JP2011257817A - Patent specification analyzer and text analyzer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To precisely extract case components defining a patent claim.SOLUTION: Component separation means 11 separates a claim including plural components into several components. Modification structure analyzing means 12 executes morphological analysis on the respective components to break down into a segment including a word and a parse of the word and analyzes the modification structure of the segment. Segment restructuring means 13 restructures the modification structure of the segment into a tree-like segment structure. A pattern 15 of plural case components and a pattern 16 excluded from the case components are restructured in a tree-like configuration with segments including the word and the word parse. Case component extraction means 14 compares the segment structure to the plural case component patterns 15 and the patterns 16 exclusive case component, and extracts the same as a case component.

Description

The present invention relates to a patent specification analysis device and a text analysis device, and more particularly to a patent specification analysis device that analyzes and displays a patent claim, and a text analysis device that analyzes and displays conditions such as a contract. is there.

2. Description of the Related Art Conventionally, in a patent specification analysis apparatus and a text analysis apparatus, a technique for counting and displaying the degree of limitation of an invention defined in the scope of claims in a patent specification is known.

Patent Document 1 describes a technique of extracting a character string described in a specific column in a patent specification converted into electronic data, and then counting all the case components and displaying them as the degree of limitation of the invention. Yes.

JP 2009-259154 A

However, the conventional patent specification analyzer and the sentence analyzer are
There were problems (A) to (C).

(A) Since the case component is extracted using only the character string of the cue phrase as an index, even if the part of speech is completely different, it matches the cue phrase character string, and the character string that is not the case component is mistakenly used as the case component. There was a possibility of extracting as.

(B) Case components were extracted using the character string of the clue phrase as an index. Therefore, there is a possibility that the phrase is erroneously determined not to be a case component due to a ending change of the part of speech included in the phrase of the case component.

(C) For example, a pattern that is not a case component is extracted using only the duplication of the character string of the cue phrase as an index, and this is excluded as a case component. If the active and passive voices are described in duplicate, there is a risk that it will be erroneously judged as not being duplicated.

The patent specification analysis apparatus of the present invention includes a component requirement classifying unit that classifies a claim having a plurality of component requirements for each component requirement, a morphological analysis of the component requirement, and a word and a part of speech of the word. A dependency analysis means for decomposing the phrase into phrases and analyzing the dependency of the phrase; a phrase structure configuration means for configuring the dependency of the phrase into a tree-like phrase structure; and a word and a part of speech of the word A plurality of case component patterns and case component exclusion patterns configured in a tree shape with clauses, and a case component that compares the clause structure with the plurality of case component patterns and case component exclusion patterns and extracts them as case components And an extraction means.

According to the patent specification analysis apparatus and the sentence analysis apparatus of the present invention, there are the following effects (A) and (B).

(A) When extracting as a case component in comparison with a phrase structure and a plurality of case component patterns and case component exclusion patterns, it is not a character string of a clue phrase but a phrase having a word and a part of speech of the word The tree-like phrase structure is compared. Therefore, the case component can be extracted more accurately than the character string of the clue phrase.

(B) Since the case component pattern and the case component exclusion pattern are determined based on the original form of the verb, for example, even if the verb is described as active and passive, the same is used. It can be judged as a thing.

FIG. 1 is a diagram illustrating a schematic configuration of a patent specification analysis apparatus according to Embodiment 1 of the present invention. FIG. 2 is a diagram showing a case component pattern (part 1) of FIG. FIG. 3 is a diagram showing a case component pattern (part 2) of FIG. FIG. 4 is a diagram illustrating a case component exclusion pattern in FIG. 1. FIG. 5 is a flowchart showing the operation of the patent specification analyzer of FIG. FIG. 6 is a flowchart showing the operation of the component requirement classifying means of FIG. FIG. 7 is a flowchart showing the operation of the dependency analyzing means of FIG. FIG. 8 is a flowchart showing the XML operation in the phrase structure construction unit of FIG. FIG. 9 is a flowchart showing the treeing operation in the phrase structure constructing means of FIG. FIG. 10 is a flowchart showing the operation of the case component extraction means of FIG. FIG. 11 is a diagram showing a dependency analysis result of the phrase in FIG. FIG. 12 is a diagram showing dependency analysis XML by the processing shown in FIG. FIG. 13 is a diagram showing the tree-like phrase structure in FIG. FIG. 14 is a diagram showing a tree-like phrase structure from which the case components in FIG. 1 are extracted. FIG. 15 is a diagram illustrating a schematic configuration of a contract document analysis apparatus according to the second embodiment of the present invention. FIG. 16 is a flowchart showing the operation of the contract document analyzer of FIG.

