JPH07262217A - Text retrieval device - Google Patents

Abstract

PURPOSE:To provide a text retrieval device which precisely gives priority by giving priority to retrieval results by keyword and cooccurrence words. CONSTITUTION:A keyword designation part 1 receives the keywords inputted from users and respectively transmits the keywords to a text retrieval part 2 and a cooccurrence word retrieval part 4. The text retrieval part 2 retrieves a text data base 3 by the obtained keywords. On the other hand, the cooccurrence word retrieval part 4 retrieves a cooccurrence word data base 5 by the obtained keywords and obtains information on the cooccurrence word. A priority calculation part 6 calculates the priority of a retrieval item retrieved in the text retrieval part 2 based on the keywords and information on the cooccurrence word obtained in the cooccurrence word retrieval part 4 and sorts and displays the retrieval results in the order of priority by a text display part 7.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a text search device for searching a text by a keyword, and in particular, an item searched by using keyword co-occurrence information (information of a word strongly related to the keyword). The present invention relates to a text search device having a function of prioritizing.

[0002]

2. Description of the Related Art In a text search using a keyword, the number of items to be searched (hit) increases as the size of the text increases, which makes it difficult to check. Also, the relationship between keywords and hit items varies from important to lightly touched. Therefore, by prioritizing the retrieved items and checking from the important items to the required level, the necessary information can be efficiently obtained.

In order to judge whether or not the retrieved item is important, the amount of description regarding the keyword in the item is important. The description regarding the keyword may be a case where the keyword is directly described or a case where the keyword is described indirectly, and it is considered that the number of keywords reflects the amount of the direct description regarding the keyword to some extent. Therefore, in some conventional devices, prioritization is performed according to the number of keywords included in the text, and the higher the number of times a keyword appears, the higher the priority.

However, when the number of keywords contained in the text is large enough, the importance can be judged to some extent only by the amount of the direct description, but when the number of keywords is small, it cannot be judged or the user's request. There are also cases where the judgment is far from being made. Therefore, it is necessary to consider not only the amount of keywords directly described but also the amount of indirect descriptions. That is, in the conventional method, since the search results are prioritized by the number of keywords, only the amount of directly describing the keywords can be reflected. When the number of keywords is small, the prioritization can be performed accurately. Can not. In particular, in the case of dictionary-type text in which headings and their explanations are repeated, the amount of sentences per item is small, and the number of keywords contained in them is also small, which is a problem.

As another conventional apparatus, for example, Japanese Patent Laid-Open No. Hei.
There is a document search device described in Japanese Patent Publication No. 281565. In this device, the text is divided into parts such as the introductory text and the body, and the priority is calculated by weighting the part in which the keyword appears. According to this device, the text in which the keyword appears in the important part of the text can be preferentially presented to the user. Therefore, for example, even in the case of dictionary type text, the text in which the keyword is present in the headline can be presented to the user in preference to the text in which the keyword is present in the main body. However, even in this device, the amount of indirect description is not taken into consideration. So, for example,
Even if there is an indirect expression in the headline, if there is no direct description, the priority becomes low, and there is a problem that the desired result may not be obtained.

[0006]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned circumstances, and reflects the amount of indirect description by the number of co-occurrence words, and a keyword and its co-occurrence word are used as search results. An object of the present invention is to provide a text search device that can perform prioritization with higher accuracy by performing prioritization even when the number of keywords is small.

[0007]

SUMMARY OF THE INVENTION The present invention is, in a text search device for searching text by keyword, a keyword designating section for accepting input of a keyword, a text database storing text searchable by the keyword, A text search unit that searches the text database with the keyword accepted by the keyword designation unit, a co-occurrence word database that stores information on a co-occurrence word that has a strong relationship with a certain word, and a keyword accepted by the keyword designation unit A co-occurrence word search unit that searches the co-occurrence word database with a keyword, a keyword accepted by the keyword designation unit, and a search in the text search unit based on the co-occurrence word information searched by the keyword in the co-occurrence word search unit. Has a priority calculation part that calculates the priority of the result And it is characterized in Rukoto.

