JP3682535B2 - Document difference detection apparatus and program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a document difference detection apparatus and a program for detecting a difference between documents (or sentences) so that the difference between documents can be easily understood.
[0002]
[Prior art]
Conventionally, there is a technique in which a diff command is used to detect a difference between a plurality of input document data, and among the differences between a plurality of document data, one common part is output, and mismatched parts are output side by side. there were.
[0003]
Here, diff is a file comparison tool diff of UNIX (registered trademark). This diff command outputs the difference between two given files in units of lines while maintaining the order information.
[0004]
The diff command has a convenient option called the -D option. Using this option with the diff command will output not only the difference but also the common part. That is, file merging is realized. In addition, in order to display the difference portion in an easy-to-see manner, a display representing the beginning of the difference portion, the end of the difference portion, and the boundary between the two data constituting the difference is performed. Such diff when merging files is called Mdiff (M is M of merge) (for example, see Non-Patent Document 1 and Japanese Patent Application No. 2001-31329).
[0005]
Using this technique, experiments were performed to detect differences between multiple claims of a single patent. This is a new attempt. The two claims of a patent were transformed so that one word per line was inserted, and their Mdiff was taken (in the following explanation, the brackets in the claims etc. are "[" or " ] ”).
[0006]
Example 1,
17. The printer system control method according to claim 16, wherein the printer system includes a host device.
18. The printer system control method according to claim 16, wherein the printer system includes a printer.
[0007]
(Mdiff result of Example 1 above)
The printer system
; ===== begin =====
Host device
─────────
Printer
; ===== end =====
17. The printer system control method according to claim 16, further comprising:
.
[0008]
The difference between claim 17 and claim 18 can be understood very easily from the results of Mdiff of claims 17 and 18 of Example 1 above. That is,; ===== begin ===== is the beginning of the difference part,; ===== end ===== is the end of the difference part, and ──────── is the difference Represents the boundary between two data. Here, the difference is “higher-level device” and “printer”. However, if the differences are more confusing, the Mdiff results will be difficult to see.
[0009]
Example 2,
[Claim 1]
A weeding sickle characterized in that the blade portion at the tip of the blade member is formed into an irregular shape to form a wavy blade and is curved spirally, and a handle is attached to the blade member.
[Claim 2]
A weeding sickle characterized by having anti-slip parts on the upper and lower parts of the handle.
[0010]
(Mdiff result of Example 2 above)
; ===== begin =====
Blade member
─────────
Handle
; ===== end =====
of
; ===== begin =====
Tip blade
─────────
Anti-slip on top and bottom
; ===== end =====
Part
; ===== begin =====
A wavy blade is formed on an uneven surface and curved in a spiral, and a handle is attached to the blade member.
─────────
Establishment
; ===== end =====
Weeding sickle characterized by that.
[0011]
The difference between the results of taking Mdiff in claim 1 and claim 2 in Example 2 is complicated, so the result of Mdiff is difficult to see. In other words, since Mdiff is a mechanism for storing order information, when the difference is complicated, it is difficult to understand the difference, and it has been found that there is a problem with this as it is.
[0012]
[Non-Patent Document 1]
Masaki Murata, 1 outside, diff and language processing “Language understanding and communication
The Institute of Electronics, Information and Communication Engineers July 17, 2001 (NLC2001-26) Electronics
IEICE technical report, p. 29-36
[0013]
[Problems to be solved by the invention]
In the case of using the conventional Mdiff, when the difference is complicated, the result of the Mdiff is difficult to see.
[0014]
An object of the present invention is to solve the above-described problems and to provide an easy-to-understand display even when the difference is complicated.
[0015]
[Means for Solving the Problems]
FIG. 1 is a diagram illustrating the principle of the present invention. In FIG. 1, 2 is an extraction means, 3a is a storage means, and 21 is an extraction / detection area setting means.
[0016]
The present invention has the following means in order to solve the conventional problems.
[0017]
(1): An input means for inputting information; A detection area that is a unit of an area to be compared in order to detect a difference between an extraction unit that is a target unit to be output as a difference between document data and a difference between document data And the input means Setting Be done An extraction / detection area setting means 21, a storage means 3a for storing information, and an extraction means 2 are provided, and the extraction means 2 extracts all the extractions from areas other than the current detection area of the input document data. A unit corresponding to the unit is extracted and stored in the storage unit 3a, and in the current detection area, a unit corresponding to the extraction unit that is not stored in the storage unit 3a is highlighted to display the current detection area. Output of the document is repeated for each detection area. Therefore, it is possible to easily extract and display the document features and differences that are new information.
[0018]
(2): An input means for inputting information; A detection area that is a unit of an area to be compared in order to detect a difference between an extraction unit that is a target unit to be output as a difference between document data and a difference between document data And the input means Setting Be done An extraction / detection area setting means 21, a storage means 3a for storing information, and an extraction means 2 are provided. The extraction means 2 stores the input document data in the storage means 3a in the current detection area. The display of the current detection area is highlighted by highlighting the uncorresponding extraction unit, and the highlighted display is stored in the storage means 3a for each detection area. For this reason, it is possible to easily extract and display a unit corresponding to a newly appearing extraction unit (for example, a word).