  Modes for carrying out the present invention will become apparent from the following description of the preferred embodiments when read in light of the accompanying drawings. However, the drawings are only for explanation and do not limit the scope of the present invention.

(Configuration of Example 1)
FIG. 1 is a diagram illustrating a schematic configuration of a patent specification analysis apparatus according to Embodiment 1 of the present invention.

  The patent specification analysis apparatus 10 includes a component requirement classification unit 11, a dependency analysis unit 12, a phrase structure configuration unit 13, and a case component extraction unit 14. The case component extraction means 14 includes a case component pattern 15 and a case component exclusion pattern 16.

  The component requirement classifying unit 11 classifies the sentence 21 of the claim by the component requirement, and outputs the sentence 22 of the claim divided into each component requirement. The dependency analysis unit 12 performs dependency analysis on the sentence 22 of the claim divided into each constituent requirement, and outputs a dependency analysis result 23 of the phrase. The phrase structure constructing means 13 converts the phrase dependency analysis result 23 into the XML format, and then forms a tree according to the phrase dependency, and outputs a tree-like phrase structure 24. The tree-like phrase structure 24 is configured in a tree shape by phrases having a word and a part of speech of the word. The case component extraction means 14 compares the tree-like phrase structure 24 with the case component pattern 15 and the case component exclusion pattern 16 and outputs a tree-like phrase structure 25 from which the case components are extracted. The case component pattern 15 and the case component exclusion pattern 16 are configured in a tree shape with phrases having words and parts of speech of the words.

FIG. 2 is a diagram showing the case component pattern (part 1) of FIG. 1. The left side shows what the case shows, and the right side shows the case component pattern 15 having a tree-like phrase structure. It is shown. In the tree-like clause structure on the right side, one line corresponds to one clause. The tree-like phrase structure of the case component pattern 15 is configured in a tree form by phrases having words and parts of speech of the words.

In the first line of case 1 indicating the object, [noun phrase] enclosed in parentheses indicates a word string ending with a noun. “O (participant)” described next to “noun phrase” indicates that the word is “” and that the part of speech of this word “” is a particle. “-D” described next to “O (particle)” indicates the position of the previous clause on which this clause is dependent.

In the second line of the case 1 indicating the subject, [verb phrase] enclosed in parentheses indicates a phrase having the word of the verb. “− +” Described next to “verb phrase” indicates the position of the original phrase on which this phrase is dependent. “− +” In this phrase indicates the same position as the position of “−D” in the immediately preceding phrase, and thus indicates that the immediately preceding phrase is being modified. “-*” Described after “− +” indicates whether the phrase after the verb phrase is composed of words and how the dependency structure after the verb phrase is structured. It shows no. Case No. 1 indicating an object shows a pattern in which a phrase composed of a particle “wo” next to [noun phrase] depends on a phrase having a word as a verb.

Dependent phrase is not a phrase that appears next to “[noun phrase]”, as indicated by “Perform Laplacian filter processing”, but is dependent on Japanese grammar. It is a phrase. Therefore, the sentence “Laplacian filter processing is performed for each RGB signal” is included in this dependency pattern.
Similarly, Case No. 2 indicating an object indicates a pattern in which a phrase composed of the particle “ga” after [noun phrase] is dependent on a phrase having a word that is a verb.
Case 1 indicating a condition indicates a pattern in which a phrase composed of the particle “ni” next to [noun phrase] depends on a phrase having a word as a verb.
Case 2 indicating a condition indicates a pattern in which a phrase composed of the particle “ni” and the particle “more” is followed by a phrase having a word as a verb after [noun phrase].
Case 3 indicating a condition indicates a pattern in which a phrase composed of a verb “based” and a particle “te” following an arbitrary phrase depends on a phrase having a word as a verb.
Case No. 4 indicating a condition shows a pattern in which a phrase consisting of a verb “Accept” after an arbitrary phrase and a particle “te” depends on a phrase having a word as a verb.