[0008]

According to the present invention, the text search section searches the text database with the keyword specified by the user in the keyword specification section, and the co-occurrence word search section refers to the co-occurrence word database to obtain the co-occurrence word information. To get Then, the priority calculation unit prioritizes the search results by the text search unit based on the keywords and the information of the co-occurrence words obtained by the co-occurrence word search unit. In the prioritization using only keywords, only direct description about keywords is reflected. However, in the text search device of the present invention, prioritization based on keywords and their co-occurrence words is performed to directly and indirectly describe keywords. It is possible to reflect both of the amounts of, and more accurate prioritization can be performed.

[0009]

1 is a block diagram showing an embodiment of a text search device according to the present invention. In the figure, 1 is a keyword designation part, 2 is a text search part, 3 is a text database,
Reference numeral 4 is a co-occurrence word search unit, 5 is a co-occurrence word database, 6 is a priority calculation unit, and 7 is a text display unit.

The keyword designating section 1 receives an input of a keyword for a search from the user and transmits it to the text searching section 2 and the co-occurrence word searching section 4. The text search unit 2 searches the text database 3 when the keyword is obtained,
The search result is transmitted to the priority calculation unit 6. Text is stored in the text database 3, and a description about the word can be obtained by searching for an arbitrary word.

The co-occurrence word search unit 4 searches the co-occurrence word database 5 when the keyword is obtained, and transmits the co-occurrence word information to the priority calculation unit 6. The co-occurrence word database 5 stores information on words (co-occurrence words) having a strong relationship with the head word as a head word. For example, the entry word is "A
Information such as headword, co-occurrence word AI, ... artificial intelligence, computer, expert system OA, ... word processor, personal computer, electronic file is stored.

When the priority calculation unit 6 obtains the search result by the text search unit 2 and the information of the co-occurrence word from the co-occurrence word search unit 4, the priority calculation unit 6 calculates the priority of the searched item and stores the information. It is transmitted to the text display unit 7. When the text display unit 7 obtains the priority information of the search items, it sorts the search items in order of priority and displays the information to the user. In addition to displaying the search items on the text display unit 7, the results can be stored in a file or input to another device.

Next, the operation of one embodiment of the text search apparatus of the present invention will be described based on a concrete example. FIG. 2 is an explanatory diagram of an example of the operation in one embodiment of the text search device of the present invention, and FIG. 3 is an explanatory diagram of an example of text data stored in the text database. In the explanation below,
An example of input and output of each component when the user specifies "AI" as a keyword will be described.

When the keyword designation unit 1 obtains the keyword "AI" input by the user, the keyword "A"
I ”is transmitted to the text search unit 2 and the co-occurrence word search unit 4, respectively. When the keyword “AI” is obtained, the text search unit 2 searches the text database 3. Now, it is assumed that the text data shown in FIG. 3 is stored in the text database 3. As a result of the search using the keyword “AI”, the words shown in double lines in FIG. 3 match, and the search result is “law expert system ...”,
Each text of "program trading ...", "AI ...", and "pivot method ..." is obtained. These search results are transmitted to the priority calculation unit 6.

On the other hand, when the keyword "AI" is obtained, the co-occurrence word search unit 4 searches the co-occurrence word database 5,
Obtain information on the co-occurrence word of "AI". Now, it is assumed that "artificial intelligence, robot, expert system, machine translation, inference, learning" is obtained as the information of the co-occurrence word of "AI".
The obtained co-occurrence word information is transmitted to the priority calculation unit 6.

In the priority calculation unit 6, as a result of searching with the keyword "AI" in the text search unit 2, "law expert system ...", "Program trading ...",
"AI ...", "pivot method ...", and co-occurrence word information "artificial intelligence, robot, expert system, machine translation," obtained as a result of searching with the keyword "AI" in the co-occurrence word search unit 4. When "inference and learning" is obtained, the priority of the search item is calculated and the result is transmitted to the text display unit 7. When the text display unit 7 obtains the search results with priorities, the search items are sorted in order of priority and the information is displayed to the user.

