JPH08314974A - Automatic key work extracting device and document retrieving device - Google Patents

Abstract

PURPOSE: To provide an automatic key word extracting device capable of extracting only a key word securely expressing the contents of a document by the human intervention of an absolute minimum at the time of automatically extracting a key word from the retrieving object document. CONSTITUTION: The automatic key word extracting system automatically extracts the key word from the retrieving object document through the use of a dictionary 12 or a thesaurus 13 describing the high-low order relation mutually between the candidate words for the key word. The device is provided with a key word candidate word segmenting means 14 segmenting the candidate words for the key word from the retrieving object document, a key word candidate word selection means 15 confirming a user's intention and then selecting this candidate word for the key word as the key word when the segmented candidate word for the key word is described on plural places in the thesaurus, and an input/output means 16 presenting information to the user and receiving a selection input from the user. When the segmented candidate word for the key word is described in the many places in the thesaurus and the candidate word involves plural meanings, which meaning the candidate word processes is asked to the user by way of the input/output means to store the candidate word as the key word with the meaning selected by the user.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document retrieval device for retrieving a desired document from documents stored in a database or documents stored in a storage device such as a general word processor or office computer, and this document retrieval. The present invention relates to a keyword automatic extraction device that is used in a device and that automatically extracts keywords that characterize each document, and in particular, enables a highly accurate document search.

[0002]

2. Description of the Related Art In recent years, with the start of mass distribution of digitized document information such as electronic mails, electronic catalogs, and electronic publications, a document search for searching only desired documents from those document information. There is growing interest in devices.

In such a document retrieval device, conventionally,
A method called a keyword search for searching a document using a keyword assigned to each document is generally used. In this keyword search, a keyword representing the content of the document is manually added to the stored document in advance, and a transposed file for the keyword is configured.
When a user inputs a desired keyword at the time of search, a document containing this keyword is searched at high speed by using the transposed file.

In this keyword search, a person assigns a keyword by examining the content of each document, so that the document having the content desired by the user can be searched with high accuracy. However, on the other hand, it has been pointed out that manual keyword addition work cannot keep up with the increase in the number of stored documents.

In order to solve such a problem, a device for automatically extracting a keyword from a document has been proposed (for example, Haruo Kimoto "keyword automatic extraction device",
JP-A-63-136224). As shown in FIG. 19, this conventional keyword automatic extraction device includes a document storage unit 191, which stores documents to be searched, a dictionary 192 which is referred to when extracting keywords, and a thesaurus in which upper-lower relationships between words are described. 193, and a keyword automatic extraction unit 194 for extracting the keyword of the document read from the document storage unit 191
And a keyword extraction result storage unit 195 that stores the extracted keywords in a transposed file format or the like so that they can be easily used in subsequent searches.

The keyword automatic extraction unit 194 of this apparatus is
First, the search target document is read from the document storage unit 191, and the dictionary 1
Using 92, the longest match method (the longest character string that matches a word in the dictionary is classified as a word) and the morphological analysis method (the character string is classified using part-of-speech information, connection information, etc.) , Divide this document into words. For example,
When a document (document number 20) as shown in FIG. 20 is divided into words, a word string as shown in FIG. 21 is obtained.

Next, the keyword automatic extraction unit 194 selects a keyword to be given as a keyword of this document from the word strings thus obtained. In the keyword search, it is preferable to extract only the words that accurately represent the contents of this document, so when selecting a keyword,
Prepare an unnecessary word dictionary and do not select words registered in the unnecessary word dictionary, or prepare a keyword dictionary and select only words registered in the keyword dictionary. Alternatively, it is possible to take a measure such as selecting a word that is judged to be important by performing frequency calculation or the like. For example, by removing an auxiliary word such as a particle as an unnecessary word from the word string in FIG.
Only the words shown in FIG. 22 are selected.