Case No. 5 indicating a condition indicates a pattern in which a phrase consisting of a verb “corresponding” and an auxiliary particle “te” following an arbitrary phrase depends on a phrase having a word as a verb.

Case 6 showing a condition shows a pattern in which a phrase consisting of a noun “if”, a particle “ni”, and a particle “ha” depends on a phrase having a word as a verb after an arbitrary phrase. ing.
Case 1 indicating a time indicates a pattern in which a phrase consisting of a noun “to” next to an arbitrary phrase depends on a phrase having a word as a verb.

FIG. 3 is a diagram showing the case component pattern (part 2) of FIG. 1. Like FIG. 2, the left side shows what the case shows, and the right side has a tree-like phrase structure. A case component pattern 15 is shown. As in FIG. 2, in the tree-like clause structure on the right side, one line corresponds to one clause.
Case No. 1 indicating the starting point shows a pattern in which a phrase composed of the particle “kara” after [noun phrase] depends on a phrase having a word as a verb.
Case No. 1 indicating a landing point shows a pattern in which a phrase consisting of the particle “he” next to [noun phrase] depends on a phrase having a word as a verb.
Case No. 2 indicating a landing point indicates a pattern in which a phrase consisting of the particle “ni” after [noun phrase] depends on a phrase having a word as a verb.

  In case 3, the phrase that consists of the particle "ni" after [noun phrase] depends on the phrase that consists of the verb "versus" and the particle "te", and has a word that is a verb. It shows the pattern that depends on the phrase.

  Case 1 that shows mediation is that a phrase consisting of the particle "wo" next to [noun phrase] depends on a phrase consisting of the verb "me" and the particle "te", and further has a word that is a verb. The pattern that is dependent on is shown.

  In case 2 that shows mediation, the phrase consisting of the particle “ni” next to the [noun phrase] depends on the phrase consisting of the verb “mediate” and the particle “te”, and further has a word that is a verb. It shows the pattern that depends on the phrase.

Case No. 1 indicating a use role shows a pattern in which a phrase composed of a particle “as” and a particle “no” following a [noun phrase] depends on a phrase having a word as a verb.
Case No. 1 indicating a state shows a pattern in which a phrase composed of a noun “state” and a particle “de” following [noun phrase] depends on a phrase having a word as a verb.

  The case 1 that shows the raw material / material is that the phrase consisting of the particle "kara" after the [noun phrase] is dependent on the phrase consisting of the verb "nar", and the pattern depends on any phrase. Show.

In case 2 that shows the raw material / material, the phrase consisting of the particle "O" after the [noun phrase] depends on the phrase consisting of the verb "has", and the pattern depends on any phrase. Show.
Case No. 1 showing a comparison shows a pattern in which a phrase consisting of the particle “no” after [noun phrase] depends on a phrase having an adjective word.
FIG. 4 is a diagram illustrating a case component exclusion pattern in FIG. 1.

  The case component exclusion pattern 16 includes four patterns that exclude features, two patterns that exclude repetitions described above, and four patterns that exclude the dependency of the particle “” on a specific verb. . The tree-like phrase structure of the case component exclusion pattern 16 is constituted by a tree having phrases having words and parts of speech of the words.

The feature part 1 is that the phrase consisting of the particle “wo” after [noun phrase] depends on the verb “having”, and the phrase consisting of the particle “wo” next to the noun “ko”. Dependent on the phrase consisting of the particle "wo" next to "feature", depending on the verb "do", and at the end of the sentence of the claim indicated by "... apparatus / method." The pattern depending on the object is shown.

The feature part 2 is that the phrase consisting of the particle "wo" next to the [noun phrase] depends on the verb "preparation", and depends on the phrase "no" next to the noun "ko". Dependent on the phrase consisting of the particle "wo" next to "feature", depending on the verb "do", and at the end of the sentence of the claim indicated by "... apparatus / method." The pattern depending on the object is shown.

