JP2005031813A - Abstract preparation supporting system, program, abstract preparation supporting method, patent document retrieving system, and patent document rerieving method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To efficiently and properly prepare an abstract in second a language from a patent document in a first language. <P>SOLUTION: This abstract preparation supporting system comprises: an abstract rule dictionary 2 for specifying a principal part extraction rule for every fixed classification; a document structure analyzing part 11 for analyzing the chapter structure of an electronized text patent description; a principal part selecting part 12 for selecting an extraction rule dictionary according to the classification of the analyzed document structure analytic result, and for selecting a principal part from the document structure analytic result on the basis of the rule of the dictionary; a machine translating part 13 for translating the document structural analytic result into native language; an abstract candidate extracting part 14 for extracting an abstract candidate in the native language from the translated native language translation result for each principal part selected by the principal selecting part; and an abstract edition/display control part 15 for displaying the abstract candidate in the native language among the extracted native language abstract candidate and the text principal part corresponding to the native language abstract candidate or for displaying the native language abstract candidate and the text principal part by associating them with each other, and for preparing the native language abstract including the correction processing of the abstract candidate in native language. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an abstract creation support system, a program, an abstract creation support method, a patent document search system, and a patent document search method used when creating an abstract of a native language from a foreign language patent document or the like.
[0002]
[Prior art]
Patent documents (specifications including claims, drawings, etc.) written in a foreign language (for example, English) for the purpose of providing information on what patents have been filed in countries other than Japan ) It is a common practice to refer to the full text and create Japanese abstracts such as the native language (eg Japanese). The created Japanese abstracts are stored in the database of the search system and are provided for viewing and searching for end users.
[0003]
By the way, when creating a Japanese abstract from an English patent document, at present, the abstract creator reads the entire text of the English patent document for which the Japanese abstract is to be created, and the main part (hereinafter referred to as the main part) from the patent document. To understand what is, and to translate the relevant parts into their native language. The Japanese abstract to be created is described and created within a predetermined Japanese length (for example, a total of about 600 to 800 characters) from the technical field, abstract, and examples.
[0004]
Since all such abstract creation work relies on manual work by the abstract creator, it takes a great deal of time to create the abstract. Also, in the abstract creation work process, before the abstract is created, the abstract creator needs to grasp the contents of the entire English patent document. As a result, the abstract creator needs to be able to read the entire English patent document, but this reading requires a lot of work costs, and the efficiency of the abstract creation work is very poor. Was demanded.
[0005]
In order to improve the efficiency of abstract creation work, several abstract creation support systems have been proposed in recent years.
[0006]
One abstract creation support system is a text summarization system for summarizing sentences, which extracts chapter structures such as technical papers and uses them as abstracts for each section (for example, abstract part) Or processing is performed such as whether to ignore the summarization process without performing the digest process (see, for example, document) (see Patent Document 1).
[0007]
Another abstract creation support system is a document information retrieval device and a document retrieval result display method for retrieving document information, and automatically presents an abstract from the original text when presenting the retrieval result. Present the generated abstract to the user. And it is the structure which takes out the correspondence of an abstract and an original text, highlights the interlocking | linkage of an abstract display and an original text display, and the abstract part in an original text display (refer patent document 2).
[0008]
[Patent Document 1]
Japanese Patent No. 2766261
[0009]
[Patent Document 2]
Japanese Patent No. 2957875
[0010]
[Problems to be solved by the invention]
Therefore, the abstract creation support system as described above is intended for technical papers, etc., and the important sentence extraction processing method for abstract creation is based on statistical information such as the frequency of words used in the document. Based on the method (for example, selection criteria that words that are frequently used are important, and sentences containing important words are important), methods for selecting important sentences based on conjunctions and expressions at the end of sentences (for example, “In short” The selection criteria that sentences using adverbs are important are disclosed.
[0011]
However, these important sentence extraction processing methods are intended for technical papers and the like, and are methods for selecting important sentences according to uniform selection criteria. However, there is a problem that it is impossible to extract a high-quality important sentence that can be determined to be a main part from the viewpoint of patent literature.
[0012]
The abstract creation support system described above is intended to create an abstract for a single language. For example, it does not have a support function for creating a Japanese abstract from an English patent document, so a Japanese abstract cannot be created from an English patent document. There is a problem.
[0013]
In addition, ideally, it is preferable that the abstract is automatically created. However, in order to create a high-quality abstract, it is finally necessary to manually check the abstract creator. For this reason, it is necessary to provide a system that does not automatically create abstracts but supports abstracts.
[0014]
Further, in the abstract creation support system as described above, for example, the “summary” is extracted as a main part according to the chapter structure such as “summary”, “technical field to which the invention belongs”, “examples”, and the like. However, with regard to extraction rule dictionaries such as words and phrases that should be the main part, it is prepared in advance for each category of international patent classification (hereinafter referred to as IPC classification), and the IPC classification of the patent document to be abstracted or “ The contents such as switching the extraction rule dictionary according to the chapter structure such as “technical field” and “example” are not described, and it is difficult to use the system as it is from that point. In addition, there is no description about the cooperation between abstracts and machine translation.
[0015]
Furthermore, since the text created by the abstract creator is a Japanese abstract, it is necessary to extract the main part and translate it into Japanese. Conventionally, when manually translating, even if it is a patent document in the same technical field due to the work load, the abstract creator in another field is often in charge, and if limited to manual translation, translation of technical terms It is very difficult to maintain consistency against
[0016]
Therefore, it is conceivable to use a machine translation system for supporting translation from English patent documents into Japanese. As for machine translation systems, some package software is already on the market. For example, a technical term dictionary for machine translation based on IPC classification is prepared, and which IPC classification is a patent document for machine translation. A system that performs appropriate translation by determining and selectively switching and using a technical term dictionary according to the IPC classification of patent literature has also been proposed (Japanese Patent Laid-Open No. 2003-16063). However, this system is intended for machine translation and does not describe the creation of abstracts.
[0017]
Therefore, from the above description, a foreign language patent specification (herein, a foreign language patent specification includes a claim, and further, a patent related document (practical (Including new models, etc.), and the specification including the scope of claims. Therefore, it is intended for electronic patent application documents that are publicly provided, although the names differ depending on the country) It is conceivable that an abstract is created by a conventional abstract creation system, and the created English abstract is translated into Japanese by an English-Japanese translation system.
[0018]
However, in order for an abstract creator to produce a high-quality Japanese abstract, it is not sufficient to simply read the results translated into Japanese, and the corresponding English source text and the context of the text before and after the text Need to refer. Therefore, a combination of a conventional abstract creation support system and a machine translation system cannot proceed with the creation of abstracts in different languages as a series of work procedures, so a support function for creating Japanese abstracts from English patent documents Cannot be fulfilled sufficiently.
[0019]
In any case, the characteristics and patent abstracts of patent documents according to the technical field are very important documents for grasping technical trends and other practical aspects. It can be said that the real situation is waiting for the realization of a support system for creating abstracts in the native language.
[0020]
The present invention has been made in view of the above circumstances, and an abstract creation support system, program, and abstract creation support method for efficiently and appropriately creating an abstract in a second language from a digitized patent application document in a first language The purpose is to provide.
[0021]
Another object of the present invention is to search for a required first language patent application document from an electronic first language patent application document, and search the first language patent application document from the searched first language patent application document. An object of the present invention is to provide a patent document search system, a program, and a patent document search method for creating a search main part of one language and search candidates (search main part) of a second language and realizing an efficient search work.
