JPH0944497A - Title sentence preparing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically prepare a title sentence and to save manpower by judging whether a phrase which is possible to be adopted as a title sentence between extracted subject and predicate or not. SOLUTION: A WRITE condition decision part 17 decides as to whether each of the title sentence of the subject and predicate that a WHO discovery part 13 and a predicate discovery part 14 extract and the title sentence including the adaptable phrase that a first stage and second stage WHAT/WHERE discovery parts 15 and 16 extract can be adopted as the title sentence or not. A 40- character compression part 18 performs a number of character compression processing within 40 characters for the sentence which is made adaptable in the WRITE condition decision part 17 when this sentence exceeds 40 characters. The adaptability as the title sentence is decided by performing the dicision for the WRITE condition decision part 17 once from the predicate discovery part 14. When the title sentence is adapted, the title sentence as it is outputted when the number of character is within 40 characters. When the number of character exceeds 40 characters, a number of character compression processing is performed in the 40-character compression part 18.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for automatically creating a title sentence provided in a newspaper article database system from an article body.

[0002]

2. Description of the Related Art In such a newspaper article database,
There is a limit to the number of characters in the title sentence (up to 40 characters), and it is necessary to extract the information necessary for the title sentence from the article body within that number of letters. That is, who and who
You need to create a title sentence that has information about where and what.

Conventionally, the title sentences for these database systems have been manually created by experts. The expert creates the title sentence while referring to the text by referring to the headline attached to the article.

[0004]

[Problems to be Solved by the Invention] The headline of a newspaper article is
Since it is often the mission to make the reader read the text, it is not faithful to the text. In the case of manual intervention, the personality of the expert who creates the title sentence is reflected in the title sentence. In the database system, the title sentence is the most important item for the user to judge necessary information. Therefore, the title sentence as the database information is required to be a sentence that is true to the body and is standardized (having a certain quality).

The newspaper article database loses its original mission unless a large number of articles that occur every day are made into a database in a shorter time. Therefore, it is necessary to create a title sentence and register it in the database within a fixed time. In order to do this kind of work manually, it is necessary to train a large number of specialists for database creation, and the worker has to do the same work every day while being chased by time. Accompanied by.

The present invention solves such a problem, and an object of the present invention is to automatically create a title sentence from an article and promptly provide it to a database so as to maintain promptness as a newspaper article. Further, another object of the present invention is to automatically create a title sentence to save labor and prevent variations caused by manually creating the title sentence.

[0007]

The present invention requires a candidate sentence extracting means for extracting a candidate original sentence suitable for creating a title sentence from the beginning of the text of an original article in sentence units, and a title sentence from the extracted candidate original sentence. A subject and predicate extraction means for discovering and extracting a phrase representing a subject and a phrase representing a predicate, a determination means for determining whether or not the extracted phrase matches as a title sentence, and a matching sentence as a title sentence by this determination means Compression means for compressing the judged sentence to the number of characters required for the title sentence, and the judging means judges whether or not there is a phrase that can be adopted as a title sentence between the extracted subject and predicate. It is characterized by including.

When the determining means determines that there is a phrase that can be used as a title sentence between the subject and the predicate, the first object indicating the object or the complement of the sentence existing between the subject and the predicate is given. A first WHAT finding means for searching whether or not there is a WHAT adoption word of, and the judging means is the first WH found by the first WHAT finding means.
It is preferable to include means for determining whether or not there is a phrase that can be adopted as a title sentence between the AT adopted phrase and the subject.

Further, when the determining means determines that there is a phrase that can be adopted as a title sentence between the first WHAT adopted word and the subject, a sentence exists between the subject and the first WHAT adopted word or phrase. Second WHAT representing the object or complement of
It is preferable to include a second WHAT discovery means for searching whether or not there is a adopted phrase.

The subject and predicate finding means finds a subject adoptive word of a vocabulary that follows the word representing the subject in the candidate original sentence, extracts the subject, and predicates the vocabulary that precedes the word representing the predicate in the candidate original sentence. It is preferable to include means for discovering a hieroglyph and extracting a predicate.

It is preferable that a pre-processing unit for deleting unnecessary words or symbols in the title sentence from the adopted candidate original sentence is preferably included in the preceding stage of the subject and predicate finding unit.

First, the terms used in this specification will be described.

First, in the present specification, a sentence refers to a continuous character string that represents one complete meaning that is separated from the beginning of a sentence by a punctuation mark (.) And completes. Further, the adoptable words (WHO adopted words as a subject, predicate adopted words, first WHAT adopted words, second WHAT adopted words, etc.) are words that can be adopted as a title sentence. The words and phrases that can be adopted are as follows. 1. A character string of two or more characters consisting of Kanji, Katakana characters (including the signs "-" (long-sign) and "." (Middle black)), English characters, and numbers. 2.1. Codes sandwiched between the adopted words ",", "・",
"-" (Long center bar), "~", and "&" are also adopted as words constituting the adopted phrase, and the adopted phrase + the left sign + adopted phrase are collectively adopted as a series of adopted phrases. Example. Dalian, China 3. The code "" (brackets) is also included in the adopted words.
Example. "Yamazen" brand 4. Words enclosed in quotation marks (“”), including their signs, are adopted words. Example “Kuse”
(This means that hiragana is not included in the terms that can be adopted due to the condition of 1. However, if it is enclosed in quotation marks, it becomes an acceptable term.) The sending kana table (which conforms to the glossary of newspapers published by the Japan Shimbun Association) can also be adopted. Example. Obviously, in the present invention, first, the original article is read from the beginning on a sentence-by-sentence basis, a sentence that describes the subject of the article body is discovered, and candidate original sentences for creating a title sentence are extracted. Since the article has an inverted triangular sentence structure in which the so-called 5W1H is mentioned at the beginning of the sentence, the step of discovering whether there is a candidate original sentence is performed from the beginning of the sentence. This candidate sentence extraction means excludes sentences that are not qualified as title sentence candidates, and determines whether or not they can be adopted as title candidate sentences based on whether or not there are subject adoptive words and predicate adoptive words in the sentence. To do.

Next, pre-processing is performed to remove unnecessary symbols or the like from the adopted candidate original text or convert specific words into abbreviations.