Characteristic part 3 is that the phrase consisting of the particle "wo" next to [noun phrase] depends on the verb "has", and further, "... device / method." And the pattern which depends on the object of a claim is shown.

The feature part 4 is that the phrase consisting of the particle "wo" next to [noun phrase] depends on the verb "preparation", and further, "... apparatus / method." And the pattern which depends on the object of a claim is shown.

Each of the feature parts 1 to 4 is a pattern including a verb depending on a clause indicating the subject of the patent. This pattern is extremely formal and routine and does not affect the technical scope of the patent. Furthermore, the case component limits the verb for the constituent requirement. Therefore, the part which limits the verb with respect to the object of the patent excludes this from the case component even if there is a portion which formally matches the pattern of the case component.

  Repetition by the above ・ Part 1 shows a pattern in which a phrase structure having the noun “above” at the beginning, the particle “ni” and the particle “yori” at the end depends on a phrase having a word that is a verb. Yes.

  Repetition by the above (2) shows a pattern in which the phrase structure having the noun “above” at the head and the particle “ga” at the end is further dependent on the phrase having the word that is a verb.

The repetition by the above-mentioned 1 and 2 are further that the noun phrase following the noun “above” is the same as the noun phrase already appearing, and the verb phrase that appears here is the original form , It is determined that it is the same as the original form of the verb phrase that has already appeared. Thereby, even if the verb phrase described in the active voice is described again in the passive voice, it can be detected.

The above repetitions 1 and 2 are duplicated parts for clarifying the scope of the claims, and need not be described. If this is extracted as a case component, the case component will change depending on the description format, which is not desirable. Therefore, this is excluded from the case components.

Dependency on a specific verb No. 1 shows a pattern in which a phrase consisting of a particle “no” next to a noun phrase depends on the verb “do”. The verb “do” may change in ending. In this case, it is formally applicable to the case pattern indicating the object, but it cannot actually be a case.

The verb “do” does not have a specific meaning, and the meaning of the verb is determined for the first time only when the phrase consisting of the particle “wo” follows the noun phrase. If the verb “do” is described next to the noun phrase, the same meaning can be expressed, but it is not desirable because the number of case components changes by rewriting in this way.

For example, consider the sentence “Conversion means for performing RGB conversion processing on received data”. Formally, “RGB conversion processing” corresponds to a case representing an object, but this sentence can be rewritten as “conversion means for performing RGB conversion processing on received data”. This is not desirable because the case component changes only by the formal change of. Therefore, this pattern is not a case component.

Dependency on a specific verb No. 2 shows a pattern in which a phrase consisting of a particle “wo” next to a noun phrase depends on the verb “execute”. The verb “execute” may have a ending change.

Dependency on a specific verb No. 2 shows a pattern in which a phrase composed of a particle “wo” next to a noun phrase depends on a verb “process”. The verb “process” may have a ending change.

Dependency on a specific verb No. 2 shows a pattern in which a phrase consisting of a particle “wo” next to a noun phrase depends on the verb “se”. The verb “do” may have a ending change.
(Operation of Example 1)
FIG. 5 is a flowchart showing the operation of the patent specification analyzer of FIG.

  When the patent specification analysis apparatus 10 starts processing, in step S1, a sentence 21 of each patent gazette or public gazette claim is obtained from a CSV file containing information on the patent gazette or gazette for each line. To extract.

In step S <b> 2, the sentence 21 of the claim is divided by the component requirement classification unit 11 for each component requirement. In step S <b> 3, the dependency analysis unit 12 performs dependency analysis for each component and converts it into a CSV. In step S4, the dependency structure analyzing unit 13 converts the dependency analysis CSV into XML, and in step S5, the dependency analysis XML is tree-formed according to the dependency. In step S6, a case component is extracted from the tree dependency analysis XML, and the operation of FIG.
FIG. 6 is a flowchart showing the operation of the component requirement classifying means of FIG.