Finally, the keyword selected by the keyword automatic extraction unit 194 is stored in the keyword extraction result storage unit 195 as the keyword of this document. FIG. 23 is a diagram showing the contents of the keyword extraction result storage unit 195. Figure 2
For each word selected in 2, the document number 20 containing that word is recorded.

At the time of search, a keyword input from the user is accepted, and a document including this keyword is searched from the keyword extraction result storage unit 195. For example, if the user's input is "sunflower", the keyword extraction result storage unit
195 is searched, and the document with the document number 20 including “sunflower” as a keyword is obtained from the table of FIG. 23 as the search result.

A method called full-text search has also been proposed as a means of solving the keyword assignment relying on manual labor from another viewpoint. In this full-text search, a character string provided by the user is used as a search condition, and this search condition is compared with all the character strings forming the search target document to obtain a document satisfying the search condition. Therefore, in this method, it is not necessary to add a keyword to the document in advance.

[0011]

However, in the conventional keyword automatic extraction device, since all the keywords are automatically extracted from the document to be searched, it cannot be said that the contents of this document are properly represented. Sometimes things are captured as keywords. Moreover, the number of keywords to be added is considerably larger than that of the normal manual keyword addition. As a result, in the case of performing a keyword search for a document to which a keyword has been added by the automatic keyword extraction device, there is a problem that the accuracy of the search deteriorates (both the search omission and the search dust increase).

Explaining in a concrete example, when the keywords of the document shown in FIG. 20 are automatically extracted, the contents of the document such as “month”, “day”, and “Ryogoku” are expressed as shown in FIG. Even an extra word that cannot be thought of is registered as a keyword, so the search accuracy when searching a document for a keyword deteriorates. For example, if the user wants to search for a document related to the place name "Ryogoku" and searches with the keyword "Ryogoku", the document with the document number 20 is erroneously searched. Further, when a user wants to search for a document about the moon "Month", if the search is performed using the keyword "month", the document with the document number 20 will be erroneously searched. Further, when the user wants to search for a document related to a famous movie, if the search is performed using the keyword "masterpiece", the document "document number 20" is added with the keyword "masterpiece". The document with the document number 20 is retrieved. As described above, when the keywords are automatically added, as a result, the number of documents (search dust) that are erroneously searched increases, and the search accuracy deteriorates.

Further, in the conventional full-text search, all documents including the character string input by the user as a search condition are output as the search result, which is suitable for an exhaustive search, but the search There was a problem that there was a lot of garbage and the search accuracy was poor.

The present invention solves these conventional problems. Only a keyword that accurately expresses the content of a document is obtained by automatically interposing a minimum amount of human hands when the keyword is automatically extracted from the document to be searched. An automatic keyword extraction device capable of extracting a keyword is provided, and when searching for a search target document in which a keyword is automatically extracted, a minimum of human intervention is performed to perform a highly accurate and efficient keyword search. It is an object of the present invention to provide a document search device that can perform such a search.

[0015]

Therefore, in the present invention, in a keyword automatic extraction device for automatically extracting a keyword from a document to be searched using a dictionary or a thesaurus in which upper-lower relationships between keyword candidate words are described. ,
When the keyword candidate word extraction means for extracting a keyword candidate word from the document to be searched and the extracted keyword candidate word are described in multiple places in the thesaurus, this keyword candidate word is checked after confirming the user's intention. A keyword candidate word selecting means for selecting as a keyword and an input / output means for presenting information to the user and accepting selection input from the user are provided.

Further, among the keyword candidate words described in the thesaurus, the keyword candidate words which require attention in the selection of the keyword are marked in advance and the keyword candidate word cut out by the keyword candidate word cutting means is the caution word. When the above applies, the keyword selecting means is configured to select the keyword candidate word as a keyword after confirming the user's intention.