[0022]
[Means for Solving the Problems]
(1) In order to solve the above-mentioned problems, the abstract creation support system according to the present invention includes at least the titles of the chapters of the first language required for each fixed category and the explanations described in the chapters of these titles. An extraction rule dictionary that prescribes key phrases such as words and phrases in the first language included, document structure analysis means for analyzing the chapter structure of sentences described in the electronic patent application documents in the first language, and The required extraction rule dictionary is selected according to the classification described in the document structure analysis result analyzed by the document structure analysis means, and from the document structure analysis result based on the title and key phrase specified in the extraction rule dictionary Main part selection means for selecting the main part, machine translation means for translating the document structure analysis result analyzed by the document structure analysis means into the second language, and each main part selected by the main part selection means Abstract candidate extraction means for extracting second language abstract candidates from the second language translation results translated by the machine translation means, second language abstract candidates extracted by the abstract candidate extraction means, and the second Among the main parts of the first language corresponding to the language abstract candidates, at least the second language abstract candidates or the second language abstract candidates and the main parts of the first language are displayed in association with each other. An abstract edit display control means for creating an abstract in the second language including an abstract candidate correction process is provided.
[0023]
According to the present invention, the translation processing is performed based on the main part of the first language and the document structure analysis result of the first language in accordance with the field from the chapter structure of the sentence described in the patent application document of the first language. The second language abstract candidate is extracted from the translation result of the second language, and the second language abstract candidate is revised and the second language abstract is created. It is possible to automatically extract the second language abstract candidate without reading the application documents and create the second language abstract in a short period of time, including correction confirmation by the abstract creator. Abstracts can be created and appropriate abstracts can be created.
[0024]
Note that the abstract can be created by a program recorded in the program or an abstract creation method, and the same effects as the system can be achieved.
[0025]
(2) Further, the patent document search system according to the present invention includes a storage unit that stores a plurality of digitized patent application documents in a first language, and at least one first from the storage unit based on a search condition. A document search means for searching patent application documents in a language, a title of a chapter in the first language that is required at least for each predetermined classification, and a word in the first language included in an explanatory note described in the section of each title, An extraction rule dictionary that prescribes key phrases such as phrases, a document structure analysis unit that analyzes a chapter structure of a sentence described in the digitized first language patent application document searched by the document search unit, and A required extraction rule dictionary is selected according to the classification described in the document structure analysis result analyzed by the document structure analysis means, and the document is based on the title and key phrase specified in the extraction rule dictionary. The main part selection means for selecting the main part to be searched from the structure analysis result, the machine translation means for translating the document structure analysis result analyzed by the document structure analysis means into the second language, and the main part selection means. Search candidate extraction means for extracting search candidates in the second language from the translation result of the second language translated by the machine translation means for each main part to be searched, and the second language extracted by the search candidate extraction means Of the first language corresponding to the search candidate of the second language and the search candidate of the second language, and at least the search main part of the first language or the search main part of the first language and the search of the second language Candidates are displayed in association with each other, or the search part of the first language and the search candidate of the second language, the document structure analysis result of the first language, and the translation result of the second language are selectively displayed. Search result display control means It is.
[0026]
The present invention is configured as described above to search at least one first language patent application document according to a search condition from among a plurality of electronic first language patent application documents in the first language. The configuration described in (1) above can be incorporated into the retrieved patent application documents in the first language, and at least the main parts of the search in the first language and the search candidates in the second language can be extracted and displayed. It is possible to realize a typical search operation.
[0027]
The retrieval main part and retrieval candidates can be extracted by a program recorded in the program or a patent document retrieval method, and the same operational effects as the system can be obtained.
[0028]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0029]
FIG. 1 is a block diagram illustrating an embodiment of an abstract creation support system according to the present invention.
[0030]
This abstract creation support system includes a text description input unit 1 for inputting patent application documents in a first language such as digitized foreign language patent documents, and an extraction rule having predetermined elements for each predetermined category of IPC classification. An extraction rule dictionary storage unit 2 for storing a plurality of extracted extraction rule dictionaries, and an abstract for generating an abstract of a second language such as a native language from a foreign language patent document or the like input from the original text description input unit 1 When the creation process control unit 3 and the abstract creation process control unit 3 are constituted by a CPU, a recording medium 4 for storing a program for executing a series of abstract creation processes, and the abstract creation process control unit 3 An abstract-related information storage database 5 for storing the abstract of the native language created in step 1 and other information related to the abstract, and a buffer memory 6 for temporarily storing necessary information during the abstract creation process, It is constituted by an output unit 7 for outputting the Abstract other necessary information of the created mother tongue. Reference numeral 8 denotes a translation dictionary for storing knowledge on translation processing used at the time of translation processing.
[0031]
Note that the above-mentioned patent application documents in the first language refer to all text information including publicly-available gazettes, publicly-known gazettes, registered patent gazettes, and gazettes related to utility models that have characteristics as publicly provided publications. . In addition, the specification in the patent application document may be a document separate from the claims and the abstract in a strict sense, but in the patent document to be distributed, the specification and the claims and the abstract are not specified. Since they are printed together in order, they are intended to include the claims and summary. Hereinafter, the patent application documents in the first language are collectively referred to as foreign language patent specifications or English patent specifications for convenience of explanation.
[0032]
The original specification input unit 1 is generally a keyboard, a mouse, etc., and inputs various control instructions using the keyboard, and the foreign language patent specification selected by designating a specific area in the foreign language patent specification using the mouse. In addition to input, input of reading information by OCR (Optical character Reader) that reads foreign language patent specifications, input of electronic foreign language patent specifications stored on floppy disk, magnetic tape, magnetic disk, etc. are different It includes the input of foreign language patent specifications in various input forms.
[0033]
The extraction rule dictionary storage unit 2 is provided with each extraction rule dictionary for each predetermined category (for example, for each category obtained by further subdividing electrical, mechanical, chemical, physical, or other fields), for example, for each IPC category at least for each upper category. A predetermined extraction rule is defined in each extraction rule dictionary. An example of the rules of each extraction rule dictionary is composed of, for example, three elements of abstract items, fields, and key phrases, and a specific example is as shown in FIG.
[0034]
The abstract creation processing control unit 3 functionally analyzes the items (document structure) of the foreign language patent specifications input from the original text description input unit 1 and puts them into a certain format that can be recognized for each item. A document structure analysis unit 11 as a document structure analysis unit for creating a description of a tagged document structure based on the document structure, and a document structure analysis result obtained by analyzing the document structure by the structure analysis unit 11 is determined in an extraction rule dictionary. Using a rule / knowledge defined in the translation dictionary unit 8 with respect to a document structure analysis result output from the document structure analysis unit 11 and a main part selection unit 12 as a main part selection means for extracting a main part according to the rules, Machine translation unit 13 serving as a translation processing means for translating into the native language, and the main part of the full text of the foreign language patent specification translated by machine translation unit 13 for each major part that is a result of the principal part extraction result of principal part selecting part 12 In the full text of the translation results in Japanese The abstract candidate extraction unit 14 as an abstract candidate extraction unit for extracting the abstract candidates in the native language, the abstract candidate for the native language extracted from the abstract candidate extraction unit 14, and the foreign language requirement extracted by the main part selection unit 12 A part (original part of the original text) is displayed in association with each other, or the whole document structure analysis result of the foreign language and a translation result of the native language translated by the machine translation unit 13 are displayed in association with each other. There is provided an abstract editing display control unit 15 for generating an abstract of the native language including correction processing by an abstract creator of abstract candidates for the Japanese language.
[0035]
The output unit 7 outputs a translation result which is an output of the machine translation unit 13, displays a foreign language patent specification input from the input unit 1, or displays a created native language abstract, It has the function of displaying the full text of the foreign language patent specification, the translated text and the various other required formats. Usually, display means such as various displays are used. Means, a floppy disk, a magnetic tape, a writing registration means to the magnetic disk, a transmission means for transmitting to other media, and various output forms desired by the user.
[0036]
Next, the operation of the above abstract creation support system, the program according to the present invention, and the abstract creation support method will be described with reference to FIG.