When the process starts, in step S10, the component requirement classification unit 11 counts the component requirement enumeration format delimiter “to” and sets it to Sk1. In step S11, the delimiter “Shi” in the writing format is counted as Sk2. In step S12, the Jepson-format break “characteristic of...,” Is counted as Sj.
In step S13, if Sj is not greater than 0, the process of step S14 is performed, and if Sj is greater than 0, the process of step S17 is performed.

  In step S14, if Sk1 is greater than or equal to Sk2, in step S15, segmentation is performed at the configuration requirement enumeration format. Otherwise, in step S16, segmentation is performed at the segmentation of the writing format. End the process.

  In step S17, if Sk1 is greater than or equal to Sk2, in step S18, the segmentation is performed using the separation of the jepson format and the component requirement listing format. If not, the separation is performed using the separation of the jepson format and the writing format in step S19. And the process of FIG. 6 is terminated.

  The method of writing a claim is roughly classified into a component requirement enumeration format and a writing format. In the component requirement enumeration format, the component requirements are described in noun phrases, and the noun phrases are listed with the particle “to” and the symbol “,” added thereto. Therefore, it is possible to delimit each constituent requirement by delimiting with the particle “to” and the symbol “,”.

In the writing style, the component requirements are described in verbs. Constituent requirements are given “shi” which is a ending change of the verb “sei” and a symbol “,” mainly. Therefore, it is possible to delimit each constituent requirement by delimiting with the verb “shi” and the symbol “,”.

The Jepson format has “characterized by ...” at the end of the known part, and is further combined with a component requirement listing format or a writing format. Therefore, in addition to the end pattern of the configuration requirement in the configuration requirement enumeration format or the writing format, it is possible to delimit each configuration requirement by delimiting with the end pattern of the configuration requirement in the Jepson format.
FIG. 7 is a flowchart showing the operation of the dependency analyzing means of FIG.

  When the process starts, in step S20, the dependency analysis unit 12 reads a sentence 21 of a claim segmented for each component. Over steps S21 to S26, the process is repeated for each configuration requirement.

In step S22, the sentence of the current configuration requirement is morphologically converted to extract words and phrases. Over steps S23 to S25, the process is repeated for each clause included in the current configuration requirement. In step S24, it is determined to which phrase the current phrase is dependent. In step S25, if all the clauses included in the configuration requirements have been processed, the process of step S26 is performed. In step S26, when the processing of all the constituent requirements included in the current claim is completed, the processing of FIG. 7 is ended.
FIG. 8 is a flowchart showing the XML operation in the phrase structure construction unit of FIG.

  When the process starts, in step S30, the phrase structure constructing unit 13 generates an XML data structure including only the root. Over the steps S31 to S40, the input dependency analysis result 23 of the clause in the CSV format is repeated for each configuration requirement. In step S32, a node having a configuration requirement is added to the XML route. Over the steps S33 to S39, the process is repeated for each clause of the constituent requirement. In step S34, a structure node is added to the component. In step S35, a clause node is added to the constituent node. Over steps S36 to S38, the process is repeated for each word in the phrase. In step S37, a word node is added to the clause, and the word reading and the part of speech are recorded. If it is determined in step S38 that all the words in the phrase have been processed, the process in step S39 is performed. If it is determined in step S39 that all the clauses in the configuration requirement have been processed, the process of step S40 is performed. In step S40, when the processing of all the configuration requirements is completed, the processing in FIG.

  FIGS. 9A and 9B are flowcharts showing the treeing operation in the phrase structure constructing means of FIG. FIG. 9A is a flowchart showing the upper process, and FIG. 9B is a flowchart showing the upper process and the recursive process called from the recursive process.

  When the upper process starts, in step S50, the recursive process of (b) is called with the structure at the beginning of the configuration requirement as the target and the structure at the end of the configuration requirement as the designation destination, and the upper process of FIG. finish.

When the recursive process starts, the process is repeated from steps S51 to S55 with the structure from the target structure to the designated structure as the current structure. In step S52, it is determined whether or not the current structure depends on the next structure. If so, the process of step S53 is performed. If not, the process of step S54 is performed. In step S53, the current structure is added to the head of the child node in the next structure. In step S54, the recursive process is recursively called with the next structure as a new target and the dependency destination structure as a new designation destination. If all the steps up to the designated destination structure are repeated in step S55, the processing in FIG. 9B is terminated.
FIG. 10 is a flowchart showing the operation of the case component extraction means of FIG.