From the candidate original sentence that has been subjected to this preprocessing, the adoptable phrase immediately before the subject adoptive vocabulary, which is the vocabulary attached to the phrase representing the subject, is cut out together with the subject adoptive verb and adopted as the subject of the title sentence, and the necessary replacement is performed. To do. In addition, an adopted term or a specific predicate (for example, “establish” or “execute”) that follows the predicate adoptive, which is a vocabulary attached to a phrase representing a predicate from the candidate original text.
Etc.) is adopted as a predicate together with the predicate adoptive verb, and the necessary substitution is performed.

It is judged whether or not the adoptable words of the adopted subject and the adoptable words of the predicate may be described as a title sentence, and if there are other adoptable phrases as the title sentence, it is further necessary. The step of acquiring the applicable phrase is performed, and if it is good, the title sentence is created through a compression procedure of compressing the number of characters required for the title sentence.

When there is an employable phrase between the subject and the predicate, it is adopted as a title sentence through the first WHAT discovering means and further through the second WHAT discovering means for discovering whether or not there is an employable phrase. It is determined whether there is a possible phrase.

Even when there is an employable phrase found other than the subject and predicate adoptable phrase, the number of characters is compressed to the number of characters required as the title sentence to create the title sentence.

Note that some newspaper articles include articles for which title sentences cannot be automatically described, such as serialized articles and roundtable articles. For such special articles, the title sentences are manually (manually) manually used. create.

According to the present invention, except for some limited newspaper articles, the title sentence of a newspaper article can be automatically created by performing such a procedure, so that the labor of creating the title sentence can be reduced. Further, the work efficiency of the worker can be improved. Furthermore, since the title sentence can be created in a short time, news articles can be provided to the newspaper article database in a short time.

[0021]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram for explaining one embodiment of the present invention.

In one embodiment exemplifying the present invention, a candidate source text discovering unit 11 for discovering and extracting a candidate source text describing a subject of an article from an original article for creating a title sentence.
And a preprocessing unit 12 that performs preprocessing necessary to create a title sentence in the extracted candidate original sentence, a WHO discovery unit 13 that extracts the subject of the sentence from the candidate original sentence, and a predicate of the sentence from the candidate original sentence. The predicate finding unit 14 and the first stage WHAT that extracts an employable phrase indicating an object or a complement of a sentence when there is an employable phrase other than the subject and the predicate.
WHERE discovery section 15 and the first stage WHAT / WHER
Second stage WHAT / WHERE discovering unit 16 for extracting adoptable phrases other than the adoptable phrases extracted by the E discovering unit 15
And the title sentences of the subject and the predicate extracted by the WHO discovery unit 13 and the predicate discovery unit 14, the first stage WHAT / WH
It is determined whether each of the title sentence including the adoptable phrase extracted by the ERE finding unit 15 and the title sentence including the adoptable phrase extracted by the second stage WHAT / WHERE discovering unit 16 can be adopted as the title sentence. A WRITE condition determining unit 17 and a 40-character compressing unit 18 that compresses the number of characters of this sentence to within 40 characters when the sentence that can be adopted by the WRITE condition determining unit 17 exceeds 40 characters.

As a step, the predicate discovering unit 14 once goes to the WRITE condition determining unit 17 to determine the suitability as a title sentence (whether there is an employable phrase between the subject and the predicate). If the number of characters is within 40 characters, the title sentence is output as it is. If it exceeds 40 characters, the 40-character compression unit 18 performs a character number compression process. If the title text does not match, go to the first stage WHAT / WHERE discovery section 15,
The first stage WHAT / WHERE discovery processing is performed. After the processing of the first stage WHAT / WHERE discovery unit 15,
It goes to the RITE condition determination unit 17 to determine the compatibility with the title sentence, and outputs the title sentence as it is if it is within 40 characters, and if it exceeds 40 characters, the 40 character compression unit 18
The number of characters is compressed with. If the title sentence does not match, the process goes to the second stage WHAT / WHERE discovery unit 16 to perform the second stage WHAT / WHERE discovery process. The result is again judged by the WRITE condition judging unit 17, and when it is within 40 characters, the title sentence is output, and when it exceeds 40 characters, the 40-character compressing unit 18 performs character number compression processing.

The configuration and operation of each part will be described below.

FIG. 2 is a flow chart showing the operation of the candidate original sentence finding section 11. This candidate text finding unit 11
This is for reading an article in sentence units from the beginning of the original article, discovering a sentence that describes the subject of the article body, and extracting candidate original sentences for creating a title sentence. The newspaper article has an inverted triangular sentence structure in which the so-called 5W1H is mentioned at the beginning of the sentence. Therefore, the candidate original text discovering unit 11 performs a process of discovering a candidate original text that becomes a title sentence from the first sentence.

First, the original article is read in sentence units from the beginning (S21). Next, it is determined whether the beginning of the sentence starts with "" (brackets), and in that case, the process proceeds to the next sentence (S22). This is because ““ ”is a sentence starting from a spoken language and is not qualified as a title sentence candidate, and is excluded. Next, it is determined whether or not there is a WHO adopted word in the sentence (S23). If there is no WHO adopted word, there is no subject, so the processing for the next sentence is performed, so step S2
Go to 1.

The WHO adopted words are "ha", "ga" and "hara", and are words and phrases following the word expressing the subject as a vocabulary expressing the subject.

Then, if there is a WHO adopted word in the sentence, it is judged whether or not there is a predicate adopted word after this WHO adopted word (S24). If there is no predicate adopter, there is a subject but no predicate, so go to the next sentence in step S.
Go to 21.

The predicate adjectives are "wo", "ni", "de",
"Ga", "from", "about", "as", "do", "to", "succession", "do something", "to"
And is a phrase that precedes the word that is the predicate as a vocabulary that expresses the predicate.

Further, it is judged whether or not the number of characters after the WHO adopted word in the sentence is within 8 characters (S25).
This is because the number of characters including the predicate after the WHO adopted word is small, and the sentence is excluded from the candidate original sentence because it is not a sentence that describes the subject of the sentence.

Through such a judgment step, the candidate original sentence for creating the title sentence is confirmed and the candidate original sentence is acquired.

Next, the preprocessing executed in advance when creating a title sentence from the acquired candidate original sentence will be described with reference to FIG. This is an unnecessary phrase from the candidate original text,
Prior to the title sentence creation process, processes such as deleting symbols and replacing specific words with abbreviations are performed.