[0019]
(3): In the document difference detection apparatus according to (1) or (2), the storage unit 3a stores in advance the data of the extraction unit that is not highlighted. For this reason, expressions that are not so important in advance can be prevented from being highlighted and can be easily viewed.
[0020]
(4): In the document difference detection apparatus of (1) to (3), a word unit is used as the extraction unit. For this reason, a newly appearing word can be extracted and displayed.
[0021]
(5): In the document difference detection device of (1) to (4), the unit of the detection area is an itemized unit. For this reason, the difference between the bullets can be easily understood.
[0022]
(6): In the document difference detection apparatus according to (1) to (4), the unit of the detection area is a unit of claims. For this reason, the features and differences of the claims can be easily understood.
[0023]
(7): A detection area that is a unit of an area to be compared in order to detect a difference between an extraction unit that is a unit to be output as a difference between document data and a difference between document data And by input means Setting Be done Extraction / detection area setting means 21 and all of the input unit corresponding to the extraction unit are extracted from areas other than the current detection area and stored in the storage means 3a. In order to cause the computer to function as the extraction unit 2 that repeats, for each detection area, the output of the document in the current detection area by highlighting the unit corresponding to the extraction unit that is not stored in the storage unit 3a. Or a computer-readable recording medium on which the program is recorded. For this reason, it is possible to easily provide a document difference detection apparatus that can easily extract and display document features and differences by installing this program in a computer.
[0024]
(8): A detection area that is a unit of an area to be compared in order to detect a difference between an extraction unit that is a unit to be output as a difference between document data and a difference between document data And by input means Setting Be done The extraction / detection area setting means 21 and the current detection area of the input document data that are not stored in the storage means 3a are highlighted and the document in the current detection area is output. Then, the storage means 3a is stored in the storage means 3a as the extraction means 2 for repeating each detection area, and a program for causing a computer to function or a computer-readable recording medium recording the program is used. For this reason, it is possible to easily provide a document difference detection apparatus that can extract and display a unit corresponding to a newly appearing extraction unit by installing this program in a computer.
[0025]
DETAILED DESCRIPTION OF THE INVENTION
(1): Description of document difference detection apparatus
FIG. 2 is an explanatory diagram of the document difference detection apparatus. In FIG. 2, the document difference detection apparatus is provided with an input means 1, an extraction means 2, an extract storage device 3, and an output means 4. The input unit 1 inputs information such as a keyboard, a mouse, and a reading device. The extraction unit 2 extracts the difference between the input documents. The extract storage device 3 is an extract storage means for storing extracts such as words, kanji and noun phrases. The output means 4 outputs information such as a display device and a printer.
[0026]
(1) Explanation of morphological analysis system
In order to divide Japanese into words, a morphological analysis system performed by the extraction unit 2 is required. Here, ChaSen will be explained (published at the tea ceremony http://chasen.aist-nara.ac.jp/index.html.jp developed at Nara Institute of Technology).
[0027]
This splits the Japanese sentence and also estimates the part of speech of each word. For example, if “go to school” is entered, the following results can be obtained.
[0028]
School Gakkou School Noun-General
To He To Particles-Case particles-General
Go Iku Go Verb-Independence
EOS
In this way, each line is divided so that one word is included, and reading and part-of-speech information are given to each word. In the extraction means 2 of the present invention, only the part (morpheme analysis means) for dividing the word in this function is used.
[0029]
▲ 2 ▼: Explanation of English stemmer
In order to extract a word by the extraction means 2, since the word is written in English, it is only necessary to perform stemming to return the word to the basic form. There is a famous Porter algorithm for this stemming (see Porter, MF, 1980, An algorithm for suffix stripping, Program, 14 (3): 130-137).
[0030]
(3): Explanation of extraction unit and detection area
Characters, paragraphs, sentences, bullet items, etc. can be mechanically recognized from the document format. For example, a character can be recognized by a 1-byte or 2-byte code. If it is a paragraph, it can be recognized by indentation and line feed. A sentence can be recognized by the presence of a period or a period. The bullets can be recognized by indentation, line breaks, symbols at the beginning of bullet items, and the like. Word recognition is performed by the morphological analysis system and stemmer mentioned above. The recognition can be performed, for example, by providing each recognition means in the extraction means.
[0031]
(2): Explanation of difference detection
There are two methods (methods) for detecting the difference of the present invention. These methods do not use the Diff command. Hereinafter, these two methods will be described with reference to flowcharts.
[0032]
(1): Method 1
FIG. 3 is a flowchart of document difference detection processing according to Method 1. Hereinafter, description will be made according to the processing S1 to S3-2 of FIG.
[0033]
S1: An extraction unit (extraction unit) and a detection area unit are determined in advance by the input means 1 or the like. An extraction unit is a unit to be output as a difference. The extraction unit may be “word”, “kanji”, “noun phrase”, or the like. The unit of detection area is a unit of area to be compared in order to detect a difference. As a unit of the detection area, “character”, “word”, “sentence”, “item of item”, “paragraph”, “claim of patent”, and the like can be considered.
[0034]
S2: The extraction means 2 stores all input data in the storage means (in the extraction means 2).
