related document are displayed.
Here, when the user 32 clicks, for example, the button 66 of “judicial precedent 1” in Article 4 of the non-compete law, a pop-up window 70 displaying the title, outline and link information of the case associated with the article is displayed on the screen 60 Is displayed.

Also, if the text of the contract clause 42 contains a character string with annotation information, the underline and highlighting are highlighted and the annotation information is clearly indicated so that its existence is clear. Is done.
In the figure, “promptly” in the contract clause is underlined and an annotation tag 72 of “<emergency: medium>” is displayed.

  In this way, on the contract analysis screen 60, a list of related clauses for each contract clause 42 and related literature information related to each related clause are presented, and an annotation tag 72 is added to a word that should be particularly noted in legal interpretation. Therefore, even the user 32 who is not familiar with the law can surely recognize the location of the problem.

  As described above, the legal document is optimized prior to the document vectorization process (optimization of parsing, character string conversion), so an accurate document vector can be generated. , Matching accuracy can be increased.

  In addition, when vectorizing legal provisions, documents are first vectorized in units of individual articles, and then grouped across legal boundaries according to the similarity of the vectors. It becomes possible to exhaustively extract the text from multiple different jurisdictions.

  In the above, analysis, vectorization, grouping, and re-vectorization of the legal provisions are executed in parallel with the analysis and vectorization of the contract data 46. There is an advantage that matching processing is performed based on the above.

However, the present invention is not limited to this.
That is, analysis, vectorization, grouping, and revectorization are executed in advance for the legal text, and the processing results are stored in a predetermined storage means, and the related text is specified at the time of matching processing with the document vector of the contract. The system may be configured so that the unit 20 refers to this.
In this case, the legal document analysis unit 16 and the document vector generation unit 18 perform analysis, vectorization, grouping, and re-vectorization of all the legal provisions at the timing when any law is revised, thereby obtaining data. Maintain the freshness of

Also, in the above, when matching the vector 47 of each contract clause with the vector 49 of the legal text grouped document 48, one grouped document 48 that minimizes the angle between them is identified as a similar text group. However, the present invention is not limited to this example.
For example, a plurality of grouped documents having a vector 49 in which an angle between each contract clause vector 47 falls within a predetermined threshold is specified, and the legal text included in each grouped document 48 is changed to the contract. It can also be recognized as the relevant clause of the clause.
Alternatively, the angle between the vector 47 of each contract clause and the vector 49 of each grouped document 48 is aligned in ascending order, and the legal text included in the top N% grouped document 48 is changed to the relevant text of the contract clause. May be certified. In this case, the value of “N” is set in advance by the user in 1% increments.

It is a block diagram which shows the contract analysis system system containing the legal document analysis system which concerns on this invention. It is a figure which shows an example of the data stored in the related literature memory | storage part. It is a figure which shows a contract input screen and a contract analysis screen. It is a flowchart which shows the process sequence of this system. It is a figure which shows the application result of the analysis rule regarding a parallel relationship. It is a figure which shows the application result of the analysis rule regarding an example relationship. It is a schematic diagram which shows the outline | summary of a document vectorization process and a matching process.

10 Legal document analysis system
11 Contract analysis system
12 Web server
14 AP server
16 Legal Document Analysis Department
18 Document vector generator
20 Related Article Identification Department
22 Analysis screen generator
24 Analysis rule storage
26 Legal Text Storage Department
28 Related literature storage
30 Internet
32 member users
34 Client terminal
40 Contract input screen
42 Contract terms
44 Submit button
47 document vector
48 Grouped documents
49 Group unit document vector
50 Related Article Information
52 Analysis screen data
60 Contract analysis screen
62 Related Articles
64 Comment button
66 Legal Button
68 Theory Button
70 pop-up window
72 Annotation tags

In order to achieve the above-mentioned object, the legal document analysis system according to claim 1 is provided when there is a relationship of “element D that is element A, element B, and element C” between specific elements constituting the sentence. , “Element A or Element B” between an analysis rule that specifies that a sentence “element D that is element A and element B” and “element D that is element C” should be generated, and a specific element that constitutes the sentence Or, when there is a relationship of “element D that is element C”, an analysis rule that specifies that statements of “element D that is element A” and “element D that is element B and element C” should be generated is stored an analysis rule storage means is, the input legal document data a legal document analysis system comprising a decomposing morphological analysis unit into morphemes, and parsing means for identifying the dependency structure between the morphemes, the syntax analysis means, above When the analysis rule storage means is referenced and the legal document data contains a sentence having a relationship between elements corresponding to the analysis rule, the sentence subjected to the conversion process defined in the analysis rule It is characterized by generating.

The legal document analysis program according to claim 2 causes the computer to execute “element A and element B when there is a relationship“ element A, element B, and element C ” between specific elements constituting the sentence. An analysis rule that specifies that a sentence of “element D that is element B” and “element D that is element C” should be generated, and “element A, element B, or element C between specific elements constituting the sentence” An analysis rule memory storing an analysis rule that specifies that a sentence “element D as element A” and “element D as element B and element C” should be generated when the relationship “element D” exists Means, a morpheme analysis means for decomposing input legal document data into morpheme units, and a program for functioning as a syntax analysis means for specifying a dependency structure between each morpheme, wherein the syntax analysis means stores the analysis rule storage unit If the legal document data contains a sentence with the relationship between the elements corresponding to the analysis rule, generate a sentence subjected to the conversion process defined in the analysis rule. It is characterized by.

Claims (4)

It includes an analysis rule storage means for predefining the relationship between specific elements constituting a sentence and the content of conversion processing for each element as an analysis rule specific to a legal document,
Morphological analysis processing that decomposes input legal document data into morpheme units,
A parsing process that identifies the dependency structure between each morpheme,
A legal document analysis system that implements
When the syntax analysis process refers to the analysis rule storage means, and the legal document data includes a sentence having a relationship between elements corresponding to the analysis rule, it is defined in the analysis rule. A legal document analysis system that generates a sentence subjected to conversion processing. The above analysis rule defines the parallel relationship between elements,
2. The legal document analysis system according to claim 1, wherein a sentence in which a corresponding conversion process is applied to a sentence corresponding to the analysis rule is generated in the parsing process. The analysis rule defines an example relationship between elements,
3. The legal document analysis system according to claim 1, wherein a sentence in which a corresponding conversion process is applied to a sentence corresponding to the analysis rule is generated in the parsing process. Computer
Analysis rule storage means for predefining the relationship between specific elements constituting a sentence and the content of conversion processing for each element as an analysis rule specific to a legal document,
Morphological analysis means for decomposing input legal document data into morpheme units,
A syntax analysis means for specifying a dependency structure between each morpheme,
A program that functions as
The syntax analysis means refers to the analysis rule storage means, and if the legal document data includes a sentence having a relationship between elements corresponding to the analysis rule, the syntax analysis means is defined in the analysis rule. A legal document analysis program characterized by generating a sentence that has undergone conversion processing.

Images