First, the character string in the candidate original sentence is matched with the abbreviation dictionary and replaced with an abbreviation (S31). This means replacing Hitachi with "Hitachi" and Sumitomo Special Metals with "Sumikin". Next, "(photo)", "= photo", "= photo (top)" in the candidate text
The text is deleted by deleting the kanji including the characters of the photo, such as "=" and "()" (S32). further,
Characters in "()" in the candidate sentence are deleted by including characters in parentheses (S33). In addition, "" is left only for the character type of the next condition, and other characters are deleted including key brackets to fill the characters (S34).

1. All kanji, 2. Kanji and Hiragana, 3. Kanji and Katakana, 4. Kanji and English, 5. All Katakana, 6. English letters and katakana, 7. When "device" is included, the WHO (subject) is extracted by the WHO discovery unit 13 from the candidate original text for which such preprocessing has been completed. The operation will be described with reference to FIG.

A Japanese sentence has a main part including the subject at the beginning thereof. Therefore, the process to be described later is performed from the first character of the candidate original sentence toward the rear part, the subject or the main part is found, and the candidate sentence is shaped and adopted for the title sentence. This adopted phrase is called a WHO adopted phrase.

Here, the modifiers are "no", "to",
"Na", "or", "to", "to", "from", "etc." However, "etc." means the second stage WHAT
It is used only when processing in the WHERE detection section. Also,
The company / group extraction nouns mean the words "president", "chairman", "chief", "director", "director", "leader".

First, in the processing of the WHO finding section, the employable phrase immediately before the WHO adopting words (the above "wa", "ga" and "hara") are cut out together with the WHO adopting words, and the WHO adopting words ",". Replaced with (punctuation) and adopted as the subject of the title sentence. When the word immediately before this word is “to” or “ya”, the applicable word immediately before that is also adopted (S4).
1). Example 001. Candidate original text "Matsushita Electric and Matsushita Electronics are in Shanghai, China on the 12th" is adopted phrase "Matsushita Electric and Matsushita Electronics."

Next, if there is "etc." in front of the WHO adopted word, the characters from the first character to the WHO adopted word are adopted (S42). Example 002. Candidate original text "5 manufacturers of different types of machines and metals in the Kumamoto area" is adopted phrase "5 manufacturers of different types of machines and metals in the Kumamoto area."

When the WHO adoption word is surrounded by "" (quotation mark), it is adopted with a mark (S43). Example 003. The candidate text "In the situation where the" double-digit yen appreciation "of one dollar = 90 yen level is becoming established" is adopted as the phrase "" double-digit yen appreciation ".

When there is "by" or "at" immediately before the WHO adoption word, the adoptionable word immediately before that is also adopted. Also, when there is a modifier immediately before the immediately preceding adopted phrase, the immediately preceding adoptable phrase is also adopted (S44). Example 00
4. Candidate original text "The four-pole trade meeting of trade ministers of Japan, the United States, the EU, and Canada held in Los Angeles was on the 11th,"".

When the end of the WHO adopted word is a company / organization extracted noun, the words from the first character to the corresponding word are adopted.
Also, when the WHO adopted words are "both" and "both companies", the words are adopted from the first character. When "immediately before" in "each company" is "no", the applicable phrase immediately before that is also adopted (S45). Example 005. Candidate original text "Keidanren Chairman Shoichiro Toyoda at the 21st Regular Meeting" is adopted as the phrase "Keidanren Chairman Shoichiro Toyoda."

When the WHO adoption phrase is surrounded by "" (brackets), the braces are removed and employed. When there is a modifier immediately before that, the adoptable word immediately before that is also adopted (S46). Example 006. Candidate original text "The Japan-Korea technomart for promoting smooth technology transfer between Japan and South Korea is scheduled for 13 to 4 days" is adopted as the phrase "Japan-Korea technomart". Also, for example 007. "The third sector" Marina Akita ", which manages and operates the marinas of three ports in Akita Prefecture, was established in October" is adopted as the phrase "third sector Marina Akita."

"Professor et al."
When there is a "professor" and there is a modifier immediately before that phrase,
The WHO adopted phrase is not adopted, and the adoptable phrase immediately before the modifier is set as the WHO adopted phrase (S47). Example 008. Candidate original sentence "Professor Shigeru Kurosu of Koyama National College of Technology and others in collaboration with Yamatake Honeywell" is adopted as "Oyama National College of Technology."

There is a modifier immediately before the WHO adoption word,
Just before that is "()" (parentheses), and when there is a company / organization name extraction noun in the parentheses, the applicable phrase immediately before the parentheses is also adopted (S48). Example 009. Candidate original text "Hanamasa (54, Hirai, Edogawa-ku, Tokyo, 51st, Mr. Hiroshi Ono, low-priced wine among retailers)" is adopted as "Hanamasa's low-priced wine," .

When there are a plurality of "no" s immediately before and after the WHO adopted word, and before and after there are available words, the adopted words added to up to two "nos" are both adopted (S
49). Example 010. Candidate original text "The redevelopment project of Daikanyama in Shibuya is nearing its final stage" is adopted as the phrase "Redevelopment project of Daikanyama in Shibuya."

WHO adopted phrase + adapted phrase, or
WHO Adopted phrase + Admissible phrase + "No" + Immediately after the admissible phrase, "to", "jointly with", "jointly invest with",
"In joint venture", "in partnership with", "toward", "in",
When there is "from", adopt all that apply. However, "kara" is rejected when a number is included in the employable word immediately before it (S50). This is because, in such a case, there are two or more words to be adopted as the subject, and both are adopted. Example 011. Candidate original text "Nippon Steel and Finnish steelmaking company Hundia in collaboration with Mitsui & Co., Ltd." are adopted as the phrase "Nippon Steel and Mitsui & Co., Ltd.". Example 012. Candidate original text "Noritake is affiliated with American tableware maker Oneida Inc. in the United States and sold in the United States" is adopted phrase "Noritake, American tableware maker Oneida Inc. in the United States," is adopted. Example 013. Candidate original text "Matsushita Battery is the first charger in the industry for mobile phones"
The adopted phrase is "Matsushita battery, a charger for mobile phones,". Example 014. Candidate original text "Obayashi from Singapore Airlines to new headquarters building" is adopted as the phrase "Obayashi from Singapore Airlines."