[0035]
S3: The extraction unit 2 examines the input data from the left and repeats the following processing S3-1 and processing S3-2 for each detection region determined in the processing S1 from the left detection region.
[0036]
S3-1: The extraction unit 2 extracts all the extraction units (for example, words) corresponding to all the extraction units from all the regions other than the current detection region, and stores them in the extract storage device 3.
[0037]
S3-2: The extraction unit 2 outputs a sentence in the current detection region by highlighting an item (for example, a word) corresponding to an extraction unit that is not stored in the extract storage device 3 in the current detection region. Output to.
[0038]
(2): Explanation according to the method 1 example
An example of method 1 will be described with the extraction unit as a word, taking the claims (detection region) of the patent specification as an example. All words are extracted from all claims other than the currently analyzed claim, and words that do not appear in other claims are identified in the currently analyzed claim. The results are shown in Example 3 below.
[0039]
Example 3
[Claim 1] The "blade" part of the "tip" of the "blade member" is "formed and corrugated" to be "formed and wavy blade" and "curved" and "curved", and a handle is attached to the blade member A weeding sickle characterized by that.
[Claim 2] A weed sickle characterized in that a "slip prevention" portion is "provided" on the "upper and lower parts" of the handle.
[0040]
In Example 3 above, words that did not appear in other claims are enclosed in parentheses “<<” and “>>” (highlighted). This result is much easier to see than the Mdiff result in Example 2. From Example 3, it can be understood very easily that the feature of [Claim 2] is "the upper and lower anti-slip portions". If it can be understood that the feature of claim 2 is “anti-slip part”, the embodiment including this term “anti-slip part” and paragraphs in the examples can be easily extracted to claim 2. Embodiments and examples corresponding to the above can also be extracted.
[0041]
Thus, this method is very useful for extracting features and differences. Further, it is useful for extracting embodiments and examples corresponding to certain claims, that is, for associating claims with the embodiments and examples.
[0042]
Next, I tried Method 1 for another example with three claims. In this case, the result as shown in Example 4 below was obtained.
[0043]
Example 4
[Claim 1] The "blade" portion of the "tip" of the "blade member" is made into "uneven" and "formed and wavy blades" and "curved" and "curved", and a handle is attached to the blade member A weeding sickle characterized by that.
[Claim 2] A weeding sickle characterized in that an anti-slip portion is provided on the upper part of the handle.
[Claim 3] A weeding sickle characterized in that an anti-slip portion is provided on the upper and lower parts of the handle.
[0044]
As a result of Example 4 above, the “slip prevention portion” that is the feature of Claims 2 and 3 could not be extracted. In order to solve this problem, a second new method (Method 2) was considered.
[0045]
(2): Method 2
FIG. 4 is a flowchart of the document difference detection process of method 2. Hereinafter, a description will be given according to the processing S11 to S12-2 of FIG.
[0046]
S11: An extraction unit (extraction unit) and a detection area unit are determined in advance by the input means 1 or the like. An extraction unit is a unit to be output as a difference. The extraction unit may be “word”, “kanji”, “noun phrase”, or the like. The unit of detection area is a unit of area to be compared in order to detect a difference. As a unit of the detection area, “character”, “word”, “sentence”, “item of item”, “paragraph”, “claim of patent”, and the like can be considered.
[0047]
S12: Input data is input from the input means 1 for each detection region determined in the process S11, and the extracting means 2 repeats the following processes S12-1 and S12-2.
[0048]
S12-1: The extracting means 2 outputs a sentence in the current detection area by highlighting an extraction unit (for example, a word) corresponding to an extraction unit that is not stored in the extract storage device 3 in the current detection area. Output to. However, the extract storage device 3 is initially empty.
[0049]
S12-2: The expression highlighted in the process S12-1 is stored in the extract storage device 3.
[0050]
(2): Explanation by example of method 2
An example of method 2 will be described with the extraction unit as a word, taking the claims of the patent specification as an example. The second new approach takes all words from all claims above the one that is being analyzed, and all above the claims that are currently being analyzed in the claim that is being analyzed. Identify words that do not appear in the claim. The results are shown in Example 5 below.
[0051]
Example 5
[Claim 1] << A weeding sickle characterized in that the blade part at the tip of the blade member is formed into an irregular shape to form a wave-like blade and is curved in a spiral shape, and a handle is attached to the blade member. >>
[Claim 2] A weeding sickle characterized in that a "non-slip" part is "provided" on the "upper part" of the handle.
[Claim 3] A weeding sickle characterized in that an anti-slip portion is provided on the upper and lower parts of the handle.
[0052]
In this case, the “non-slip portion” that is the feature of claims 2 and 3 could be extracted. By this method, a newly appearing word can be extracted as a difference.
[0053]
-An example of an ordinary sentence using Method 2 will be described. Here, both the unit of extraction and the unit of detection area are words.