The priority calculation in the priority calculation unit 6 is as follows.
For example, the following procedure can be performed. First, each search item is divided into the following three groups. 1. Include a keyword in the item name. 2. Item names include co-occurrence words. 3. other than that. Each group is ordered in this order. Then, each item is ranked by the number of keywords and co-occurrence words included in the item in each group. At this time, the number of keywords has priority over the number of co-occurrence words.

In this specific example, a co-occurrence word is present in each text of the search result together with the keyword. In FIG. 3, the co-occurrence words are shown with an underline. First, as a search result that includes a keyword in the item name, "AI
There is a text "...". Search results that include a co-occurrence word in the item name include "Legal expert system.
There is a text "...". Other than that, there are texts such as "program trading ..." and "pivot method ...". In addition, "Program trading ..."
There is one keyword "AI" in the text, and one text "AI" and one co-occurrence word "machine translation" exist in the text "Pivot method ...". From these things, the priority is "AI ...",
"Law expert system ...", "Pivot method ...", "Program trading ..."

As another example of the priority calculation, the priority can be calculated by the following score calculation. 1. If the item name contains a keyword, add +100 to the score. 2. If the item name contains a co-occurrence word, add +10 to the score. 3. (Number of keywords) + (Number of co-occurrence words) x 0.7
Is added to the score. The score of each text obtained as a result of such score calculation is used as the priority. The text display unit 7 sorts the search items based on this score and displays the information to the user.

In this specific example, the number of keywords and co-occurrence words of each searched text, the score, and the priority order are as shown in FIG. That is, in the case of the text "Legal expert system ...", the item name includes a co-occurrence word, so +10, one keyword, and two co-occurrence words, so 1
+ 2 × 0.7 = 2.4, which is a total of 12.4. In the case of the text “Buy and sell programs ...”, since the item name does not include a keyword and a co-occurrence word, the number is one for each keyword. In the case of the text “AI ...”, the keyword is included in the item name, so +100, keyword 1
Since there are 5 co-occurrence words, 1 + 5 × 0.7 = 4.5, which is a total of 104.5. Text "Pivot method
・ ・ ”, One keyword and one co-occurrence word
1 + 1 × 0.7 = 1.7.

In this way, the score of each item is calculated, and the order of priority is given to the item having the highest score. As a result, information of prioritized (priority order, search items) ((2 legal expert system ...) (4 program trading ...)
(1 AI ...) (3 Pivot type ...)) is obtained. This information is transmitted to the text display unit 7, and the text display unit 7 sorts the search items in order of priority and displays them for the user.

Still another example of calculating the priority will be described.
For each text searched by the text search unit 2, first, a sentence including a keyword and a co-occurrence word in the text is set as a descriptive sentence. At this time, the sentence sandwiched between the descriptive sentences is also the descriptive sentence. If there are multiple statements sandwiched between descriptive sentences,
Only when the number of sentences is less than the set number, the sandwiched sentence is regarded as a descriptive sentence. When the title includes a keyword, all sentences included in the paragraph are descriptive sentences. In this way, the number of sentences determined to be description sentences is counted, and it is determined that the text having more description sentences has higher priority. In this way, the texts of the search results are given priority, sorted in order of priority and displayed on the text display unit 7.

It is assumed that the text "fuzzy ..." Stored in the text database shown in FIG. 3 is searched by the keyword "fuzzy". In addition, the co-occurrence word corresponding to the keyword “fuzzy” is defined as “ambiguity, AI”. At this time, the keyword "fuzzy", the co-occurrence word "ambiguity", "AI"
"Fuzzy is uncertain.", "The engineering field to which this uncertainty is applied is fuzzy engineering.", "Experience and intuition including ambiguity that cannot be expressed numerically are the subjects of research." It is judged that the three sentences of "I am doing." If one sentence sandwiched between the descriptive sentences is also a descriptive sentence, the third sentence that does not include a keyword or a co-occurrence word, "Professor Zaday of the University of California is the founder." Is also a descriptive sentence. However, in this example, since the title includes the keyword, all sentences are descriptive sentences. In this example, there are four descriptive sentences. Similarly, in other searched texts, the description sentences are counted and sorted in descending order of description sentences.
It will be displayed on the text display unit 7.