[0037]
First, when a computerized foreign language patent specification, for example, an English patent specification, is input from the source text specification input unit 1 (S1), the abstract creation processing control unit 3 follows a program stored in the recording medium 4. A series of processing is executed. The abstract creation processing control unit 3 can perform processing by a hardware configuration including, for example, a logic circuit, but here, a series of processing is executed according to a program.
[0038]
Generally, as shown in FIG. 3, some of the English patent specifications include Abstract, Claims, Field of Invention (FIELD OF THE INVENTION), Background of Invention (BACKGROUND OF THE INVENTION), etc. In addition to the items (document structure) in the specification, the inventor, the IPC classification code, and the like are described, which are not shown here.
[0039]
Therefore, when the abstract creation processing control unit 3 receives the digitized English patent specification input from the original text description input unit 1, for example, after temporarily storing it in the buffer memory 6, the document structure analysis unit 11 performs the English patent. The document structure analysis process of the specification is performed (S2: document structure analysis function, corresponding to the document structure analysis step). This document structure analysis processing unit 11 analyzes each item (document structure) of a predetermined English patent specification when it is assumed that the format of the patent specification is an XML format, and an XML tag for each item. The document structure is replaced with a tag based on the information, and stored in the buffer memory 6 or the database 5 as a document structure analysis result (S3).
[0040]
FIG. 4 is a diagram showing a part of the document structure analysis result obtained by the structure analysis by the document structure analysis unit 11. Here, the document structure according to each item such as Abstract, Claims, FIELD OF THE INVENTION is expressed in XML format. As shown in the document structure analysis result, a range from the tag <field header = “XXXX”> to </ field> corresponding to the tag is a portion corresponding to the title “XXXX”. Here, the example is based on the premise that the format of the patent specification is the XML format, but in the case of other formats, each item (document structure) of the English patent specification is analyzed, and each item is analyzed. It goes without saying that the document structure is replaced with another format.
[0041]
After obtaining the document structure analysis result as described above, the main part selection unit 12 sequentially extracts the main part from the document structure analysis result according to the rules of each extraction rule dictionary stored in the extraction rule dictionary storage unit 2. Extraction (S4: corresponding to a main part selection function and a main part selection step).
[0042]
This extraction rule dictionary is based on a certain technical category (for example, about 30 technical categories close to the Japanese publication issue category, etc.) based on the IPC classification that is subdivided from class (superordinate category) to subgroup (subordinate category). (The technical classification according to the need of the field to which the present invention is applied), and the format of the rules in each extraction rule dictionary is composed of three elements: abstract items, fields, and key phrases as shown in FIG. ing. In the abstract item, the contents to be output as the main part are described in Japanese, and here, three items of the technical field, the gist, and the example (including the embodiment) are illustrated. The field indicates the title of each chapter in the document structure. For keyphrases, search for each chapter set in the field, and if the word or phrase defined in the keyphrase is included in the sentence explaining the contents of the title, extract it as the main part. Means.
[0043]
For example, in the document structure of an English patent specification that is the subject of an abstract, the technical fields defined by the rules of the dictionary in the sentence of paragraphs corresponding to titles such as “BREIF DESCRIPTION OF THE PRIOR ART”, “FIELD OF THE INVENTION”, “ABSTRACT”, etc. If a word or phrase such as “relates to”, “is directed tower”, or “the inventor concerns”, which is a key phrase corresponding to, is included, a candidate for the main part corresponding to the abstract item “technical field” It means that. In the paragraphs corresponding to the titles “SUMMARY OF THE INVENTION”, “SUMMARY”, and “ABSTRACT”, “is promoted”, “is disclosed”, “is provided”, “ If a word or phrase such as “disclosed is” is included, it means that the paragraph is a candidate for the main part corresponding to the abstract item “summary”. Since a plurality of conditions can be registered for each abstract item, for example, the rule corresponding to “summary” is divided into two extraction rules R2 and R3. In the key phrase corresponding to the fourth extraction rule R4, “FIG.?” Is described. This “?” Means matching with an arbitrary number. In other words, the key phrase “FIG.?” Means that it can be collated with any of “FIG. 1”, “FIG. 2”, and the like.
[0044]
Therefore, the main part selection unit 12 executes the main part selection process shown in FIG. 6 using the rules in the extraction rule dictionary as described above. Specifically, the document structure analysis result of the English patent specification shown in FIG. 4 is read from the buffer memory 6 or the database 5 (S11). Generally, since the IPC classification is described in the patent specification, the IPC can be recognized by the tag information also in the document structure analysis result shown in FIG. That is, in the example of FIG. 4, the range surrounded by “<ipc>” and “</ ipc>” is a portion where the IPC classification is described, and is described in detail as “C08F 002/00”. ing. By using this classification information, it is possible to determine the extraction rule dictionary corresponding to the technical classification based on the IPC classification that is subdivided and expanded from the class (upper classification) to the subgroup (lower classification).
[0045]
When the IPC classification is extracted from the document structure analysis result, the main part selection unit 12 reads the extraction rule dictionary corresponding to the corresponding IPC classification from the extraction rule dictionary storage unit 2 (S12), and then proceeds to the next step S13. As the extraction rule (ID = R1) to be referred to first in the extraction rule dictionary, 1 is set to the variable i. Then, based on the extraction rule of the variable i = 1, it is determined whether or not a paragraph to be collated with the field and key phrase of the rule exists in the English patent specification (document structure analysis result) to be abstracted (S14). . Here, if there is a paragraph to be collated, position information (specific information) of the paragraph to be collated is extracted (S5 in FIG. 2, S15 in FIG. 6) and stored in the buffer memory 6 or the database 5.
[0046]
After extracting the position information of this paragraph or when there is no paragraph to be collated in step S14, the process proceeds to the next step S16, and the extraction rule (ID = R2) to be processed next specified in the extraction rule dictionary is as follows. After incrementing the variable i by +1, it is determined whether or not all the extraction rule items corresponding to the relevant variable have been completed (S17). If there are still extraction rule items, the process returns to step S14. It is determined whether or not a paragraph to be matched with the extraction rule field and key phrase of the variable i exists in the patent specification (document structure analysis result) to be abstracted, and is repeatedly executed for all extraction rules.
[0047]
FIG. 7 is a diagram schematically showing the extraction part of the main part in order to explain the gist of the present invention. For example, in the second extraction rule (ID = R2) of the extraction rule dictionary shown in FIG. 5, “ABSTRACT” is described in the field, and “is disclosed” is described in the key phrase. Therefore, since this second extraction rule is collated with “is disclosed” (indicated by underline) included in the explanatory text 21 enclosed in the dotted dotted frame related to the title “Abstract”, this “ The description sentence 21 of “Abstract” or the paragraph position (specific information) in the sentence is extracted as the “summary”. In the first extraction rule (ID = R1) of the extraction rule dictionary shown in FIG. 5, “FIELD OF THE INVENTION” is described in the field, and “relates to” is described in the key phrase. Therefore, since this first extraction rule is collated with “relates to” (indicated by underlining) included in the explanatory text 22 of FIG. 7 surrounded by the dotted frame of “FIELD OF THE INVENTION”, this “ The description sentence 22 of “FIELD OF THE INVENTION” or the paragraph position (specific information) in the sentence is extracted as “technical field”. Further, in the third extraction rule (ID = R3) of the extraction rule dictionary shown in FIG. 5, “DATAILED DESCRIPTION OF THE INVENTION” is described in the field, and “this invention addresses” is described in the key phrase. Therefore, since this third extraction rule is collated with “this invitation addresses” (indicated by underlining) included in the explanatory text 23 enclosed in the dotted line frame of “DATAILED DESCRIPTION OF THE INVENTION”, this “DATAILED The description 23 of “DESCRIPTION OF THE INVENTION” or the paragraph position (specific information) in the sentence is extracted as the “summary”.
[0048]
FIG. 8 is a diagram illustrating an example of an expression format of an extraction result by the main part selection unit 12.