[0054]
Example 6
《The purpose of this study is to use ‘changed’ and ‘machine particle’ as ‘changed’ in ‘passive sentence’ in ‘Japanese’ and ‘converting’ to ‘active’ sentence. Automatically. >> “Examples” of Japanese passive sentences and usage sentences are given in “Figure 1” and “Figure 2”. The Japanese << suffix ">>"<<"in the sentence in Figure 1 is" passive "is a" showing auxiliary verb "," is ", and" this "is a passive sentence. The Japanese suffix “<< se” ta ”in the sentence in Fig. 2 is an auxiliary verb indicating a working part, and this sentence is a working sentence. Figure << 3 >> shows an active sentence that corresponds to << the >> sentence. When the sentence << in Figure 1 is transformed into an active sentence, << (i) >> the case particle "ni" is the case particle "ga", << (ii) >> the case particle "ga" is the case particle " Is converted to. When the sentence in Fig. 2 is converted to an active sentence, (i) the << part >> of the case particle "ga", << the clause >> of << he >> is << erased >>, and (ii) the case particle "ni" Is converted to the case particle "ga", and << (iii) >> In this study, conversion of these case particles <<(>> example <<: >> case << auxiliary >><< verb >> conversion of the case particle "ga""to"<<)>> and elimination of the << unnecessary >> part ( Example: Erasing “He is” is the “subject” of the study. (From 《After》 and 《This paper》 《Convenience》 《He is》 《Erase》 《Erase》 《Also》 《Call》 Conversion of case particles.
The conversion of passive sentences and active sentences into active sentences is performed by using the sentence << Generation >>, << paraphrase >>, sentence << simplification / language >><< operation support >>, << natural >> language sentence << from >><< knowledge acquisition and information extraction 》, 《Question Answering System》 and 《Many》 Research 《Fields》 are 《Useful》. << For example >>, in the question answering system, the question sentence is a << noh >><< motion >> sentence and the << answer >> is << written >> with a << passive >> sentence. There are cases where it is “difficult” to “take out” an answer to a question because the “structure” of the sentence is “different”. These “like” “questions” and “titles” also become “resolve” and “solve” so that “passive” and active sentences can be converted into active sentences. In this way, the conversion of passive sentences and usage sentences into active sentences is “important” in the natural language “processing”.
[0055]
From the display of this example 6, it can be seen in the second paragraph that a story such as “Generation”, “Parallel”, “Simplification / Language”, etc. is newly generated. Also, in the second paragraph, the language expression used in the evaluation of “useful”, “difficult”, “can”, “important”, etc. is often used, so the validity and usefulness of the method is described. It is easy to understand that
[0056]
An example of a detailed description of the invention using method 2 will be described. Here, both the unit of extraction and the unit of detection area are words.
[0057]
Example 7
Next, the present invention will be described with reference to the drawings. FIG. 1 is a front view of the weeding sickle according to the present invention, FIG. 2 is a rear view of the weeding sickle according to the present invention, and FIG. 3 is a right side of the weeding sickle according to the present invention. It is.
<< [0007] >>
The weeding sickle 1 is formed as shown in FIG. 3 <<, and the blade << part >> 2 << b >> of the << blade member >> 2 is << formed to the wavy blade << 5 >> of the << front end >><< one side is wavy >>. The blade member 2 << and >> whose back is formed to be "flat" and "handle" 3 << are constructed.
[<< 0008]
As shown in FIGS. 1, 2 << and >> FIG. 3, the blade member 2 has an << extension >> portion 2 << a >><< the length >> of the handle 3 << about >> 2 << double >>. . The blade portion 2b of the wavy blade 5 is << curved >> in one direction.
[<< 0009 >>]
Fig. << 4 >> is a front << enlarge >> figure of the blade portion of the weeding sickle of the present invention. As shown in the figure, the blade portion 2b that << reapses << weed >> has a << convex >> portion 5a and a << concave >> portion 5b << existing >> alternately and is wavy.
[<< 0010 >>]
FIG. 5 is an enlarged view of the blade portion of the weeding sickle according to the present invention. The tip of the convex portion 5a that constitutes the blade portion 2b is << inclined >> slightly << to the left. “This” means “to hook” the weeds to “make it easy to mow”.
[<< 0011 >>]
Fig. << 6 >> is an enlarged view of << the part >> showing the curve << state >> of the blade portion of the weeding sickle according to the present invention. As shown in the figure, the tip 2 << c >> of the blade 2b is more curved from the << vertical line >> 6 than the extension 2a of the blade 2b.
[<< 0012 >>]
The figure << 7 >> is a << cross section >> figure taken along the <<A->> A line of the figure << middle >>. The “upper surface” 7 of the blade portion 2 b is “inclined”, and the tip 5 c of the convex portion 5 a is “pointed”. << And >>, the blade part 2b << itself >> is curved and << spirals >>.
[<< 0013 >>]
Figure << 8 >> is a front view of << Other >><< Example >> of the weeding sickle according to the present invention, Figure << 9 >> is a rear view of another Example of the weeding sickle according to the present invention, and Figure << 10 >> is A right side view of another embodiment of the weeding sickle according to the present invention, FIG. 11 is a partially enlarged view of another embodiment of the weeding sickle according to the present invention.
[<< 0014 >>]
In the weeding sickle 1a of this example, the extension part 2a of the blade member 2 is << short >> and the << part >> of the blade part 2b is slightly << large >> formed.