Finally, the WHO adopted word and "," (punctuation)
The number of words that can be hired between the numbers is "number" + "year", "month",
When “day” is included, the applicable phrase is deleted (S51). Example 015. Candidate original text "The four-way trade meeting held by the trade ministers of Japan, the United States, the EU, and Canada in Los Angeles was held on the 11th of the morning." , ”.

WHO by such WHO discovery unit 13
After the adopted phrase is extracted, the predicate finding unit 14 extracts a predicate from the candidate original sentence. This process will be described with reference to FIG. A Japanese sentence has a predicate and a predicate including the predicate at the end of the sentence. Therefore, the candidate original sentence is traced back from the end toward the beginning of the sentence, the processing described below is performed, the predicate or predicate is discovered, and the sentence is formatted for the title sentence and adopted. The adopted phrase is called a predicate adopted phrase.

First, check from the beginning of the candidate original sentence,
"Establish", "Implement", "Equipment", "Develop",
If there are phrases such as "open", "raise", and "inquiry", the subsequent character strings are ignored and adopted as a predicate.
The subsequent predicate extraction processing is not performed, and the process goes to the WRITE condition determination unit (S61). This is because a specific predicate is unconditionally adopted as a predicate, and after this specific predicate, it enters the detailed portion of the article and is unnecessary as a title sentence, so it is ignored. Example 016. Candidate original text "Symantec, Inc., a leading US computer and software house, established a wholly owned Japanese corporation, Symantec, and started full-scale sales activities for the Japanese market."
Is adopted as "Symantec, Inc., a wholly-owned Japanese corporation, Symantec, is established."

Next, if there is a predicate adoptive word, which is traced back from the end of the candidate original sentence, and there is an applicable phrase immediately after that, it is adopted together. If there is no applicable phrase immediately after the predicate adoptive word, a check is performed toward the rear part, and if there is an applicable phrase, the character string up to that is adopted. When there is no conflicting adopted phrase, the character string up to "." (Phrase) is adopted except the phrase (S62). Example 017. Candidate original text "The life insurance industry is actively investing in stocks in small and medium-sized enterprises and local companies" is adopted as the phrase "life insurance industry is actively investing in stocks in medium and small-sized enterprises and local companies."

Next, "year", which includes a number in the predicate adopted word,
If there are "month" and "day", they are deleted (S63). Example 018. Candidate original text "Omron Soft is a U with a function to import CAD drawings and images into word processing documents.
Japanese software for NIXWS will be released for two days. "Omron software releases Japanese software for UNIXWS equipped with a function to import CAD drawings and images into word processing documents."

After determining the predicate adoption phrase, the sentence is traced back again and the matching process with the predicate table is performed. If there is a phrase in the predicate table, that phrase is used as the predicate adoption phrase, and the first predicate adoption phrase is rejected (S64). Example 019. Candidate original text "Sanden officially announced that it will start integrated production of scroll compressors for car air conditioners in the United States on February 12, 1995." Adopted phrase "Sanden, integrated production of scroll compressors for car air conditioners in the United States To start. " This is because only "start" is put in the predicate table, and words and phrases below it are deleted to remove unnecessary words and phrases in the title sentence.

When such predicate extraction processing is completed, W
Go to the RITE condition determination unit 17.

The WRITE condition judging unit 17 judges whether or not WRITE may be performed as the title sentence. The WHO adopted word / phrase output from the predicate finding unit 14+
Predicate adopted phrase, 1st stage WHAT / WHERE discovery section 1
WHO adoption phrase output in 5 + 1st stage WHAT
WHERE adopted phrase + predicate adopted phrase, 2nd stage WHAT
・ WHO adopted word / phrase output from WHERE discovery unit 16 + second stage WHAT / WHERE adopted word + first stage W
HATE / WHERE adopted words + predicate adopted words are judged. In addition, the case where the second stage WHAT / WHERE adopted words and phrases are plural is also included. The character string to be processed by the WRITE condition determination unit 17 is called a title candidate sentence.

Of the processing of the WRITE condition judging section 17,
The WRITE condition determination for the output of the predicate finding unit 14 will be described with reference to FIG.

First, in the candidate original sentence, the word W is immediately before the predicate adopted word.
It is determined whether or not it is a HO adoption phrase (S111). YES
In this case, it is determined whether the title candidate sentence is within 40 characters (S112). If it is within 40 characters, it is adopted as the title sentence and the candidate original sentence finding unit 11 is used to process the next sentence (article). Go to (S113). Example 020. The original text of the candidate, "Papkin withdraws from Miyagi prefecture.", Is referred to as the title sentence "Papkin withdraws from Miyagi prefecture." If the title candidate sentence exceeds 40 characters in step S112, the process proceeds to the 40-character compression unit 18. Step S11
In the case of 1 in the candidate original text, if the word immediately before the predicate adopted word is not the WHO adopted word, the first stage WHAT
Go to the WHERE discovery section 15 (S115).

If the WRITE condition determining unit 17 determines that the condition of the title sentence is not satisfied, WHAT.W
Move to WHAT / WHERE discovery processing in HERE discovery section. This process is for adopting the object, complement, etc. in the title sentence. The WHAT / WHERE discovery unit is the first stage WHAT / WHERE discovery unit 15 and the second stage.
The logic is divided into the stage WHAT / WHERE discovery unit 16. Japanese sentences are arranged backward from the end of the sentence in the order of words that represent the important properties of the sentence. Therefore, it operates so as to find the WHAT / WHERE retroactively from the predicate adoption phrase, shape it for the title sentence, and adopt it.

Step S1 of the WRITE condition judging section 17
If there is no WHO adopted word immediately before the predicate adopted word in the judgment of 11, the first stage WHAT • WHERE finding processing in the first WHAT • WHERE finding unit 15 is performed. This first
The processing operation of the stage WHAT / WHERE discovery unit 15 is shown in FIG.
This will be described with reference to FIG.