[0049]
FIG. 7 shows the main parts extracted while being associated with the document structure analysis result of FIG. 4 with the main part specifying information 21, 22, 23, but FIG. 8 more specifically shows the fields in the patent specification, It can be extracted by associating with a pointer such as a paragraph position in the sentence. Incidentally, FIG. 8 includes a combination of a field, a paragraph (paragraph position, specific information), an abstract item, and an extraction rule used for extracting a main part. Specifically, the information on the first line in FIG. 8 is a chapter “FIELD OF THE INVENTION”, which indicates that the first paragraph matches the extraction rule R1, and the extraction rule R1 is an abstract item. It is a rule corresponding to the “technical field”. The information on the third line is a section called “DATAILED DESCRIPTION OF THE INVENTION”, which indicates that the first paragraph matches the extraction rule R3. This extraction rule R3 is the “summary” of the abstract item. Indicates that the rule corresponds to.
[0050]
As described above, after the main part selection unit 12 extracts the main part as shown in FIG. 8 from the English document structure analysis result, for example, or a machine that performs processing in parallel with the main part selection processing by the main part selection unit 12 The translation unit 13 translates the document structure analysis result of the English patent document using the translation processing knowledge of the translation dictionary unit 8 (S6), and obtains the translation result of the native language, for example, Japanese ( S7). These steps S6 and S7 correspond to a translation processing function and a translation processing step.
[0051]
FIG. 9 is a diagram showing a translation result obtained by translating a document other than the XML tag into Japanese with respect to the English document structure analysis result shown in FIG. The reason for the translation process excluding the XML tag is that the original English document structure analysis result needs to be retained for the translation result, so that the translation process is not performed for the XML tag. For English-Japanese machine translation, some package software is already on the market, and a translation processing technique realized by these package software may be used. Thereby, it is possible to easily obtain the translation result as shown in FIG.
[0052]
Subsequently, the abstract candidate extraction unit 14 compares the main part extraction result (see FIG. 8) extracted by the main part selection unit 12 with the machine translation result (see FIG. 9) acquired by the machine translation unit 13. For example, a Japanese abstract candidate, which is a native language, is extracted (S8: equivalent to abstract candidate extraction function, abstract candidate extraction step).
[0053]
Specifically, the abstract candidate extraction unit 14 extracts abstract candidates related to the first paragraph of “FIELD OF THE INVENTION” on the first line in FIG. 8. In the example shown in FIG. 9, in <field header = “FIELD OF THE INVENTION”>, a paragraph beginning with “[0001]” corresponds to an abstract candidate. Further, abstract candidates relating to the first paragraph of “ABSTRACT” on the second line in FIG. 8 are extracted. From the example shown in FIG. 9, in <field header = “ABSTRACT”>, a paragraph beginning with “process” corresponds to an abstract candidate. Similarly, abstract candidates are sequentially extracted from the machine translation results in accordance with the information on each line of the main part extraction results shown in FIG. 8 and stored in a predetermined area of the database 5. FIG. 10 is a diagram showing extraction results of the abstract candidates.
[0054]
By the way, in the embodiment described above, an example in which machine translation processing is performed on the premise that the number of lines and the number of paragraphs does not change has been described. However, in the machine translation processing, a single sentence is divided into a plurality of sentences and translated. May be performed. In such a case, as shown in FIG. 8, there is a possibility that the correspondence is shifted in correspondence between paragraphs and sentences. However, since the document structure analysis unit 11 allocates and processes XML tags having unique numbers for all sentences and paragraphs on the English patent specification side, based on the document structure analysis result, If machine translation is performed with the XML tag information held, there is no problem.
[0055]
Therefore, in this abstract creation support system, at the stage of extracting abstract candidates as described above, the abstract related information storage database 5 stores the entire English patent specification (see FIG. 3) and the English document structure analysis result (FIG. 4). ), Japanese machine translation results (see FIG. 9), Japanese abstract candidate extraction results (see FIG. 10), and the like are stored, the abstract editing display control unit 15 performs the selection operation from the input unit 1. Then, required information is selectively extracted from the abstract related information storage database 5 and displayed on the output unit 7 in accordance with the required format (S9).
[0056]
FIG. 11 is a diagram showing a screen display example of the main part display in Japanese / English (upper part) or the full text display in Japanese / English (lower part) displayed on the output unit 7 under the required operation by the abstract creator. It is.
[0057]
First, when the abstract creator takes out the initial screen, the main screen button 31, the full text button 32, the open button 33, the copy button 34, the save button 35, the end button 36, and other necessary buttons (see FIG. In addition, in the upper left display area 37 of FIG. 11, Japanese translation essential parts (abstract candidates) are sequentially displayed. On the other hand, in conjunction with the display of the translation essential parts, the English original sentence essential parts are sequentially displayed in the display area 38 on the upper right side of FIG. In addition, since each character of the translation main part and the original text main part cannot be described due to the space of the drawing, a part of the translation main part and the original text main part are shown in FIG. In each of the display areas 37 and 38, one paragraph can be selected by a pointing device such as a mouse constituting the input unit 1.
[0058]
That is, the display area 37 and the display area 38 are linked, and when a paragraph with the display area 37 is selected, the main part of the English text corresponding to the original patent specification is linked to the display area 38 and selected. Is displayed. 39 is an abstract creation area. When the abstract creator determines that the abstract is the best abstract while confirming the translation essential parts and original text essential parts of the display areas 37 and 38, the relevant translation essential parts are copied and necessary. In response to this, this is an area that is subjected to correction processing such as character deletion and insertion to make a Japanese abstract. That is, the abstract creator creates a Japanese abstract by appropriately editing the contents of the abstract creation area 39. Also, for reference by the abstract creator, bibliographic information such as the inventor, application date, and IPC is displayed in the display area 40, and the translation result of the patented invention name is displayed in the display area 41. Is done.
[0059]
Further, when the full text button 32 is operated, the translated full text and the original full text are displayed in conjunction with each other as shown in the lower part of FIG. Thereafter, when the main part button 31 is operated, the Japanese / English main part shown in the upper part of the figure is displayed.
[0060]
FIG. 13 is a diagram illustrating a state transition diagram when actually used by an abstract creator or user (hereinafter collectively referred to as an abstract creator) and a process associated with the state transition of the abstract editing display control unit 15.
[0061]
In this figure, state 51 is the initial state when the system is started up, state 52 is a state in which the main part of the translation result is displayed in the display area 37 by the abstract creator opening the abstract processing screen, and state 53 is the system The end state means the end of, and the state 54 means a state in which the translation result of the full text of the specification is displayed in the display area 37 by the operation of full text display by the abstract creator.
[0062]
Further, the state transition will be described in detail. When the abstract creator operates the button 33 opened in the state 51, the English patent specification is obtained by inquiring and obtaining the file name of the English patent specification. The translation result of the main part extracted from the above, that is, the translation main part is taken out from the required area of the database 5 and displayed in the display area 37. After the display and the display area 39 are initialized, the state transitions to the state 52.
[0063]
In this state 52, when a paragraph displayed in the display area 37 is selected and the copy button 34 is operated, the paragraph is inserted into the display area 39. The state 52 is a state of the main part (translation), and the abstract items (see FIGS. 11 and 12) “technical field”, “summary”, and “example” are set for each paragraph. Then, it is sequentially inserted into the display area 39 by copying according to the abstract item (S52).
[0064]
Further, in the state 54, the full-text translated (full-text translation) is displayed in the display area 37 by operating the full-text button 37, and the original full-text is displayed in the display area 38 in conjunction with this (S53). In this state 54, since the abstract item has not been determined, after setting the abstract item through a process such as requesting input from the abstract creator, the paragraph displayed in the display area 37 is copied as necessary. Copying to the display area 39 by operation (S54). When the save button 35 is operated in the state 52 or the state 54, the contents of the display area 39 are saved in the database 5 based on the input of the file name of the Japanese abstract of the save destination by the abstract creator. (S55, S56). If the open button 33 is operated in the state 54, the same processing as step S51 is executed (S57), and if the main button 31 is operated, the same processing as step S52 is executed (S58). When the end button 36 is operated in each state, the abstract creation and the required information extraction and display processing are completed.