[<< 0015 >>]
<< Also >> The handle 3 is << long >> and << the top >> of the << grip >> part 3b is << prepared >> with an << anti-slip >> part 3a with a diameter slightly larger than the << diameter >> of the grip part 3b. >> is also provided with << slip >> stopper 3 << cb >> below << large >> diameter from grip part 3b.
[<< 0016 >>]
As shown in FIG. 10, the blade portion 2b of the weeding sickle 1a of this example is also curved in a spiral shape like the weeding sickle 1 shown in FIGS.
[<< 0017 >>]
Like this, it is easier and easier to “uproot” the weeds that are “grown” on the “lawn etc.” by making the tip curve spirally.
[<< 0018 >>]
[0058]
In Example 7, it can be seen that “spiral” is first appearing here at paragraph number 0012. In paragraph 0015, it is understood that the “slip prevention part” is important. Also, in paragraph number 0017, an interesting expression “remove roots” is found here for the first time.
[0059]
-An example of English text by Method 2 will be explained. Here, both the unit of extraction and the unit of detection area are words. Also, stemming was not performed, and words were recognized based on being separated by spaces.
[0060]
Example 8
《In the PATENT task of NTCIR-3, we participated in》 the 《optional task,》 《where》 the 《participants can perform any kind》 of 《research related to》 《patents. We think that》 in 《a》 PATENT 《 attempt,》 the optional task 《is very》 《interesting, because》 we 《have already heard》 that 《some》 participants in 《previous contests wanted》 to 《make their studies as freely》 as 《they》 《wanted. Various new ideas or》 new << topics will come up >> in << an >> optional << task. These attempts would be novel and valuable. >> In the << other >><< contests, too, >> we << hope that << such >> attempts will be << made .
In 《this contest,》 we 《made》 the 《following three》 studies 《for》 the optional task of 《PATENT.》
We 《extracted rewriting rules using data》 of patents.
We 《aligned》 the 《claim》 of a 《patent》 and 《its embodiment.》
We extracted << differences among plural claims >> in a << patent. >><< The first two >> topics << were given by organizers >> of PATENT as << examples >> of the optional task.We << consider these >> studies to be very << interesting. >> The << last topic >> is << our idea. >> We << sometimes write >> a << patent, >> and << had >> the << experience >> of << wanting >> to << know >> the << difference >> of << claims. So, >> we << did》 this 《study.》 We have 《been studying natural language processing》 using the 《Unix》 《command Diff.》 We 《previously proposed ways》 to 《use Diff》 in natural language 《processing.》 The Diff command is very 《 suitable》 for 《doing》 the 《above》 three 《studies.》 We have already extracted rewriting rules by using Diff in some research 《topics. For example,》 we 《used》 a 《pair》 of 《definition sentences having》 the 《 same word entry》 in two 《different》 《dictionaries》 and extracted the differe nces 《between them.》 Theseextracted differences can be used as 《synonym phrases》 because the definition sentences in the same entry have the same 《meaning.》 In 《another situation,》 we used aligned 《spoken-language》 and 《written-language These extracted differences can be used as rewriting rules 《transforming》 spoken-language sentences 《into》 written-language sentences or transforming written-language sentences into spoken-language 《sentences.》 Diff can 《also》 be used for << alignment. >> Diff << has >> a << function >> of << merging >> data << like >> a << DP-matching algorithm.So >> we can << align >> two related texts by using Diff. In this << study, >> we used this function for the 《alignment》 of a patent claim and its 《embodiment (working》 《example). Finally,》 we used Diff for 《extracting》 the differences of patent claims. 《Extracting》 differences is an 《original》 function of Diff.Extracting differences between claims 《enables us》 to 《understand》 the claims of a patent 《more deeply.》
[0061]
In Example 8, the main word in the itemized section is emphasized in the itemized item in the middle. That is, (<< extracted rewriting rules using data >>, << aligned >> the << claim >> of a << patent >> and << its embodiment. >> and << differences among plural claims >>) can easily understand the main points of each item.
[0062]
In the last paragraph, you can see that the story of Diff begins. Also, key keywords (key phrases) such as << definition sentences >><< synonym phrases >><< spoken-language >><< written-language texts >><< DP-matching algorithm >> are immediately visible. Convenient for understanding contents.
[0063]
(3): Description of a document difference detection apparatus provided with a user dictionary
Each user has a user dictionary in advance and does not emphasize anything in the dictionary. As a result, it is possible to make it easy to see unimportant expressions without emphasizing them in advance.
[0064]
FIG. 5 is an explanatory diagram of a document difference detection apparatus provided with a user dictionary. In FIG. 5, the document difference detection apparatus is provided with an input means 1, an extraction means 2, an extract storage device 3, an output means 4, and a user dictionary 5. The input unit 1 inputs information such as a keyboard, a mouse, and a reading device. The extraction unit 2 extracts the difference between the input documents. The extract storage device 3 is an extract storage means for storing extracts such as words, kanji and noun phrases. The output means 4 outputs information such as a display device and a printer. The user dictionary 5 is a dictionary registered in advance by each user.