Next, the case where a keyword has a plurality of meanings will be described using a concrete example. In the explanation below,
It is assumed that "CD" is input as the keyword. Further, it is assumed that “compact disc, digital, laser, music” is stored in the co-occurrence word database 5 as a co-occurrence word corresponding to “CD”. On the "CD"
"Compact disc", "Negotiable certificates of deposit (Certif
"ICATION OF DEPOSIT", "Cash Dispenser", "Conventional Disarmament"
t) ”,“ cyclodextrin ”, etc., and many items will be hit by text search.

FIG. 5 is an explanatory diagram of the number, score, and rank of keywords and co-occurrence words in each search item searched by the keyword "CD". Although the text stored in the text database is not shown, the keyword "C
As a result of the search by "D", it is assumed that the search items as shown in the leftmost column of FIG. 5 are searched. Here, "semiconductor laser ...", "PCM music broadcasting ...", "CD-RO"
"M ..." and "LD ..." are "compact discs"
Is related to "Free interest rate products ..." are "negotiable deposits", "Banks ..." are "cash dispensers", "Geneva Conference on Disarmament ..." is "normal disarmament",
"Molecular capsules ..." is related to "cyclodextrin".

Assuming the number of keywords and the number of co-occurrence words as shown in FIG. 5, a method of calculating a score is adopted as the priority calculation method, and the score calculation is performed by the above calculation method. Then, priorities are given according to the calculated points.
Here, by prioritizing only the search items for which the co-occurrence word is used, only the four items related to the "compact disc" are prioritized. As a result, it is possible to obtain a search result in which the meaning of “CD” is limited to “compact disc”. In this way, by using the co-occurrence word, it is possible to limit the search for the meaning of a word having a plurality of meanings.

In the above description, it is assumed that the text to be searched is a dictionary type text that is a repetition of the item name and its description as shown in FIG. However, the present invention is not limited to this, and may be one large text. At this time,
It is applicable if the unit of items is a chapter, a section, or a paragraph. When the unit of item is a paragraph, the item name and title may not exist. In addition, a search can be performed on a document group in which some documents are collected. At this time, the title may be the item name and the document may be the one item. In this case as well, it can be applied even if there is no title.

[0028]

As is apparent from the above description, according to the present invention, not only when a keyword is directly described, but also when an indirect description is taken into consideration, a keyword and its co-occurrence word are used to generate a search result. Since the prioritization is performed, the prioritization can be performed more accurately than the conventional prioritization of only the keyword that considers only the direct description of the keyword. Especially in texts where the amount of text per item is small, the number of keywords is small,
The present invention is also effective for such text. Further, even when a keyword has a plurality of meanings, the use of a co-occurrence word has an effect that the meaning of the keyword can be limited to be searched.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a text search device according to the present invention.

FIG. 2 is an explanatory diagram of an example of an operation in one embodiment of the text search device of the present invention.

FIG. 3 is an explanatory diagram of an example of text data stored in a text database.

FIG. 4 is an explanatory diagram of the number of keywords and co-occurrence words, scores, and priorities in each search item searched by the keyword “AI”.

FIG. 5 is an explanatory diagram of the number of keywords and co-occurrence words, scores, and priorities in each search item searched by the keyword “CD”.

[Explanation of symbols]

1 ... Keyword designation part, 2 ... Text search part, 3 ... Text database, 4 ... Co-occurrence word search part, 5 ... Co-occurrence word database, 6 ... Priority calculation part, 7 ... Text display part.

Claims (1)

[Claims] 1. A text search device for searching for text by a keyword, a keyword designating section for accepting input of a keyword, and a text database storing text searchable by the keyword.
A text search unit that searches the text database with the keyword accepted by the keyword designation unit, a co-occurrence word database that stores information on a co-occurrence word that has a strong relationship with a certain word, and a keyword accepted by the keyword designation unit A co-occurrence word search unit that searches the co-occurrence word database with a keyword, a keyword accepted by the keyword designation unit, and a search in the text search unit based on the co-occurrence word information searched by the keyword in the co-occurrence word search unit. A text search device having a priority calculation unit for calculating the priority of a result.