[0065]
Therefore, according to the embodiment as described above, after analyzing the item (document structure) of the foreign language patent specification and obtaining the tagged document structure analysis result based on a certain format, the document structure analysis result is obtained. Based on the above, the main part selection unit 12 extracts the main part according to the rules of the extraction rule dictionary determined in advance, and the machine translation unit 13 converts the contents of the document structure analysis result into, for example, the native language based on normal translation processing. translate. After that, after the abstract candidate extraction unit 14 extracts the abstract candidates from the translation results of the whole Japanese sentence translated for each main part, which is the main part extraction result, the translation main part and the original main part are associated with each other. Display, copy the translation main part, refer to the main part of the original text, and if necessary, refer to the entire document structure analysis result of the English patent specification and the translated translation result, and extract the Japanese abstract by deleting and inserting characters. Therefore, Japanese abstracts can be efficiently created, and manual Japanese abstract work can be appropriately supported.
[0066]
The main part display is not limited to the form shown in FIGS. For example, as shown in FIG. 14, the main part button 31 is divided into a translation main part button 31a and an original sentence main part button 31b, and the full sentence button 32 is also divided into a translated full sentence button 32a and an original full sentence button 32b. It may be configured to select and display in four display modes by performing a selection operation. In short, it is possible to freely display the English text of the main part selected from the English patent specification, the translation result thereof, the original English patent specification, the translated full text specification, etc. freely and sequentially.
[0067]
Next, FIG. 15 is a diagram for explaining another embodiment of the abstract creation support system according to the present invention.
[0068]
In this embodiment, priorities are set in advance in the rules of each extraction rule dictionary stored in the extraction rule dictionary storage unit 2, and the display amount or display number of abstract candidates is controlled based on the importance of the extracted paragraphs. This is an example.
[0069]
In the abstract creation support system according to the present invention, if the importance level is set in advance in the extraction rule dictionary shown in FIG. 15, if the main part is selected by the main part selection unit 12, the main part extraction result shown in FIG. The main part extraction result shown in FIG. 16 with the importance added is obtained. That is, the importance information of the extraction rule used at the time of extraction is added to the extraction result of the main part for each paragraph to be extracted. As a result, although not shown in the drawing, importance level information is added to the translation results of abstract candidates shown in FIG. It is assumed that “1” has the highest importance and that the importance decreases as the numerical value increases.
[0070]
As a result, the abstract editing display control unit 15 executes abstract candidate extraction processing in consideration of the importance as shown in FIG. That is, the abstract edit display control unit 15 first sets 1 as the first extraction rule (ID = R1) to the variable i (S61), and then the importance corresponding to this extraction rule (ID = R1). Is greater than a predetermined threshold, that is, whether the importance is higher than the threshold (S62). If it is determined that the importance is high, the Japanese translation corresponding to the paragraph of the field The main part is displayed in the display area 37 (S63). After displaying this translation main part or when it is determined that the importance is lower than the threshold value in step S62, the process proceeds to step S64, and after adding +1 to the variable i, the first extraction rule (ID of this variable i (ID = Ri) is determined whether it still remains (S65). If it still remains, the process returns to step S62 to execute the same processing. When the importance level is checked for all the extracted paragraph numbers and the main part translation is displayed, the process related to the importance level ends.
[0071]
Therefore, according to the embodiment as described above, it is possible to create a Japanese abstract by displaying in the display area 37 only the translation essential parts having a higher importance than the threshold value determined in advance in the abstract creation stage. Japanese abstracts can be created efficiently using quantities.
[0072]
The importance threshold can be changed by the abstract creator at any time, and the Japanese abstract required by the abstract creator can be easily created.
[0073]
FIG. 18 is a diagram for explaining still another embodiment of the abstract creation support system according to the present invention.
[0074]
This embodiment is an example in which “description of representative diagram” is extracted as an abstract candidate from the specification described in the document structure analysis result shown in FIG.
[0075]
The abstract creation support system according to the present invention expands the expression format of the extraction rule dictionary by adding not only phrase matching but also special symbols to the key phrases in the rules defined in the extraction rule dictionary. The explanation of the figure is extracted as the main part.
[0076]
Among the symbols to be introduced into this key phrase, “$ RepFig $” is a symbol meaning a representative figure, and “$ ^ Line $” is a key phrase search from the beginning of the specified field to the Line paragraph. It is a symbol that means a range. "$ ^* “Line $” means that the Line paragraph below the paragraph to be collated is taken out. "$ ^* For example, “$” means that all the paragraphs are extracted as the main part for the “SUMMARY” field.
[0077]
In this system, an extraction rule as shown in FIG. 18 is described in the extraction rule dictionary unit, so that the main part extraction unit 12 relates to the first extraction rule (ID = R1). If it is 1, it means searching for a key phrase “FIG. 1 shows” in the document structure analysis result corresponding to “FIELD OF THE INVENTION” defined in the field. In addition, regarding the second extraction rule (ID = R2), the search range is up to the third paragraph in the document structure analysis result corresponding to “DATA ILED DESCRIPTION OF THE INVENTION” defined in the field, and “FIG.? Shows”. This means searching for the presence of a key phrase ("?" Is an arbitrary number). Furthermore, regarding the third extraction rule (ID = R3), it means that two paragraphs in the document structure analysis result corresponding to “CLAIMS” defined in the field (because no other key phrase is specified) are extracted. To do.
[0078]
Incidentally, “pap-v16-2002-01-01.” In “-// USPTO // DTD PAP V1.6 2002-01-01 // EN” which is the definition of the XML format of the US published patent specification. According to Dtd, a number surrounded by <representative figure> and </ representative figure> represents a representative diagram. When a patent specification described in such a format is input, what the representative diagram is can be easily extracted from the patent specification.
[0079]
Therefore, in step S14 shown in FIG. 6 of the main part selection unit 12, after replacing with a specific character string of the symbol “$ RepFig $” in the extraction rule dictionary using the above definition information, The search may be performed according to the key phrase. That is, for example, if the representative figure of the patent specification to be abstracted is “FIG. 1”, “$ RepFig $” can be replaced with “FIG. 1”, so the key phrase of the extraction rule R1 is “FIG. 1 shows ". Therefore, the key phrase “FIG. 1 shows” is searched from the document structure analysis result corresponding to “DATA ILED DESCRIPTION OF THE INVENTION” defined in the field.
[0080]
In addition, if the key phrase described in the extraction rule includes the special symbol “$ RepFig $” in step S14 (FIG. 6) of the key phrase search in the main part selection unit 12, the key phrase search is performed. Limit the range to the Line paragraph, or use the special symbol “$ ^* If "Line $" is included, the line paragraph below the paragraph to be matched with the key phrase is extracted as the main part, or the special symbol "$" is included in the key phrase. ^* If “$” is included, processing such as selecting all the paragraphs of the field to be processed as the main part is performed.
[0081]
Therefore, according to such an embodiment, it is possible to easily extract the explanation of the representative diagram from the patent specification as an abstract candidate.
[0082]
If the special symbols as described above are newly defined and the main part selection unit 12 is configured accordingly, it is possible to easily extract the contents as the main parts other than the description of the representative diagram.
[0083]
FIG. 19 is a diagram for explaining still another embodiment of the abstract creation support system according to the present invention.
[0084]
This embodiment has a configuration in which an unnecessary word dictionary is newly provided in the extraction rule dictionary storage unit 2, and even a paragraph that matches the conditions of the extraction rule dictionary is not extracted only in a paragraph that matches an unnecessary word. This is an example.