[0065]
(1): Explanation of Method 1 for providing a user dictionary
FIG. 6 is a flowchart of a document difference detection process of Method 1 for providing a user dictionary. Hereinafter, a description will be given according to the processes S21 to S23-2 of FIG.
[0066]
S21: The extraction unit (extraction unit) and the detection area unit are determined in advance by the input means 1 or the like, and the user dictionary 5 is registered. An extraction unit is a unit to be output as a difference. The extraction unit may be “word”, “kanji”, “noun phrase”, or the like. The unit of detection area is a unit of area to be compared in order to detect a difference. As a unit of the detection area, “character”, “word”, “sentence”, “item of item”, “paragraph”, and the like can be considered.
[0067]
S22: The extraction unit 2 stores all input data in the storage unit (in the extraction unit 2).
[0068]
S23: The extraction unit 2 examines the input data from the left and repeats the following processing S23-1 and processing S23-2 for each detection region determined in S21 from the left detection region.
[0069]
S23-1: The extraction unit 2 extracts all the extraction units (for example, words) corresponding to all the extraction units from all the regions other than the current detection region, and stores them in the extract storage device 3.
[0070]
S23-2: The extraction unit 2 highlights an extraction unit (for example, a word) that is not stored in the extract storage device 3 and is not stored in the user dictionary 5 in the current detection area. The text in the current detection area is output to the output means 4.
[0071]
(2): Method 2 for providing a user dictionary
FIG. 7 is a document difference detection process flowchart of Method 2 for providing a user dictionary. Hereinafter, a description will be given according to the processes S31 to S32-2 in FIG.
[0072]
S31: An extraction unit (extraction unit) and a detection area unit are determined in advance by the input means 1 and the like, and the user dictionary 5 is registered. An extraction unit is a unit to be output as a difference. The extraction unit may be “word”, “kanji”, “noun phrase”, or the like. The unit of detection area is a unit of area to be compared in order to detect a difference. As a unit of the detection area, “character”, “word”, “sentence”, “item of item”, “paragraph”, and the like can be considered.
[0073]
S32: Input data is input from the input means 1 for each detection area determined in the process S31, and the extraction means 2 repeats the following processes S32-1 and S32-2.
[0074]
S32-1: The extraction unit 2 highlights an extraction unit (for example, a word) that is not stored in the extract storage device 3 and is not stored in the user dictionary in the current detection area. The text in the current detection area is output to the output means 4. However, the extract storage device 3 is initially empty.
[0075]
S32-2: The expression highlighted in the process S32-1 is stored in the extract storage device 3.
[0076]
(3): Explanation of method 2 (other realization methods) for providing a user dictionary
FIG. 8 is a document difference detection process flowchart of Method 2 (another realization method) for providing a user dictionary. Hereinafter, a description will be given according to processing S41 to S43-2 in FIG.
[0077]
S41: An extraction unit (extraction unit) and a detection area unit are determined in advance by the input means 1 and the like, and the user dictionary 5 is registered. An extraction unit is a unit to be output as a difference. The extraction unit may be “word”, “kanji”, “noun phrase”, or the like. The unit of detection area is a unit of area to be compared in order to detect a difference. As a unit of the detection area, “character”, “word”, “sentence”, “item of item”, “paragraph”, and the like can be considered.
[0078]
S42: The extraction means 2 stores all the contents of the user dictionary 5 in the extract storage device 3.
[0079]
S43: Input data is input from the input unit 1 for each detection region determined in the process S41, and the extraction unit 2 repeats the following processes S43-1 and S43-2.
[0080]
S43-1: The extraction means 2 outputs a sentence in the current detection area by highlighting an extraction unit (for example, a word) corresponding to an extraction unit that is not stored in the extract storage device 3 in the current detection area. Output to.
[0081]
S43-2: The expression highlighted in the process S43-1 is stored in the extract storage device 3.
[0082]
-When not using a user dictionary, take a text example (example 9) as follows. Here, using method 2, both the extraction unit and the detection area unit are words.
[0083]
Example 9
《The purpose of this study is to use ‘changed’ and ‘machine particle’ as ‘changed’ in ‘passive sentence’ in ‘Japanese’ and ‘converting’ to ‘active’ sentence. Automatically. >> “Examples” of Japanese passive sentences and usage sentences are given in “Figure 1” and “Figure 2”. The Japanese 《suffix》 《same》 in the sentence of Fig. 1 is 《Passive》.
《Present auxiliary verb》 is 《Yes》, and 《This》 sentence is passive. The Japanese suffix “<< se” ta ”in the sentence in Fig. 2 is an auxiliary verb indicating a working part, and this sentence is a working sentence. Figure << 3 >> shows an active sentence that corresponds to << the >> sentence. When the sentence << in Figure 1 is transformed into an active sentence, << (i) >> the case particle "ni" is the case particle "ga", << (ii) >> the case particle "ga" is the case particle " Is converted to. When the sentence in Fig. 2 is converted to an active sentence, (i) the << part >> of the case particle "ga", << the clause >> of << he >> is << erased >>, and (ii) the case particle "ni" Is converted to the case particle "ga", and << (iii) >> In this study, conversion of these case particles <<(>> example <<: >> case << auxiliary >><< verb >> conversion of the case particle "ga""to"<<)>> and elimination of the << unnecessary >> part ( Example: Erasing “He is” is the “subject” of the study. (From 《After》 and 《This paper》 《Convenience》 《He is》 《Erase》 《Erase》 《Also》 《Call》 Conversion of case particles.