Further, a keyword is automatically extracted from a document to be searched using a dictionary or a thesaurus in which upper-lower relationships between keyword candidate words are described, and the extraction result is stored in the keyword extraction result storage means and used. In a document retrieval device that retrieves a document having this keyword from the keyword extraction result storage means by using the retrieval character string input by the user as a keyword, when the retrieval character string is described in a plurality of locations in the thesaurus, the user's intention is A keyword search means for confirming and searching a document having a keyword intended by the user from the keyword extraction result storage means, and an input / output means for presenting information to the user and accepting a selection input from the user are provided. There is.

As means for automatically extracting a keyword from the search target document, a keyword candidate word cutting means for cutting out a keyword candidate word from the search target document and the cut out keyword candidate words are described at a plurality of locations in the thesaurus. At this time, a keyword candidate word selecting means for selecting this keyword candidate word as a keyword after confirming the user's intention is provided.

Further, a full-text search means for searching the search target document including the search character string by character string collation is provided so that the search by the keyword search means or the full-text search means can be selected from the input / output means.

Further, a synonym dictionary describing the correspondence between word groups having a synonym relationship and a keyword candidate word of the thesaurus, and a search character for converting a search character string input by the user using the synonym dictionary Column conversion means is provided, and the keyword search means is configured to give the keyword candidate words of the thesaurus converted by the search character string conversion means as a search character string.

Further, the search character string conversion means converts the search character string input by the user into a word group of synonyms using a synonym dictionary, and the full text search means selects from among the word groups. The character string selected by the user is given as the search character string.

[0022]

In the keyword automatic extraction device of the present invention, when the cut-out keyword candidate word appears in a plurality of positions in the thesaurus, that is, when the candidate word has a plurality of meanings, it is input through the input / output means. The user is asked what meaning it has, and the candidate word is stored as a keyword having the meaning selected by the user. In this way, since the user is given the opportunity to select the correct meaning for the ambiguous candidate word, highly accurate keyword extraction becomes possible.

Also, regarding keyword candidate words that are likely to cause an error in keyword extraction, a caution word is displayed on the thesaurus, and even when this caution word is cut out as a candidate word, the user must select it. Wait, it will be stored as a keyword.

Further, in the document search device of the present invention, when the search character string input by the user is described in a plurality of locations in the thesaurus, that is, when the search character string has a plurality of meanings, input / output is performed. The user is asked what meaning the search character string means through the means, and a document having a keyword intended by the user is searched. Therefore, the search according to the user's intention can be advanced with high accuracy, and the search as a whole can be efficiently performed.

Further, by constructing the keyword extracting means in this document retrieval device by the keyword automatic extraction device of the present invention, it is possible to perform a highly accurate document retrieval with less omission of retrieval and retrieval dust.

Further, in the document search apparatus equipped with the full-text search means as well, the keyword search having a relatively high accuracy and the full-text search having a relatively high degree of completeness are properly used, or the target documents collected by the full-text search are used as keywords. It is possible to perform a complementary search, such as narrowing down the search, and a search that meets the needs of the user can be performed.

In addition, in the document search device provided with the synonym dictionary, as keywords of each document, in addition to the extracted keyword candidate words, the superordinate words on the thesaurus of the candidate words are also stored. On the other hand, at the time of keyword search, the synonym dictionary is searched for an upper word on the thesaurus of the input search character string, and the target document is searched using this upper word as a keyword. In this case, since it is possible to obtain a desired document only by searching for the synonyms, which are subordinate concepts of the synonyms, without searching individually, it is extremely efficient.

Further, at the time of full-text search, the synonyms of the input search character string are searched from the synonym dictionary and disclosed to the user via the input / output means, and one or more selected by the user. Perform a full-text search with the synonyms of as the search string. By doing so, it is possible to efficiently perform a search that meets the needs of the user.