[0085]
FIG. 19 is a diagram showing an example of an unnecessary word dictionary. For example, unnecessary words of the first and second rules F1 and F2 are phrases used in explaining the second embodiment, May not be appropriate for selection. This is because the unnecessary words in the third and fourth rules F3 and F4 are phrases that are used when they are dependent claims, and these are often not to be selected as essential parts.
[0086]
FIG. 20 is a flowchart showing a series of processing examples of the main part selection unit 12 when an unnecessary word dictionary is added to the extraction rule dictionary storage unit 2 in addition to the extraction rule dictionary.
[0087]
In the main part selection unit 12, the difference from the processing in FIG. 6 is that in step S 74, it is determined whether or not there is a paragraph that matches the extraction rule i. It is determined whether or not an unnecessary word exists in the collated paragraph based on (S75), and if an unnecessary word is included, the process is not extracted as a main part. Accordingly, the other steps S71 to S74 correspond to steps S11 to S14 in FIG. 6, and steps S76 to S78 correspond to steps S15 to S17 in FIG. 6, and detailed processing will be left to the description of FIG.
[0088]
Therefore, according to the embodiment as described above, when an unnecessary word defined in the rules of the unnecessary word dictionary is present in a paragraph corresponding to the main part of the machine sentence translation result, it is determined that it is not appropriate as the main part. Is not extracted as the main part, so a more accurate Japanese abstract can be created.
[0089]
The unnecessary word dictionary shown in FIG. 19 assumes that an unnecessary word dictionary is prepared for each IPC classification. For example, an unnecessary word is set for each abstract item or each extraction rule in the extraction rule dictionary. Such a configuration may be adopted.
[0090]
Further, in each of the embodiments described above, paragraphs are used as processing units in the selection of the main part and the generation of abstract candidates. However, these are processing units in other language units (sentences, sections, phrases, words, etc.). It can also be configured.
[0091]
Furthermore, in the present embodiment, an extraction rule dictionary is selectively used according to the IPC classification described in, for example, an English patent specification, which is an abstract creation target. If the technical term dictionary is selectively used, an improvement in translation accuracy is expected, and it is also possible to modify the dictionary to be used for machine translation according to the IPC classification.
[0092]
Furthermore, in recent years, the function of a translation memory has been realized in a machine translation system (for example, The Translation Professional V8.0). With this translation memory, if a Japanese translation of a certain English sentence with high accuracy was created in the past, it would be possible to re-use the same Japanese translation for an English sentence similar to the English sentence. Based on the fact that it is high, remember the past English-Japanese translation combination, search for past similar sentences when translating a new English sentence, and if there are similar sentences, the Japanese corresponding to the similar sentences This is a function for presenting a translation to the user.
[0093]
Therefore, also in the abstract creation support system according to the present invention, for example, the translation result related to the main part extracted by the main part selection unit 12 is presented to the abstract creator, and finally the Japanese abstract is created while referring to the translation result. The structure to do may be sufficient. In this case, if the English main part extracted by the main part selection unit 12 and the Japanese abstract of the correction result are stored, the translation memory function in the machine translation unit 13 can be used. That is, the system of the present invention is a machine further provided with a storage means for storing a pair of an English main part extracted by the main part selecting part 12 and a Japanese abstract after completion of editing by the abstract creator, and a translation memory function. If the translation unit 13 is used, a Japanese abstract can be created while effectively utilizing the translation memory function. At this time, if it is configured to store the English main part-Japanese abstract corresponding to the IPC classification, the past examples of the English main part-Japanese abstract searched according to the IPC classification of the English patent specification subject to the abstract generation can be efficiently used. Selective use is possible.
[0094]
In the above embodiment, an example in which an abstract creator whose native language is Japanese is a Japanese abstract has been described. For example, English to Chinese, English to German, Japanese to English, etc. The same applies to the above. Therefore, in this case, if an extraction rule in the extraction rule dictionary or a language to which machine translation is applied is incorporated in accordance with another language pair, an abstract creation support system corresponding to the other language pair can be realized. Needless to say.
[0095]
Further, in the above embodiment, the Japanese abstract candidate and the main part of the original text are displayed in association with each other, but it is also possible to display only the Japanese abstract candidate extracted by the abstract candidate extracting unit 14. .
[0096]
Next, FIG. 21 is a functional configuration diagram illustrating an embodiment of a search system with a main part creation function according to the present invention.
[0097]
This search system includes a document search unit 61, an extraction rule dictionary storage unit 62 having the same function as the extraction rule dictionary storage unit 2 in FIG. 1, and a document structure analysis unit having the same function as the document structure analysis unit 11 in FIG. 63, a main part selection unit 64 having the same function as the main part selection unit 12 of FIG. 1, a machine translation unit 65 having the same function as the machine translation unit 13 of FIG. 1, a search candidate extraction 66, and a search result display control unit 67, and various other components such as a patent document storage database 68, a translation dictionary (not shown), a recording medium for recording a series of search processing programs, and the like are provided. In the figure, the components 2, 63 to 65 having the same functions as those in FIG. 1 have already been described in detail in FIGS.
[0098]
First, a large number of foreign language patent specifications, for example, English patent specifications, are digitized and stored in the patent document storage database 68.
[0099]
In this state, when a search instruction is input from a search input unit (not shown) under a search condition such as a technical field or an IPC classification, the document search unit 61 performs a series of processes from the recording medium according to the search processing program. Execute. That is, when the search condition is input, the document search unit 61 searches the database 68 for a plurality of English patent specifications that match the search condition (71). As the document search unit 61, various known document search techniques are used.
[0100]
When a plurality of English patent specifications are searched by the document search unit 61, after storing them in a temporary buffer memory or the like, for example, the patent specifications as search results are sent to the document structure analysis unit 63 in the searched order, and the patent specifications Analyzes the document structure. The document structure analysis of the patent specification has already been described with reference to FIGS. As a result, a document structure analysis result as shown in FIG. 4 can be extracted (72).
[0101]
The document structure analysis result extracted in this way is sent to the main part selection unit 64 and the machine translation unit 65. The main part selection unit 64 extracts a main part from the document structure analysis result according to the rules defined in the extraction rule dictionary (see FIG. 5) stored in the extraction rule dictionary storage unit 62 from the document structure analysis result according to the processing flow shown in FIG. (73). The main part extraction result is as shown in FIG. 7 or FIG. On the other hand, the machine translation unit 65 translates the document structure analysis result into the native language using the rules and knowledge prescribed in the translation dictionary unit 8, and acquires the machine translation result as shown in FIG. 9 (74). . The main part selection unit 64 and the machine translation unit 65 have been described in detail in the description related to FIGS. 1 and 2, and will be omitted here.
[0102]
Subsequently, the search candidate extraction unit 66 performs a search translated from the Japanese translation results of the full text of the foreign language patent specification translated by the machine translation unit 65 for each major part as the extraction result of the major part selection unit 64. The main part is taken out and stored, for example, in a search result storage database (not shown). The search result storage database includes a full-text English patent specification that is a searched original, a full-text translated native-language patent specification that is machine-translated by the machine translation unit 65, and an original text that is extracted by the main part selection unit 64. And the translation main part generated by the search main part generating part 66 are stored.
[0103]
Therefore, the search result display control unit 67 displays the result on the output unit corresponding to 7 in FIG. 1 according to the processing procedure shown in FIG. FIG. 23 is a diagram showing a screen display example. Note that a full text display button 81, a main part display button 82, a next selection button 83, and a previous selection button 84 are formed on the initial screen.
[0104]
When the main part button 82 is operated, the search result display control unit 67 displays the display areas 85 and 86 with overlapping in the screen depth direction when there are a plurality of searched English patent specifications. In this state, when 1 is set as the variable i, the main part of the original text (the main part extraction result) extracted from one English patent specification i searched from the search result storage database is extracted. The parts are classified and displayed in the display area 85. Then, in conjunction with the original text main part, the main translation part corresponding to each original text is displayed in the display area 86 (S82).