The conversion of passive sentences and active sentences into active sentences is performed by using the sentence << Generation >>, << paraphrase >>, sentence << simplification / language >><< operation support >>, << natural >> language sentence << from >><< knowledge acquisition and information extraction 》, 《Question Answering System》 and 《Many》 Research 《Fields》 are 《Useful》. << For example >>, in the question answering system, the question sentence is a << noh >><< motion >> sentence and the << answer >> is << written >> with a << passive >> sentence. There are cases where it is “difficult” to “take out” an answer to a question because the “structure” of the sentence is “different”. These “like” “questions” and “titles” also become “resolve” and “solve” so that “passive” and active sentences can be converted into active sentences. In this way, the conversion of passive sentences and usage sentences into active sentences is “important” in the natural language “processing”.
[0084]
-As the user dictionary, words that appear frequently in other papers of the inventor are registered.
(User dictionary registration example)
,,,,. , In, is, is, is, is, is, is, is, is, like, part, is, data, difference, is, this is Murata, is, "research can, diff,) , Correspondence, well, system, processing, language, (, file, use, thing
Such as Note that word delimiters in user dictionary registration examples are represented by ",".
[0085]
In this case, the result of Example 9 is as follows.
The “Purpose” of the “Book” study is “Changed” and “Critical particles” to be “changed” when “converting” a “passive sentence” and “usefulness” sentence into an “active” sentence in “Japanese”. It is an “automatic” conversion using machine learning. An example of a passive sentence and a usage sentence in Japanese is given in Figure 1 and Figure 2. The Japanese “suffix” and “re” in the sentence in Figure 1 are “passive”.
《Present auxiliary verb》 is 《Yes》, and 《This》 sentence is passive. The Japanese suffix “<< se” ta ”in the sentence in Fig. 2 is an auxiliary verb indicating a working part, and this sentence is a working sentence. Figure << 3 >> shows an active sentence that corresponds to << the >> sentence. When the sentence in Fig. 1 is transformed into an active sentence, << (i) >> the case particle "ni" becomes the case particle "ga"<< (ii) >> the case particle "ga" becomes the case particle "" Converted. When the sentence in Fig. 2 is converted to an active sentence, (i) the part of the case particle "ga""<< he >>"<< clause >> is << erased >> and (ii) the case particle "ni" is The particle “ga” is converted into << (iii) >> and the case particle “ha” is converted into << less >><< remains as it is >>. In this study, conversion of these case particles (example <<: >> case << help >><< noun >> conversion of the case particle "ga"<< to ") and removal of the << unnecessary >> part (example:" he Is the “subject” of the study. (In 《After》 and 《This paper》, 《Convenience》 is also called `` Conversion of case particles ''.
The conversion of passive sentences and active sentences into active sentences is performed by using the sentence << Generation >>, << paraphrase >>, sentence << simplification / >> language << operation support >>, << natural >> language sentence << from >><< knowledge acquisition and information extraction 》, 《Question Answering》 system and 《Many》 research 《Fields》. << For example >> In a question answering system, a question sentence is a << noh >><< motion >> sentence, and an << answer >> is a << passive >> sentence << written >>, a question sentence and an answer << including >> In some cases, it is difficult to “take out” the answer to a question because the “structure” of the sentence is “different”. Such “questions” and “titles” are also “solved” and “resolved” so that passive sentences and usage sentences can be converted into active sentences. In this way, the conversion of passive sentences and usage sentences into active sentences is «important» in natural language processing.
[0086]
Although the above results do not change so much, for example, the less important expressions such as the first “research” and “it is” are not taken, and it becomes a little easier to see. By registering more unimportant words in the user dictionary 5, it is possible to make it easier to see.
[0087]
In the above-described embodiment, description has been given with double angle brackets as the highlight display, but other highlight displays such as underline, color coding, background change, font change, and blinking can also be performed.
[0088]
Such a method can also be used to consider new and old information issues in anaphora analysis. The only method that can be used for this “problem between new information and old information in anaphora analysis” is “method 2”, and method 1 cannot be used. Next, in the case of “method 2”, a newly appearing expression is highlighted, but linguistically, an object that newly appears in such a sentence is called “new information”. Therefore, Method 2 for highlighting newly appearing expressions is to extract new information in terms of linguistics, and the result of Method 2 is the consideration of new information in terms of linguistics. Can also be used. However, in the case of language expression, the same thing may be expressed in different language expressions. In that case, even old information may be highlighted by method 2 because it is a new language expression. That is, not all “new information” and “old information” are correctly distinguished. Nevertheless, Method 2 is useful for considering “new information” and “old information”.
[0089]
Furthermore, by setting the extraction unit as kanji, it is possible to easily understand the appearance of new kanji in school education and the like. In the case of kanji, morpheme analysis means is not required like words because they can be compared with kanji codes.