First, the candidate original sentence is traced back from the predicate adoptive word and the immediately preceding adoptable word is adopted (S71). Example 02
1. Candidate original text "Symantec, a leading US computer / software house, established a wholly-owned Japanese corporation, Symantec, and started full-scale sales activities for the Japanese market."
Established Symantec, a wholly owned Japanese subsidiary. "
In this case, "Symantec" just before "Establishing" is the first W
HAT / WHERE adopted phrase.

Next, if there is a modifier immediately before the first WHAT / WHERE adopted word (hereinafter referred to as the first WHAT adopted word), the adopted word immediately before that modifier is also adopted. If there is a further modifier immediately before the adopted word, the words are successively adopted while going back (S72). However, if the word that can be adopted immediately before "kara" contains a number, it is not adopted.
Example 022. Candidate original text "The life insurance industry is actively investing in stocks in small and medium-sized companies and local companies."
"Actively invest in stocks in the life insurance industry, small and medium-sized companies, and regions". Here, "small and medium-sized enterprises and regional stock investment" is the first W
It is a HAT adoption phrase.

Next, in step S73, the following processing is performed. First, when adoptable words + “XXX” (word / phrase with parentheses) + predicate adopter, adoptable words before the brackets are adopted. Example 023. Candidate original text "Mori Seiki is OA
The surface roughness of the aluminum roll used in the machine is simply cut. Special processing machine "SFL
-10/600 "was developed. "Mori Seiki develops a dedicated processing machine that finishes the surface of aluminum rolls used in OA equipment with high precision with a surface roughness of 0.2S simply by cutting." Here, the "dedicated processing machine" immediately before the parentheses is the first WHAT adoption phrase. Further, when the word is not an adoptable word + the word with key brackets + the predicate adoptive word, the predicate adoptive word is replaced with "," (a reading point). Example 02
4. Candidate original text "Hanazono Tool Co., Ltd. has released" Lead Cleaner, "which sucks chips from the lead wires of printed wiring boards with air. ”Is adopted as the adoption phrase,“ Hanazono tools, air suctioning chips of lead wires of printed wiring boards ”. Here, "o" just before the release is replaced with ",". However, the words that are not applicable words + the words with parentheses + the predicate are the adopted words, and the predicates are "establish", "install", "implement", "developing", "complete", "sign", At the time of “formulation”, the inside of the brackets is adopted with the brackets. Example 025. Candidate original text "The Material Center has established an" Overseas Consultation Office "within the center as a measure to support overseas procurement and overseas expansion, which are increasing in the domestic material industry such as casting and forging. ”Is adopted as the phrase“ establishment of an “overseas consultation room” within the center as a measure to support overseas procurement and overseas expansion, which is increasing in the domestic raw material industry such as the material center and casting / forging ”. Here, because the predicate adoption phrase is "install", the "overseas consultation room" is the first with parentheses.
Adopted as a WHAT adoption word. Note that words that are not applicable words + words with parentheses + modifiers + first WHAT
When hiring a phrase, the modifier inside the brackets with brackets,
The first WHAT adoption word is adopted together. Example 026. Candidate original text "Kuroto No. 3 power plant was certified as a" Good Design Facility "which was newly established by Tsutsumi in 1994." Has been certified as a "Good Design Facility".

Incidentally, in the case of adoptable words / phrases including numerical, year, month, day, or yen + predicate adoptive words, they are rejected and the process goes back (S74). Example 027. Candidate original text "NEC Home Electronics will release the 9.5-inch LCD TV" NEXTV "with the largest screen size in the industry standard on October 1st. "
"The NEC Home Electronics launches a 9.5-inch LCD TV with the largest screen size in the industry standard."
And

When the first-stage WHAT / WHERE discovery processing is completed, the first condition of the WRITE condition judging section 17
The process goes to the WRITE condition determination process in the stage WHAT / WHERE discovery section (S75).

Here, the WRITE condition determination processing for the output of the first stage WHAT / WHERE discovery section 15 in the WRITE condition determination processing section 17 will be described with reference to FIG.

First, it is determined whether or not the word just before the first WHAT adopted word in the candidate original text is the WHO adopted word (S12).
1). In the case of YES, it is judged whether the title candidate sentence is within 40 characters (S122), and when it is within 40 characters, it is adopted as a title sentence and the candidate original sentence discovering unit 11 is processed to process the next sentence. (S123). Example 028. Candidate original text "NEC has developed its own inorganic resist formation method and succeeded in forming the world's smallest pattern of 5 nm by scanning electron beam exposure."
Developed original inorganic resist formation method. " Step S
122 when the candidate title sentence exceeds 40 characters, 4
Go to the 0 character compression unit 18. If it is determined in step S121 that the word immediately before the first WHAT adopted word is not the WHO adopted word in the candidate original sentence, the second stage WHAT / WHERE finding unit 1
Go to 6 (S125).

Step S1 of the WRITE condition judging section 17
When there is no WHO adopted word immediately before the first WHAT adopted word in the judgment of 21, the second step WHAT • WHERE finding processing is performed by the second step WHAT • WHERE finding section 16. The processing operation of the second stage WHAT / WHERE discovery unit 16 will be described with reference to FIG.

Second stage WHAT / WHERE discovery unit 17
Processing of the first WHAT adoption phrase in the candidate original text,
Go back to the beginning of the sentence and perform the following actions to perform the second stage WH
Discover AT ・ WHERE and format it for title sentences. Here, the word / phrase adopted by the second stage WHAT / WHERE discovery unit is referred to as the second WHAT / word adopted word / phrase. Also,
2nd WHAT / WHERE 2nd WH
It is called AT Adjective. This second WHAT Adjective is
“Oh”, “Ohto”, “Ototo”, “Ha”, “By”,
“Is”, “in”, “even”, “only”, “can”, “can”, “to”, “by”, “in”, “about”, “mo”, “ga” , "To", "to", and "as".