[0105]
Thereafter, it is determined which button has been operated (S83), and when there is no button operation, it is repeatedly determined. If it is determined in step S83 that the button is operated, it is determined whether or not the full text display button 81 has been operated (S84). If it is determined that the full text button 81 has been operated, the corresponding patent specification is determined. The full text of i is displayed in the display area 85, and the full text translation that is the machine translation result corresponding to the full text of the original is displayed in the display area 86 (S85). Next, it is determined which button has been operated (S86). If it is determined that the main display button 82 has been operated (S87), the process returns to step S82 and the English patent specification is similarly applied. The original text main part and the translation main part extracted from the book i are displayed.
[0106]
If it is determined in step S84 that the button operation has been performed but the full-text display button 81 is not operated, it is determined whether the previous selection button 84 or the next selection button 83 is operated (S89, S90). . Here, when it is determined that the previous selection button 84 has been operated, i-1 is set to the variable i within the range of 1 (the lower limit number) (S91), and the process proceeds to step S82, and the previous English patent specification The main part of the original text and the main part of translation extracted from the book i are displayed in the display areas 85 and 86, respectively. On the other hand, if it is determined that the next selection button 83 has been operated, i + 1 is set to the variable i within the range of N (upper limit number) (S92), and the process proceeds to step S82, where it is extracted from the next English patent specification i. The original text main part and the translation main part are displayed in the display areas 85 and 86, respectively.
[0107]
In step S87, if the button operation is not the operation of the main part display button 82 but the previous selection button 84 is operated (S93), i-1 is set to the variable i (S94), and the process proceeds to step S85. Then, the original text and the translated text are displayed in the display areas 85 and 86, respectively. When the next selection button 83 is operated (S95), i + 1 is set to the variable i (S96), and the process proceeds to step S85 to display the original text and the translated text in the display areas 85 and 86, respectively.
[0108]
Therefore, according to the embodiment as described above, at least one foreign language patent specification is searched from a number of foreign language patent specifications based on a search condition, and the searched foreign language patent is searched. For the description, the full-language translated native language patent specification, the original text part, the translation main part are taken out, and the foreign language patent specification is selectively displayed on the display screen-the full-text translated native language patent specification, the original text part- Since the translation is displayed in a paired relationship like the main parts of the translation, the searcher can easily grasp the required technical contents, and thus the search efficiency can be greatly improved.
[0109]
In FIG. 23, the main part is displayed for each item. However, the main part may be displayed in the order of appearance of the paragraphs extracted by the main part selection unit 64 instead of for each item.
[0110]
Although not shown in FIG. 23, in the case of the full text display, since the main part is already extracted by the main part selection unit 64, the main part corresponding part in the whole sentence is highlighted with a different color, for example. For example, it is possible to recognize at a glance which part is important in the whole sentence, and to further improve the search efficiency.
[0111]
Note that the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the scope of the invention.
[0112]
In addition, the embodiments can be implemented in combination as much as possible, and in that case, the effect of the combination can be obtained. Further, each of the above embodiments includes various higher-level and lower-level inventions, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent elements. For example, when an invention is extracted because some constituent elements can be omitted from all the constituent elements described in the means for solving the problem, the omitted part is used when the extracted invention is implemented. Is appropriately supplemented by well-known conventional techniques.
[0113]
【The invention's effect】
As described above, according to the present invention, it is possible to provide an abstract creation support system, a program, and an abstract creation support method that can efficiently and appropriately create an abstract of a second language from patent documents in the first language.
[0114]
In addition, the present invention quickly creates a second language patent specification, a source text main part and a translation main part corresponding to the searched first language patent specification, and displays them selectively. In addition, it is possible to provide a patent document search system, a program, and a patent document search method that can grasp the technical contents and thus greatly improve the search efficiency.
[Brief description of the drawings]
FIG. 1 is a configuration diagram showing an embodiment of an abstract creation support system according to the present invention.
FIG. 2 is a diagram for explaining a flow of a series of processes of the abstract creation support system shown in FIG. 1;
FIG. 3 is a diagram showing an example of an English patent specification, for example, which is an abstract creation target.
4 is a diagram showing a document structure analysis result of an English patent specification analyzed by a document structure analysis unit shown in FIG. 1. FIG.
FIG. 5 is a diagram for explaining rules in an extraction rule dictionary stored in an extraction rule dictionary storage unit;
6 is a flowchart for explaining a specific processing flow of a main part selection unit shown in FIG. 1;
FIG. 7 is an image diagram of a main text extraction result extracted by a main text selection unit.
FIG. 8 is a diagram illustrating an example of an expression of a main part extraction result extracted by a main part selection unit.
FIG. 9 is a diagram showing a machine translation result of, for example, an English patent specification that is an abstract creation target shown in FIG. 3;
10 is a diagram for explaining an example of Japanese abstract candidates extracted by the abstract candidate generation unit shown in FIG. 1; FIG.
FIG. 11 is a display example of abstract candidates, in particular a display example of the extracted translation main part and original text main part, an example of editing a Japanese abstract from this translation main part, and a display example of the full translation and original full text Illustration to explain.
FIG. 12 is a diagram showing an example in which a translation main part and an original text main part are displayed in association with each other.
FIG. 13 is a diagram for explaining state transitions by the abstract edit display control unit shown in FIG. 1 and processing associated with the state transitions.
FIG. 14 is a diagram showing a display example of another abstract candidate.
FIG. 15 is a diagram showing an example of an extraction rule dictionary for explaining another embodiment of the abstract creation support system according to the present invention.
16 is a diagram showing an example of a main part extraction result extracted by the main part selection unit using the extraction rule dictionary shown in FIG.
FIG. 17 is a flowchart illustrating a flow of a series of processes in a main part selection unit.
FIG. 18 is a diagram showing an example of an extraction rule for explaining still another embodiment of the abstract creation support system according to the present invention.
FIG. 19 is a diagram showing an example of an unnecessary word dictionary for explaining still another embodiment of the abstract creation support system according to the present invention.
FIG. 20 is a flowchart illustrating an example of processing for extracting a main part using an extraction rule dictionary and an unnecessary word dictionary in a main part selection unit;
FIG. 21 is a functional configuration diagram illustrating an embodiment of a search system according to the present invention.
22 is a flowchart for explaining a series of processing of a search result display control unit in FIG.
FIG. 23 is a diagram showing a display example of search results.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Original text description input part, 2 ... Extraction rule dictionary storage part, 3 ... Abstract creation process control part, 4 ... Recording medium, 5 ... Abstract related information storage database, 7 ... Output part, 8 ... Translation dictionary part, 11 Document structure analysis unit, 12 Main part selection unit, 13 Machine translation unit, 14 Abstract candidate extraction unit, 15 Abstract edit display control unit, 61 Document search unit, 62 Extraction rule dictionary storage unit, 64 Main part selection part, 65 ... Machine translation part, 66 ... Search candidate extraction part, 67 ... Search result display control part.