[0029]

【Example】

(First Embodiment) A first embodiment is an automatic keyword extraction device. As shown in FIG. 1, this device cuts off a keyword candidate word from a document storage unit 11 for storing a search target document. A dictionary 12 that is referred to for output, a thesaurus 13 that describes the upper-lower relationship between keyword candidate words, and a keyword candidate word cutout that cuts out keyword candidate words from the search target document using the dictionary 12 and thesaurus 13. Department
14, a keyword candidate word selection unit 15 that selects a keyword suitable for the target document from the extracted keyword candidate words through interaction with the user, and an input that accepts input from the user and presents information to the user. The output unit 16,
A keyword extraction result storage unit 17 for recording the correspondence between the search target document and the selected keyword is provided.

The thesaurus 13 is, as illustrated in FIG.
It defines the relationship between a broader word that represents a broader concept and a lower word that is semantically included in it. It also has multiple meanings, such as "month" (the moon as a satellite and the month and day of the month). A caution word mark (*) is attached to a keyword candidate word that is easily held and is likely to be mistaken. The number after the word indicates the word number.

In this keyword automatic extraction device, first,
The keyword candidate word cutout unit 14 sequentially reads the documents stored in the document storage unit 11, divides each document into words using the longest matching method or the morphological analysis method with reference to the dictionary 12, and then divides the words into words. Unnecessary words are deleted from the divided result to obtain a string of keyword candidate words.

In this case, words that can be keywords can be recorded in the dictionary 12 or thesaurus 13 in advance, and the keyword candidate word cutout unit 14 collates the entire text of the read document with the dictionary 12 or thesaurus 13 to find matching words. You may extract | require only and you may obtain | require a keyword candidate word string.

Thus, for example, the document number 2 shown in FIG.
A keyword candidate word string as shown in FIG. 22 is obtained from the document of 0.

Next, the keyword candidate word selecting unit 15 sequentially refers to the thesaurus 13 for each word of this keyword candidate word string to obtain a superordinate word of each word. At that time,
If the thesaurus 13 has a word registered at a plurality of locations, all the pairs of the word and the superordinate word of this word are displayed on the input / output unit 16 to allow the user to select one of them. For example, among the keyword candidate word strings in FIG. 22, the “name” is registered in two places in the thesaurus 13 in FIG. 3, so this word is displayed in the input / output unit 16 together with its superordinate word. FIG. 4 shows an example of the display screen at this time.

When the user confirms the displayed document and determines that the word "masterpiece" in this document means the "masterpiece" of the painting, the user selects "2 paintings" on the display screen. To do. The keyword candidate word selection unit 15 receives an input from the user and, as a keyword of this document (document number 20), “masterpiece” (word number 201) in the meaning of “painting”.
And its higher-ranking word “painting” (200) are stored in the keyword extraction result storage unit 17.

Further, the keyword candidate word selection unit 15 sequentially sets the thesaurus 13 for each word in the keyword candidate word string.
To find the broader term of each word, refer to
If there is a word written as a caution word in 13, the input / output device 16 once confirms that the word is a caution word.
It is displayed on the screen, and the user is warned, and is asked to select whether or not to register this word as a keyword. Of the keyword candidate word strings in FIG. 22, “month” is described as “attention word” in the thesaurus 13 as shown in FIG. FIG. 5 shows an example of the display screen at this time.

The user confirms the displayed document, confirms that the word "month" in this document is not "month" in the sense of the satellite, and selects "2 selection" on the display screen. Select "No". The keyword candidate word selection unit 15 receives an input from the user and determines not to register “month” as a keyword.

The operation procedure of such an automatic keyword extracting device will be described with reference to FIG.

Step 1: The number N of search target documents stored in the document storage unit 11 is stored, and the document number i is set to the initial value (1).

Step 2: keyword candidate word cutout unit 14
Reads the document with the document number i from the document storage unit 11, and step 3: divides this document into words using, for example, the dictionary 12, deletes unnecessary words, and obtains a keyword candidate word string of length K.

Step 4: The order j of keyword candidate words is set to an initial value (1).