When the process is started, the case component extraction unit 14 reads in step S60. Over steps S61 to S67, the process is repeated for all the configuration requirements. Over steps S62 to S66, the processing is repeated in order from the end for all the clauses included in the current configuration requirement. In step S63, it is determined whether or not the pattern is a case component pattern 15. In step S64, it is determined whether or not it is a case exclusion pattern. The process of step S65 is performed only when the pattern is not the case component pattern 15 and the case component exclusion pattern 16, and the process of step S66 is performed otherwise. In step S66, if all the clauses have been processed, the process of step S67 is performed. In step S67, when the processing of all the configuration requirements is completed, the processing in FIG.
FIG. 11 is a diagram showing a dependency analysis result of the phrase in FIG.

  As a rule, one word is written in each line. In the line in which the word is described, a character string that constitutes the word, a reading, a character string that constitutes the word, a part of speech, and detailed information related to the part of speech are separated by tabs.

Words are separated into clauses by lines beginning with “*” asterisks. The line in which “*” asterisk is written at the beginning describes the ID number of the clause, the ID number of the previous clause on which the clause is associated, and “D” separated by a space.
The clause is claused for each constituent requirement by a line having “EOS” at the beginning.
FIG. 12 is a diagram showing dependency analysis XML by the processing shown in FIG.

  A “configuration requirement” node is connected to a root node (not shown). A plurality of “structure” nodes are connected to the “configuration requirement” node. In the attribute of the “structure” node, an ID of a “clause” node that is a child node of the “structure” node and an ID of a “clause” node as a dependency destination are described. This “structure” node is all connected directly under the “configuration requirement” node, and is not a connection reflecting the dependency of the “structure” node.

One “phrase” node is connected to this “structure” node. In the attribute of the “phrase” node, the ID of the “phrase” node and the “phrase” node ID of the dependency destination are described.

A plurality of word nodes are connected to the “phrase” node. The part of speech of the word is described in the attribute of the “word” node, and the reading of the word is described in the value of the “word” node.
FIG. 13 is a diagram showing the tree-like phrase structure in FIG.

A “configuration requirement” node is connected to a root node (not shown). A plurality of “structure” nodes are connected to the “configuration requirement” node. The end of the sibling node of this “structure” node is the “clause” node of the dependency destination.
FIG. 14 is a diagram showing a tree-like phrase structure from which the case components in FIG. 1 are extracted.

The attribute of the “structure” node describes whether it is a case component, the type of this case component, and what pattern the case component was extracted from. For example, the “structure” node on the fifth line is a case component representing a landing point, and indicates that it is extracted by the pattern “˜”.
(Effect of Example 1)

According to the patent specification analysis apparatus of the first embodiment, when extracting a case component in comparison with the phrase structure, the plurality of case component patterns 15 and the case component exclusion pattern 16, the character string of the clue phrase is Rather, it compares tree-like phrase structures composed of words and parts of speech and phrases. Therefore, there is an effect that the case component can be extracted more accurately than the character string of the clue phrase.

(Configuration of Example 2)
FIG. 15 is a diagram illustrating a schematic configuration of a contract analysis apparatus according to the second embodiment of the present invention, and the same elements as those in FIG.

The contract analysis apparatus 10A according to the second embodiment includes clause classification means 11A different from the component requirement classification means 11 included in the patent specification analysis apparatus 10 according to the first embodiment, and excludes case components as illustrated in the first embodiment. It has the same configuration as that of the patent specification analysis apparatus 10 shown in the first embodiment except that it has a case component exclusion pattern 16A different from the pattern 16.
(Operation of Example 2)
FIG. 16 is a flowchart showing the operation of the contract document analyzer of FIG.

  When the process starts, in step S70, the element classification means gives the attribute that is not subject to case extraction when the name of the clause is extracted, and in step S71, the attribute that is not subject to case extraction is extracted by extracting the preface of the clause. Assign the attribute of case extraction to the text of the article. In steps S72 to S74, the processing of all the articles is repeated.