First, the candidate original sentence is traced back from the first WHAT adopted word, and if there is a second WHAT adopted word, the adoptable words before and after that are adopted. Alternatively, the WRITE condition determination unit 1
From 7 on, if there is a second WHAT adopted word, the applicable words before and after that word are adopted. If there is one Kanji character or Hiragana character immediately after, adopt up to the next applicable phrase or reading point. When hiring a phrase that can be adopted, and if it is followed by hiragana, a reading point is added. If the last word is only numbers,
It is not adopted, and the second WHAT adopted word is replaced with a reading point (above S81). Example 029. Candidate original text "Health food manufacturer brown rice yeast has entered the food and retail sector in Tokyo."
Is the hiring phrase, "Brown Rice Yeast, Entering the Food and Retail Field in Tokyo". Here, “de” is the second WHAT adopted word and “Tokyo de” is immediately before the first WHAT adopted word “restaurant / retail field”, so that it is adopted as the second WHAT adopted word. Example 030. Candidate original text "France Bet consolidates and abolishes logistics base warehouses to reduce transportation costs and enhance user services." Adopts the phrase "France Bet, logistics bases to reduce transportation costs and enhance user services. The warehouse is consolidated. here,
“To” of “to strengthen” is the second WHAT adopted word, and it is adopted including the words before and after. Example 031. Candidate original text "Sequent, Computers Japan corresponds to Microsoft's Daytona" WinService
The "r" 520 and 3500 will be shipped within the year. ”, The adopted phrase“ Sequent, Computers Japan,
It will be compatible with Microsoft's Daytona and will be shipped within the year. " Numerical words are not adopted as a reading point of "" of "3500 wo".

Next, in step S82, the second W
When there is a modifier immediately before the HAT adopted phrase, the applicable phrase immediately before that is also adopted. If there is a qualifier immediately before the adopted phrase, it is adopted while going backward. However,
If the immediately preceding adoptable phrase contains a number, it will be rejected. If there are modifiers “no” and “from” immediately after the second WHAT adopted word, the scoring word immediately after that is also adopted. If there is one Kanji character or Hiragana character immediately after, adopt up to the next applicable phrase or reading point. When hiragana is used immediately after the adoption of an applicable phrase, a reading mark is added. Example 032. Candidate original text "France Bed will consolidate the logistics base warehouse to reduce transportation costs and enhance user services." Adopted the phrase "France Bed, logistics base to reduce transportation costs and enhance user services. The warehouse is consolidated. Here, "no" of "user service" is a modifier immediately before the second WHAT adoption word, and "user service" immediately before that is adopted together with the second WHAT adoption word. Example 033. Candidate original text "Mori Seiki has developed a dedicated processing machine" SFL-10 / 600 "that finishes the surface of aluminum rolls used in OA equipment with high precision with a surface roughness of 0.2S only by cutting. "Mori Seiki has developed a special processing machine that finishes the surface of aluminum rolls used in OA equipment with high precision only by cutting." Here, the "face ancestry degree 0.2 S" immediately before the modifier "no" immediately before the second WHAT adopted word is deleted because it includes a numeral. Example 034. Sakae shopping center of the candidate original "discount stores, we developed a vending machine" robot vending machine "that can sell beer to leave the case, to sell to the whole country of the DS for the prospect of the 1995 summer. "Sakae Shopping Center develops vending machine that can sell beer as it is." “Sell as is” is the second word adopted.

Then, the process proceeds to step S83. Here, adoptable words + "XXX" (words with brackets) + 2W
In the case of HAT adoptive words, the adopted word before the brackets is adopted. Example 035. Candidate original text "The mail order company Nikko has divided the existing main catalog" Seasoning "into three product categories and distributed them to each customer. Is adopted as "Hikari, existing main catalog is divided into three product categories and distributed to customers." Here, the "main catalog" before the brackets before the second WHAT adopted word "o" is adopted as the second WHAT adopted word. Also, "no", punctuation, second WHAT Adjective + Keyword phrase + Second WH
When using the AT hieroglyph, the inside of the parenthesis is adopted with the parenthesis. Otherwise, the second WHAT adoption word is replaced with a reading point. Example 036. Candidate original text "Matsushita Electric Co., Ltd. will release" Electronic Meeting Unit "for video conferencing system that can be connected to a personal computer from desktop meetings to multi-party meetings from November 25th. "
Is adopted as the phrase "Matsushita Electric releases" electronic meeting unit for video conferencing system "." Here, since the character immediately before the parentheses is "no", "for the video conference system" including the inside of the parentheses is the second WHAT adoption phrase. However, the words that are not applicable words + the words with parentheses + the second WHAT adoption words are the predicate adoption words, "establishment", "installation", "implementation", and "development"
When, the inside of the key bracket is adopted with the key bracket. Example 03
7. Candidate original text "The Ministry of Posts and Telecommunications has established a policy of establishing a secretariat and a subcommittee in the" Information and Communication Infrastructure Development Promotion Liaison Conference "which will be established on the 12th as a full-scale promotion organization for multimedia promotion on the 14th. Is adopted as the hiring phrase "Ministry of Posts and Telecommunications, launched on the 14th, and the" Information and Communication Infrastructure Development Promotion Liaison Conference "has a secretariat and a subcommittee." In addition, when adoptable phrase + phrase with parentheses + modifier + second WHAT adoptable phrase, adoptable phrase before the bracket is adopted. Note that words that are not applicable words + words with brackets + modifiers + 2nd W
In the case of the HAT adoption word, the key parentheses are removed and adopted (above step S83).

Upon completion of step S83, the WRI
Go to the TE condition determination unit 17.

Now, with reference to FIG. 10, the WRITE condition judgment processing for the output of the second stage WHAT / WHERE discovery section 16 in the WRITE condition judgment processing section 17 will be described.

First, it is judged whether the word just before the second WHAT adopted word in the candidate original sentence is the WHO adopted word, or whether there is no second WHAT adopted word even if it goes back from the last adopted word (S1).
31). In the case of YES, it is determined whether the title candidate sentence is within 40 characters (S132). If it is within 40 characters, it is adopted as a title sentence, and the candidate original sentence finding unit 11 is processed to process the next sentence. (S133). Example 038.
Candidate original text "Life insurance industry is aggressively investing in stocks in small and medium-sized enterprises and local companies." The title sentence is "Active insurance investment in life insurance industry, small and medium-sized enterprises and companies."
Example 039. Candidate original text "Taisei Construction promotes and develops facility management business as a whole company." The title sentence is "Taisei Construction promotes facility management business." When the title candidate sentence exceeds 40 characters in step S132, the process proceeds to the 40-character compression unit 18 (step S134). Further, in step S131, the word just before the second WHAT adopted word in the candidate original text is not the WHO adopted word or is traced from the last adopted word, but the second WHAT
If there is an admission word, it is determined in step S135 whether or not the title candidate sentence exceeds 40 characters, and if it exceeds, the process proceeds to the 40-character compression unit 18 (S136).
If the title candidate sentence does not exceed 40 characters, the process goes to the second stage WHAT / WHERE discovering unit 16 again (S137).