Step 5: If j has not reached K, Step 6: The keyword candidate word selection unit 15 checks whether or not the jth word of the keyword candidate word string is registered in a plurality of locations in the thesaurus 13. Step 7: When the word is registered in a plurality of places, the word is displayed on the screen of the input / output unit 16 together with the upper word in the thesaurus.

Step 8: When the user looking at the screen selects an appropriate word and its superordinate word, Step 12: The word and superordinate word are stored as keywords in the keyword extraction result storage unit 17.

In step 6, if the j-th word of the keyword candidate word string is not registered in a plurality of locations in the thesaurus 13, step 9: checks whether the word is a caution word in the thesaurus, and 10: When the word is a caution word, the word is displayed on the screen of the input / output unit 16 together with the upper word in the thesaurus. Step 11: The user who looks at the screen determines whether to select the word as a keyword. .

Step 12: Keyword candidate word selection unit 15
Stores the word and the broader term as keywords in the keyword extraction result storage unit 17 except when not selected in step 11.

Step 13: The order j of the keyword candidate words is incremented by 1, and the steps 5 to 12 are repeated. Step 5: When j exceeds K, that is, the document number i
When the keyword selection processing for each word of the keyword candidate word string obtained from the document is finished, step 14: increment the document number by one, and step 2
The procedure of step 13 is repeated, and step 15: When the document number i exceeds N, that is, when the keyword extraction for all the documents stored in the document storage unit 11 is completed, the process ends.

As described above, in the keyword automatic extraction device of the first embodiment, the keyword is registered with the correct meaning upon the user's check. Therefore, in the document search, detection of an erroneous document is suppressed, and the search accuracy is improved.

(Second Embodiment) A second embodiment is a document retrieval apparatus. This apparatus, as shown in FIG. 6, has a document storage unit 61 for storing a retrieval target document and a keyword candidate word from the retrieval target document. A dictionary 62 referred to when extracting a keyword, a thesaurus 63 in which upper-lower relations among keyword candidate words are described, and a keyword automatic extraction unit that extracts a keyword from a document to be searched using the dictionary 62 or thesaurus 63. 64,
The keyword extraction result storage unit 65 that records the correspondence between the search target document and the keyword, the input / output unit 67 that receives the input from the user and presents the information to the user, and the character string input by the user A keyword search unit 66 for searching a target document as a keyword is provided.

The thesaurus 63 is shown in FIG. 8, which is the same as the thesaurus (FIG. 3) of the first embodiment.

The automatic keyword extraction unit 64 of this apparatus corresponds to the keyword candidate word cutout unit 14 and the keyword candidate word selection unit 15 of the first embodiment (FIG. 1), and the document storage unit 61.
The documents stored in the above are sequentially read, the keywords of each document are automatically extracted using the dictionary 62 and thesaurus 63 while exchanging with the user, and the extraction result is stored in the keyword extraction result storage unit 65. For example, for the document (document number 20) shown in FIG.
22 extracts the keywords shown in FIG. 22 and also extracts the keywords shown in FIG. 25 for the document (document number 24) shown in FIG. At this time, as shown in the first embodiment, with respect to the keywords registered in a plurality of locations of the thesaurus 63, the keyword having the correct meaning is extracted through the interaction with the user.

As shown in FIG. 9, the extracted keywords are registered in the keyword extraction result storage unit 65 in the form of the keyword and its word number, the total number of documents containing the keyword, and each document number of the document. It Here "masterpiece"
The two keywords are registered, but "masterpiece"
The (word number 101) is a “masterpiece” having the meaning of a movie which is the keyword of the document in FIG. 24, and the “masterpiece” (word number 201) has the meaning of a painting which is the keyword of the document in FIG. It is a masterpiece.

When the registration of the keyword of the document to be searched is completed in this way, the document search device executes the document search according to the procedure shown in FIG.