[0090]
(4): Explanation of program installation
The input means 1, the extraction means 2, the extract storage device 3, the output means 4, the user dictionary 5, the extraction / detection area setting means 21 and the like can be configured by a program and executed by a main control unit (CPU). It is stored in the main memory. This program is generally processed by a computer. This computer is composed of hardware such as an input device which is an input means such as a main control unit, a main memory, a file device, a display device, and a keyboard. The program of the present invention is installed on this computer. In this installation, these programs are stored in a portable recording (storage) medium such as a floppy disk or a magneto-optical disk, and a drive device for accessing the recording medium provided in the computer is used. Alternatively, it is installed in a file device provided in the computer via a network such as a LAN. Then, the program steps necessary for processing are read from the file device into the main memory and executed by the main control unit.
[0091]
【The invention's effect】
As described above, the present invention has the following effects.
[0092]
(1): The extraction unit extracts all the extracted unit corresponding to the extraction unit from the region other than the current detection region of the input document data and stores it in the storage unit. Easily extract and display document features and differences, which are new information, in order to repeat for each detection area the output corresponding to the extraction unit that is not stored and highlight and output the document in the current detection area Can do.
[0093]
(2): The extraction means highlights the current detection area of the input document data corresponding to the extraction unit that is not stored in the storage means, and outputs the document in the current detection area. Since storing the display in the storage means is repeated for each detection area, it is possible to easily extract and display a display unit (for example, a word) corresponding to a newly appearing extraction unit.
[0094]
(3): Since the data of the extraction unit that is not highlighted in advance is stored in the storage unit, expressions that are not so important can be prevented from being highlighted in advance and can be easily viewed.
[0095]
(4): Since the extraction unit is a word unit, a newly appearing word can be extracted and displayed.
[0096]
(5): Since the unit of the detection area is a unit of bullets, the difference between the bullets can be easily understood.
[0097]
(6): Since the unit of the detection region is the unit of the claims, the features and differences of the claims can be easily understood.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating the principle of the present invention.
FIG. 2 is an explanatory diagram of a document difference detection apparatus according to an embodiment.
FIG. 3 is a document difference detection processing flowchart of Method 1 in the embodiment;
FIG. 4 is a document difference detection processing flowchart of Method 2 in the embodiment;
FIG. 5 is an explanatory diagram of a document difference detection apparatus provided with a user dictionary in the embodiment.
FIG. 6 is a document difference detection process flowchart of Method 1 for providing a user dictionary in the embodiment;
FIG. 7 is a document difference detection processing flowchart of Method 2 for providing a user dictionary in the embodiment;
FIG. 8 is a document difference detection process flowchart of Method 2 (another realization method) for providing a user dictionary in the embodiment;
[Explanation of symbols]
2 Extraction means
3a Storage means
21 Extraction / detection area setting means

Claims (8)

An input means for inputting information;
An extraction / detection area setting means in which an extraction unit that is a unit to be output as a difference between document data and a detection area that is a unit of an area to be compared in order to detect a difference between document data are set by the input means ;
Storage means for storing information;
Extraction means,
The extraction means extracts all of the extraction unit corresponding to the extraction unit from areas other than the current detection area of the input document data, stores the extracted data in the storage means, and the storage means in the current detection area A document difference detection apparatus that repeats, for each of the detection areas, highlighting the one corresponding to the extraction unit that is not stored in the document and outputting the document of the current detection area. An input means for inputting information;
An extraction / detection area setting means in which an extraction unit that is a unit to be output as a difference between document data and a detection area that is a unit of an area to be compared in order to detect a difference between document data are set by the input means ;
Storage means for storing information;
Extraction means,
The extraction means highlights the current detection area of the input document data corresponding to the extraction unit that is not stored in the storage means, outputs the document in the current detection area, and An apparatus for detecting a difference in a document, characterized in that storing the displayed item in the storage means is repeated for each detection area. 3. The document difference detection apparatus according to claim 1, wherein the data of the extraction unit not highlighted is stored in advance in the storage means. The document difference detection apparatus according to claim 1, wherein the extraction unit is a word unit. The document difference detection apparatus according to claim 1, wherein the unit of the detection area is a unit of bullets. The document difference detection apparatus according to claim 1, wherein the detection area unit is a unit of a claim. An extraction / detection area setting unit in which an extraction unit that is a unit to be output as a difference between document data and a detection area that is a unit of an area to be compared in order to detect a difference between document data are set by an input unit ;
The input document data is extracted from the areas other than the current detection area corresponding to all the extraction units and stored in the storage means, and the current detection area is not stored in the storage means. As an extraction unit that repeats for each detection area, highlighting the one corresponding to the extraction unit and outputting the document of the current detection area,
A program that allows a computer to function. An extraction / detection area setting unit in which an extraction unit that is a unit to be output as a difference between document data and a detection area that is a unit of an area to be compared in order to detect a difference between document data are set by an input unit ;
In the current detection area of the input document data, the document corresponding to the extraction unit not stored in the storage means is highlighted to output the document in the current detection area, and the highlighted one is stored in the storage area. As an extraction means that repeats storing for each detection area,
A program that allows a computer to function.

Images