  In step S73, each sentence is separated, and in step S74, a case component is extracted for a portion to which an attribute of case extraction is given. When the processing of all the sentences is finished, the processing of FIG. 16 is finished.

As described above, in the contract, the case component can be automatically extracted as in the patent specification. In addition, the case component can be used as a quantitative index for limiting the effective range of the contract as in the patent specification.
(Effect of Example 2)

According to the contract analysis apparatus 10A of the second embodiment, when extracting a case component in comparison with the phrase structure, the plurality of case component patterns 15 and the case component exclusion pattern 16, the character string of the clue phrase is Rather, it compares tree-like phrase structures composed of words and parts of speech and phrases. Therefore, there is an effect that the case component can be extracted more accurately than the character string of the clue phrase.
(Modification)
The present invention is not limited to the above-described embodiments, and various usage forms and modifications are possible. As this usage pattern and modification, for example, there is the following (a).

(A) In the first and second embodiments, the dependency analysis result is converted into XML having a tree-like structure. However, a binary format may be used as long as the tree-like data structure can be handled.

There is a high correlation between the case component and the acceptance rate of the invalidation trial, and the case component and the infringement acceptance rate in the infringement suit are highly correlated. This means that there is a high correlation between the number of case components and the breadth of the patent's technical scope, and it is possible to quantitatively evaluate the quality of patents using case components. .
The present invention shows a case component automatic extraction apparatus. According to the present invention, it is possible to judge the quality of a patent mechanically and quantitatively without depending on human judgment.

DESCRIPTION OF SYMBOLS 10 patent specification analyzer 10A contract document analyzer 11 component requirement classification means 11A clause classification means 12 dependency analysis means 13 clause structure structure means 14 case component extraction means 15 case component pattern 16, 16A case component exclusion pattern 21 Item sentence 22 Claim sentence divided into constituent elements 23 Clause dependency analysis result 24 Tree-like phrase structure 25 Tree-like phrase structure extracted from case components

Claims (6)

A component requirement classifying means for classifying a claim having a plurality of component requirements for each component requirement;
A morpheme analysis of the constituent requirements, disassembled into phrases having words and parts of speech of the words, and dependency analysis means for analyzing the dependency of the clauses;
Clause structure configuring means for configuring the clause dependency into a tree-like clause structure;
A plurality of case component patterns and case component exclusion patterns configured in a tree shape with clauses having words and word parts of speech;
Case component extraction means for comparing the phrase structure with the plurality of case component patterns and case component exclusion patterns and extracting as case components;
A patent specification analysis apparatus comprising: The component requirement classifying unit classifies the claim into the component requirements according to a pattern corresponding to the statement format after determining the statement format of the claim. Patent specification analyzer. The case component extraction means compares the phrase structure with the plurality of case component patterns and case component exclusion patterns by at least the original form of the first word and the part of speech of the first word, and extracts the case component as a case component;
The patent specification analysis apparatus according to claim 1. The exclusion pattern of the case component is
4. The patent specification analysis apparatus according to claim 3, further comprising a verb depending on a clause indicating a subject of the patent. The exclusion pattern of the case component is
A pattern in which a phrase ending with the particle "wo" depends on the verb "do",
A pattern in which a phrase ending with the particle "wo" depends on the verb "execute",
A pattern in which a phrase ending with the particle "wo" depends on the verb "processing",
4. The patent specification analysis apparatus according to claim 3, wherein the phrase ending with the particle “s” is one of patterns depending on the verb “s”. Element classifying means for classifying a sentence including a plurality of elements for each element;
Dependent analysis means for performing morphological analysis of the element, decomposing the phrase into words and parts of speech of the word, and analyzing the dependency of the phrase;
Clause structure configuring means for configuring the clause dependency into a tree-like clause structure;
A plurality of case component patterns and case component exclusion patterns configured in a tree shape with clauses having words and word parts of speech;
Case component extraction means for comparing the phrase structure with the plurality of case component patterns and case component exclusion patterns and extracting as case components;
A sentence analyzing apparatus characterized by comprising:

Images