Step 21: When the user inputs, for example, "masterpiece" as the search keyword from the input / output unit 67, the keyword search unit 66 causes the thesaurus 63 to search for the higher-ranking word ("movies"
"Painting").

Step 23: When there are a plurality of broader terms of the search keyword in the thesaurus, that is, when the search keyword is registered in plural places of the thesaurus, Step 24: The number of documents including this search keyword is set as the keyword. Obtained from the extraction result storage unit 65, the search keyword, its superordinate word, and the number of documents containing the search keyword are displayed on the screen of the input / output unit 67 (an example of this display screen is shown in FIG. 10).

Step 25: The user looks at the screen and selects which superordinate word the keyword to be searched is.

Step 26: The keyword search unit 66 selects the search keyword when the search keyword is not registered in a plurality of locations in the thesaurus, and the selected search keyword when the selection is made in step 25. The document number of the included document is searched from the keyword extraction result storage unit 65.

Therefore, if the search keyword input by the user is "masterpiece", the screen of FIG. 10 is displayed, and the user can select "1 movie" or "2" from this screen according to his / her search intention. Select "Painting". For example, when "2 paintings" is selected, the keyword search unit 66 receives this selection and searches the keyword extraction result storage unit 65 for documents having only the keyword "masterpiece (201)", which is the search result. The document number 20 is displayed.

As described above, in the document search apparatus of the second embodiment, when the search condition character string input by the user is registered in a plurality of locations in the thesaurus, that is, when the character string has a plurality of meanings. , Confirm the user's search intention,
Then, the search is executed. Therefore, an efficient and highly accurate search can be performed.

(Third Embodiment) The document search apparatus according to the third embodiment can select between keyword search and full text search.
As shown in FIG. 11, this apparatus is provided with a full-text search unit 117 for matching the full text of the search target document with the character string input from the input / output unit 118 and searching for the target document containing this character string. There is. The other structure is the same as that of the document retrieval apparatus (FIG. 6) of the second embodiment.

In this apparatus, the keyword automatic extraction unit 114
, Sequentially reads the documents stored in the document storage unit 111, automatically extracts the keywords, and stores them in the keyword extraction result storage unit 115. This operation is the same as in the second embodiment. Now, as a document to be searched, the document storage unit 111 is shown in FIG.
Document (document number 20), the document of FIG. 24 (document number 2)
4) and the document of FIG. 26 (document number 26) are stored, and from these documents, FIG. 22, FIG. 25, and FIG. 27, respectively.
It is assumed that the keyword is extracted and registered in the keyword extraction result storage unit 115 in the state shown in FIG.

A user who searches for a document is input / output unit 118.
From, input the search character string and select the keyword search mode or the full text search mode. For example, when the user selects the keyword search mode and inputs the search character string "masterpiece", as described in the second embodiment,
The keyword search unit 116 executes a highly accurate search according to the user's search intention while interacting with the user via the input / output unit 118 as appropriate.

However, when the user selects the keyword search mode and inputs a search character string that is not registered as a keyword itself such as "travel boom", the number of documents in the search result is 0. Become. In that case, when the user newly selects the full-text search mode, the full-text search unit 117 executes character string matching between the character string “travel boom” and each document stored in the document storage unit 111,
Document that contains the character string "Travel boom", that is, document number 2
6 documents are detected. In other words, when the user wants to comprehensively obtain documents including the character string "travel boom", the user can search for the documents by designating the full-text search mode.

The procedure of the search operation of this apparatus is shown in the flowchart of FIG.

Step 31: When the user inputs the character string of the search condition and the search mode from the input / output unit 118, Step 32: When the search mode is the keyword search mode, the steps 33 to 38 are executed. To do. This procedure is the same as the procedure of the second embodiment (FIG. 7) (step 22 to step 26).