Claims (12)

An extraction rule dictionary that defines key phrases such as words and phrases of the first language included in the captions of the first language chapters and the description sentences described in the chapters of the respective titles, which are required for each certain classification;
A document structure analyzing means for analyzing the chapter structure of the sentence described in the electronic first language patent application document;
The required extraction rule dictionary is selected according to the classification described in the document structure analysis result analyzed by the document structure analysis means, and the document structure analysis result based on the title and key phrase specified in the extraction rule dictionary Main part selection means for selecting the main part from
Machine translation means for translating the document structure analysis result analyzed by the document structure analysis means into a second language;
Abstract candidate extraction means for extracting second language abstract candidates from the translation result of the second language translated by the machine translation means for each main part selected by the main part selection means;
Only the abstract candidates of the second language extracted by the abstract candidate extracting means are displayed, or the abstract candidates of the second language and the main parts of the first language are displayed in association with each other, and the abstract of the second language is displayed. An abstract creation support system comprising an abstract edit display control means for creating an abstract in a second language including candidate correction processing. The essential part selecting means specifies an abstract item of a second language to be classified into the extraction rule of the extraction rule dictionary when the abstract is created, specifies the essential part to be collated with the extraction rule, 2. The information output device according to claim 1, further comprising means for adding and outputting an abstract item of the second language corresponding to the main part to the main part specifying information when outputting the information to be specified. Abstract creation support system. The abstract edit display control means includes a second language abstract candidate extracted by the abstract candidate extraction means and a main part of the first language, a full sentence structure analysis result of the first language, and a translation by the machine translation means. And further providing means for selectively displaying the processed second language translation results in correspondence with each other and generating a second language abstract including a correction process for the second language abstract candidate. The abstract creation support system according to claim 1, wherein Importance information is newly defined in the extraction rule of the extraction rule dictionary,
The main part selection means specifies a main part to be collated with the extraction rule, and outputs the information specifying the main part with the importance information added to the main part specifying information,
The abstract candidate extraction unit compares the importance information output from the main part selection unit with a threshold value indicating a predetermined importance level, and the main part to which the importance information higher than the threshold value is added. 3. The abstract creation support system according to claim 1, further comprising means for extracting a second language abstract candidate from the translation result of the second language translated by the machine translation means. . A key phrase having a symbol including a symbol related to a representative diagram is added to the extraction rule dictionary,
The main part selection means selects the required extraction rule dictionary according to the classification described in the document structure analysis result analyzed by the document structure analysis means, and uses the title and key phrase defined in the extraction rule dictionary. The abstract creation support system according to claim 1, wherein a location for explaining a representative diagram is selected based on the document structure analysis result. In addition to the extraction rule dictionary, an unnecessary word dictionary is provided,
The main part selection means selects the required extraction rule dictionary according to the classification described in the document structure analysis result analyzed by the document structure analysis means, and uses the title and key phrase defined in the extraction rule dictionary. Based on the result of the document structure analysis, a main part is selected, and if there is an unnecessary word to be described in the unnecessary word dictionary in the selected main part, means for excluding the selected main part The abstract creation support system according to any one of claims 1 to 5, further provided. Extraction rules having key phrases such as words and phrases of the first language included in the captions described in the chapters of the first language and the descriptions of the chapters of the respective titles are stored at least for each certain classification, A computer that creates an abstract in the second language from the electronic patent application documents in the first language,
A document structure analysis function for analyzing the chapter structure of a sentence described in an electronic first language patent application document;
Select the required extraction rule dictionary according to the classification described in the analyzed document structure analysis result, and select the main part from the document structure analysis result based on the title and key phrase specified in the extraction rule dictionary The main part selection function to
A machine translation function for translating the document structure analysis result analyzed by the document structure analysis function into a second language;
Abstract candidate extraction function for extracting second language abstract candidates from the second language translation result translated by the machine translation function for each main part selected by the main part selection function;
Only the abstract candidates of the second language extracted by this function are displayed, or the abstract candidates of the second language and the main parts of the first language are displayed in association with each other, and the abstract candidates of the second language are corrected. A program characterized by realizing a function of creating an abstract of a second language including processing. The title of chapters in the first language required for each fixed category and the extraction rules for key phrases such as words and phrases in the first language included in the explanations described in the chapters of each title are stored, and electronic An abstract creation support method for creating an abstract in a second language from a patent application document in a first language,
A document structure analysis step for analyzing a chapter structure of a sentence described in an electronic first language patent application document;
Select the required extraction rule dictionary according to the classification described in the analyzed document structure analysis result, and select the main part from the document structure analysis result based on the title and key phrase specified in the extraction rule dictionary The main part selection steps to be
A machine translation processing step for translating the document structure analysis result analyzed in the document structure analysis step into a second language;
An abstract candidate extraction step for extracting an abstract candidate in a second language from the translated second language translation result for each essential part selected in the essential part selection step;
Only the extracted abstract candidate of the second language is displayed, or the abstract candidate of the second language and the main part of the first language are displayed in association with each other, and the correction process of the abstract candidate of the second language is performed. And an abstract creation support method comprising the step of creating an abstract in a second language. Storage means for storing a plurality of digitized patent application documents in a first language;
Document search means for searching for at least one patent application document in the first language from the storage means under search conditions;
An extraction rule dictionary that defines key phrases such as words and phrases of the first language included in the captions of the first language chapters and the description sentences described in the chapters of the respective titles, which are required for each certain classification;
Document structure analyzing means for analyzing a chapter structure of a sentence described in the electronic first language patent application document searched by the document searching means;
The required extraction rule dictionary is selected according to the classification described in the document structure analysis result analyzed by the document structure analysis means, and the document structure analysis result based on the title and key phrase specified in the extraction rule dictionary Main part selection means for selecting the main part to be searched from,
Machine translation means for translating the document structure analysis result analyzed by the document structure analysis means into a second language;
Search candidate extraction means for extracting second language search candidates from the translation result of the second language translated by the machine translation means for each search target principal part selected by the principal part selection means;
After the search candidate of the second language is extracted by the search candidate extraction means, the display of the search main part of the first language or the correspondence display between the search main part of the first language and the search candidate of the second language or A patent document search system, comprising: search result display control means for displaying a correspondence between the document structure analysis result of the first language and the translation result of the second language. Importance information is newly defined in the extraction rule of the extraction rule dictionary,
The main part selection means specifies a main part to be collated with the extraction rule, and outputs the information specifying the main part with the importance information added to the main part specifying information,
The search candidate extraction unit compares the importance level information output from the main part selection unit with a threshold value indicating a predetermined importance level, and controls a search amount based on the threshold value. 9. The patent document search system according to 9. Multiple electronic patent application documents in the first language and the titles of the chapters in the first language required for each certain category, and the words and phrases in the first language included in the explanatory texts described in the chapters of each title An extraction rule dictionary that defines key phrases such as: a computer for extracting search candidates in at least the second language from the patent application documents in the first language;
A document search function for searching for at least one patent application document in the first language from the stored information under a search condition;
A document structure analysis function for analyzing a chapter structure of a sentence described in the electronic first language patent application document searched by this function;
The required extraction rule dictionary is selected according to the classification described in the document structure analysis result analyzed by this function, and the search object is searched from the document structure analysis result based on the title and key phrase specified in the extraction rule dictionary. The main part selection function to select the main part to be
A machine translation function for translating the document structure analysis result analyzed by the document structure analysis function into a second language;
A search candidate extraction function for extracting a search candidate of the second language from the translation result of the second language translated by the machine translation function for each search target main part selected by the main part selection function;
A search result display control function that displays search candidates in the second language extracted by the search candidate extraction function or displays the search candidates in the second language in association with the search main parts in the first language is realized. A program characterized by letting Multiple electronic patent application documents in the first language and the titles of the chapters in the first language required for each certain category, and the words and phrases in the first language included in the explanatory texts described in the chapters of each title An extraction rule dictionary that defines key phrases such as: a search method for extracting at least a search candidate in the second language from the patent application document in the first language,
A document search step of searching for at least one first language patent application document from among the stored information under a search condition;
A document structure analysis step of analyzing a chapter structure of a sentence described in the electronic first language patent application document searched by this step;
The required extraction rule dictionary is selected according to the classification described in the document structure analysis result analyzed in this step, and the search object is searched from the document structure analysis result based on the title and key phrase specified in the extraction rule dictionary. A main part selection step for selecting the main part to be
A machine translation processing step for translating the document structure analysis result analyzed in the document structure analysis step into a second language;
A search candidate extraction step of extracting a search candidate of the second language from the translation result of the second language translated by the machine translation function for each search target main part selected in the main part selection step;
The second language search candidates extracted in the search candidate extraction step are displayed, or the second language search candidates and the first language search main part are displayed in association with each other. Patent literature search method.

Images