When the title candidate sentence exceeds 40 characters, the WRITE condition determining unit 17 reformulates the sentence so that it is within 40 characters by the 40-character compressing unit. The operation will be described below with reference to FIGS. 11 and 12.

Here, "a series of adopted words" used in the following 40-character compression process means an adopted word adopted by a modifier adopted word (for example, "having a customs duty function and a bonded warehouse function"). (Adopted phrase) is included.

FIG. 11 shows the operation of the 40-character compression processing for the outputs of the predicate finding unit 14 and the first stage WHAT / WHERE finding unit 15.

First, when the next word is present in the title candidate sentence, it is converted into a reading point. “About”, “as”, “do”, “sequel”, “do”,
"Etc.", "etc.", "is", "only in",
“Can”, “Can”, “By”, “In”, “About”. It is determined whether or not this reading point substitution is within 40 characters, and when it is within 40 characters, it is adopted as a title sentence, and the candidate original sentence finding unit 11 is processed to process the next sentence (S141).

When the number of characters is less than 40 characters, the adopted word of "no" is deleted together with "no" in the title candidate sentence, and the sentence is packed. In this process, it is determined whether the number of characters is within 40 characters, and if the number of characters is within 40, it is adopted as a title sentence (S142, S143). As a result, when there are two "no" s, the adopted word of the previous "no" is deleted. 40 also in step S142
If it is not within the characters, go to step S143,
The title candidate sentence is added with “(hand correction)” and output as a title sentence. Example 040. Candidate original text "Furukawa Kikinzoku develops 3 large wheel loaders with bucket capacities of 2.5, 3.0 and 3.8 cubic meters and will be released from the 16th.""Develop a model".

Next, FIG. 12 shows the second stage WHAT / WHE.
The operation of the 40-character compression process for the output of the RE finding unit 16 is shown.

First, in the title candidate sentence, the adopted phrase +
When there is a connection by "no" such as "no" + adopted phrase + "no" ..., the adopted phrase by "no" at the beginning is deleted together with "no" and the sentence is packed. As a result, if it is within 40 characters, it is adopted as a title sentence, and thereafter, it goes to the candidate original sentence finding unit 11 (S151). If it is not within 40 characters, go to the next step S152. Example 041. Candidate original text "NEC has doubled the brightness of the display screen to 1
Developed 2-inch TFT type liquid crystal display, 9
Samples will be shipped from February 5th. The title candidate sentence is "NEC, the brightness of the display screen is doubled from the conventional one.
Developed 2-inch TFT type LCD device
Since (45 characters) is over 40 characters, the process of step S151 deletes "12-inch size", and the title sentence "NEC, the liquid crystal of the TFT system in which the brightness of the display screen is doubled from the conventional one" is displayed. Development of display device "(36
Character).

Next, in step S152, when there is a series of adopted words + “for” in the title candidate sentence, this is deleted. In addition, when a series of adopted words + "for" + a series of adopted words + "to", all are deleted. If it is within 40 characters, it will be adopted as the title sentence (S
158). If it is not within 40 characters, go to the next step S153. Example 042. Candidate original text "Matsushita Electric and Matsushita Electronics announced on the 12th that they will establish a joint venture with a local company to manufacture and sell semiconductors for household VTRs in Shanghai, China." Denshi established a joint venture with a local company to manufacture and sell semiconductors for home VTRs in Shanghai, China (47 characters), and deleted "for home VTRs" and gave the title sentence "Matsushita Electric and Matsushita Electronics". Established a semiconductor manufacturing and sales company in a joint venture with a local company in Shanghai, China "(39 characters). Example 0
43. Candidate original text "Large blast furnace companies such as Nippon Steel and NKK will increase the export price of steel products in Southeast Asia mainly for Southeast Asia." The title candidate sentence is "Nippon Steel, NKK, etc. Raising the export price of steel products mainly to Southeast Asia ”(48
Character) and deleted "Mainly for Southeast Asia",
Title sentence "Nippon Steel, NKK and other major blast furnace companies raise the export price of steel products loaded in the third quarter of 1994" (37 characters)
And

Next, in step S153, if the title candidate sentence contains a series of adopted words + “such as” + adopted words, all are deleted. Also, when there is a series of adopted words + "etc." + adopted words, a series of adopted words + "etc."
Delete up to. If the number of characters is within 40 characters by this processing, it is adopted as a title sentence (S158). If it is not within 40 characters, go to the next step S154.
Example 044. Candidate original "NTT is 12 days, announced the outline of preparation company 9 companies towards the commercialization of the personal handyphone PHS, carry out the planning and investigation to establish the end of October to the prospect."
Is the title of candidate statement, is "NTT, towards the commercialization of PHS, founded in prospect the end of October, announced the outline of preparation company nine companies to do the planning and investigation" (49 characters), "Planning and
"Do an investigation, etc." is deleted, and the title sentence "NTT, PH
Toward the commercialization of S, founded in prospect the end of October, announced the outline of preparation company nine companies "and (39 characters). Example 045. Candidate original text "Mitsubishi Motors Co., Ltd. announced on the 20th that it launched a specially equipped vehicle" Shario Road "equipped with RV related equipment such as a front bumper with a grill guard on the" Shario. " The candidate sentence for the title is "Mitsubishi Motors," Shario "equipped with RV related equipment such as front bumper with grill guard, specially equipped vehicle is on sale" (46 characters), and "Front bumper with grill guard etc." was deleted. "Mitsubishi Motors attaches RV-related equipment to" Shario "and sells specially-equipped vehicles" (28 characters).

Next, in step S154, when there are a plurality of adopted words + reading points immediately before "(numerical) country", "(numerical) company", and "no (numerical) company" in the title candidate sentence, the head is found. Add "etc." to the adopted word and delete the others. If the number of characters is within 40 characters by this processing, it is adopted as a title sentence (S158). If it is not within 40 characters, go to the next step S155. Example 046. Candidate original text "Pacific Exchange Center will organize an alumni association of returnees from China, Russia, and Vietnam in 1994." The title candidate sentence is "Pacific Exchange Center, 9
"Alumni association organized by returnees from three countries, China, Russia, and Vietnam during the 4th fiscal year" (45 characters), replacing "China, Russia, Vietnam" with "China, etc." Organized an alumni association of returnees from three countries such as China during 1994 "(38 letters).

Next, in step S155, when the title candidate sentence includes a series of adopted words + “prepared”, all of them are deleted. If the number of characters is within 40 characters by this processing, it is adopted as a title sentence (S158). If it is not within 40 characters, go to the next step S156. Example 047. Candidate original text "Nippon has opened and started operation of Inland Depot, an inland logistics base with a customs function and a bonded warehouse function in the Ritto branch in Ritto-cho, Shiga Prefecture.""Inland Depot, an inland logistics base with customs and bonded warehouse functions, was opened in the Ritto branch in Ritto-cho, Shiga Prefecture" (50 characters), and "with both customs and bonded warehouse functions" Deleted and added the title sentence "Nippon Express, Inland Depot, which is an inland logistics base, in the Ritto Branch in Ritto Town, Shiga Prefecture" (35 characters).

Next, in step S156, when the candidate original sentence includes "term settlement", "WHO adopted word + reading point +" (number) year "+" (number) month "term settlement forecast" +
The phrase "first point + first WHAT adopted word + predicate adopted word" is output. If this process results in less than 40 characters, it is adopted as a title sentence. If it is not within 40 characters, the process goes to the next step S157. This is an example of converting a standard sentence such as a financial statement announcement into a standard title sentence. Example 048. Candidate original text "Matsuya forecasted ordinary income of 350 million yen to 580 million yen on the 19th February 1995 financial forecast, and net income also increased from 150 million yen to 207 million yen. The title candidate sentence of "Announced upward revision to 10 million yen each," was "Matsuya, initial forecast ordinary income of 350 million yen, 580 million yen, and net income of 150 million yen. From 207
It is announced that 10 million will be revised upwards, and the title sentence is,
"Matsuya forecasts the financial results for the fiscal year ending February 1995. It will be announced when the upward revision is made."

Finally, in step S157, the final second
By adopting the staged WHAT adoption word, it is over 40 characters, so the word is not adopted. Example. Candidate original text "Shin-Oh began selling" Apogee Spark Compilers ", a compiler software for UNIX machines by Apogee, USA as part of strengthening the network division. The candidate sentence for the title is "Shin Oh, we are beginning to sell compiler software for UNIX machines by Apogee, USA as part of strengthening the network division." Since it exceeds 40 characters, the second stage WHAT adoption phrase "Network division" As a part of strengthening, ”
The title sentence is "Shin Ou, UNIX of Apogee, USA
Start selling the compiler software for machines. "

[0087]

According to the present invention, by performing such processing,
Since the title sentence can be created automatically, there is an effect that labor can be saved. Further, since the processing can be performed in a short time, there is an effect that the article database can be provided to the user in a short time.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of a title sentence creation device according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating an operation of a candidate original text finding unit.

FIG. 3 is a flowchart illustrating an operation of a preprocessing unit.

FIG. 4 is a flowchart illustrating an operation of a WHO discovery unit.

FIG. 5 is a flowchart illustrating an operation of a predicate finding unit.

FIG. 6 is a flowchart for explaining the operation of the first stage WHAT / WHERE discovery unit.

FIG. 7 is a flowchart for explaining the operation of the second stage WHAT / WHERE discovery unit.

FIG. 8 is a flowchart illustrating the operation of a WRITE condition determination unit.

FIG. 9 is a flowchart illustrating the operation of a WRITE condition determination unit.

FIG. 10 is a flowchart illustrating the operation of a WRITE condition determination unit.

FIG. 11 is a flowchart illustrating the operation of the 40-character compression unit.

FIG. 12 is a flowchart illustrating an operation of a 40-character compression unit.

[Explanation of symbols]

 11 Candidate Original Text Finding Section 12 Pre-Processing Section 13 WHO Finding Section 14 Predicate Finding Section 15 First Stage WHAT / WHERE Discovering Section 16 Second Stage WHAT / WHERE Discovering Section 17 WRITE Condition Determining Section 18 40 Character Compressing Section

Claims (5)

[Claims] 1. A candidate sentence extracting unit (11) for extracting a candidate original sentence suitable for creating a title sentence from the beginning of a main body of an original article, sentence by sentence, and a phrase representing a subject necessary for a title sentence from the extracted candidate original sentence, and Subject and predicate extraction means (13, 14) for finding and extracting a word representing a predicate, judgment means (17) for judging whether or not the extracted word is suitable as a title sentence, and this judgment means for the title sentence A compression means (1) for compressing a sentence determined to match as the number of characters required for a title sentence.
8) and the determination means includes means for determining whether or not there is a phrase that can be adopted as a title sentence between the extracted subject and predicate. 2. A first object indicating an object or a complement of a sentence existing between the subject and the predicate when the determining unit determines that there is a phrase that can be adopted as a title sentence between the subject and the predicate. A first WHAT finding means (15) for searching whether or not there is a WHAT adopted word or phrase, wherein the determining means is between the first WHAT adopted word and the subject found by the first WHAT finding means. The title sentence creating apparatus according to claim 1, further comprising means for determining whether or not there is a phrase that can be adopted as a title sentence. 3. The sentence between the subject and the first WHAT adopted word, when the judgment unit judges that there is a word that can be adopted as a title sentence between the first WHAT adopted word and the subject. 3. The title sentence creating device according to claim 2, further comprising second WHAT finding means (16) for searching whether or not there is a second WHAT adopted word or phrase that represents the object or complement of the above. 4. The subject and predicate finding means finds a subject adoptive word of a vocabulary that follows a word representing a subject in a candidate original sentence, extracts the subject, and adds the vocabulary before a word representing a predicate in the candidate original sentence. 4. The title sentence creating device according to claim 1, further comprising means for finding a predicate adopting word and extracting the predicate. 5. The title sentence creating apparatus according to claim 1, further comprising a pre-processing unit for deleting unnecessary words or symbols in the title sentence from the adopted candidate original sentence, before the subject and predicate finding unit.