In step 32, when the search mode is not the keyword search mode, step 34: the full-text search section 117 sequentially reads the search target document from the search storage section 111, and extracts the full-text of the document and the character string of the search condition. And the target document containing the character string is obtained.

As described above, in the document search apparatus of the third embodiment, the search mode can be selected, and the user selects the full-text search mode when he or she desires an exhaustive search, and the accuracy is high. You can select the keyword search mode if you want to search. In addition, if you perform a search in the full text search mode and change the number of search results to a keyword search mode to narrow down the search results, or execute the keyword search mode once and the number of search results is 0 or When it is close, a full-text search mode is selected and further related documents are searched for, and a flexible and efficient search can be performed.

(Fourth Embodiment) The document retrieval apparatus according to the fourth embodiment can improve the efficiency of retrieval by synonyms. This device, as shown in FIG. 14, includes a search character string conversion unit 148 that converts a character string of a search condition input from the input / output unit 150 into a synonym, and a synonym dictionary 149 used for this conversion. I have it. The other structure is the device of the third embodiment (FIG. 11).
There is no change.

As shown in FIG. 16, the synonym dictionary 149 describes a relationship between a synonym word group having the same meaning and an upper word in the thesaurus 143 corresponding to this word group.

In this document retrieval apparatus, the keyword automatic extraction unit 144 is used to extract the keywords of the retrieval target document.
Is the word cut out from each document and the thesaurus of this word
All upper terms in 143 are extracted as keywords and stored in the keyword extraction result storage unit 145.

For example, when "computer" is defined in the thesaurus 143 as a superordinate word for "electronic computer", "computer", and "computer", the keyword automatic extraction unit 144 uses the document (document number) shown in FIG. Two
8), the document (document number 30) shown in FIG. 30 and the document (document number 32) shown in FIG.
The keywords 1, 33 are extracted. These keywords are registered in the keyword extraction result storage unit 145 in the state shown in FIG.

After registering the keyword, the user who wants to perform the document search inputs the search character string from the input / output unit 150 and selects the keyword search mode or the full-text search mode. For example, if the keyword search mode is selected and the search string "computer" is entered,
When the selected mode is the keyword search mode, the search string conversion unit 148 searches the synonym dictionary 149 for a keyword “computer” that is a superordinate word of a synonym group including “electronic computer”, Search unit 14
Tell 6.

In response to this, the keyword search unit 146 searches the keyword extraction result storage unit 145 for a document having the keyword "computer". As shown in FIG. 17, since the documents of document numbers 28, 30, and 32 are all registered in the keyword extraction result storage unit 145 as having the keyword “computer”,
These three documents will be detected as a result of the search.

As described above, once the search character string is converted into a keyword that is a high-order word, it is not necessary to individually search for "computer", "computer", etc., which are synonymous with "electronic computer". , The keyword search can be made efficient.

On the other hand, when the full-text search mode is selected and the search character string "electronic computer" is input, the search-character-string conversion unit 148 determines that the selected computer is "electronic computer" when the selected mode is the full-text search mode. The synonym group including “” is searched from the synonym dictionary 149, and words belonging to this synonym group are presented to the user via the input / output unit 150. The display screen at this time is illustrated in FIG.

The user selects one or a plurality of words from the displayed words. For example, if the word "computer" is selected, the search string conversion unit 14
8 transmits the selected character string to the full-text search unit 147, and the full-text search unit 147 matches the full-text of the document read from the document storage unit 141 with the word "computer", and detects the document containing this word. To do. As a result, the document with the document number 32 including the word "computer" is detected, but the documents with the document numbers 28 and 30 are not detected.

FIG. 15 shows an operation procedure of the document search device which performs such an operation. Step 41: When the user inputs the character string of the search condition and the search mode from the input / output unit 150, Step 42: When the search mode is the keyword search mode, Step 43: The search character string conversion unit 148 has the same meaning. From the word dictionary 149, the upper word of the synonym word group to which the search character string belongs is found, and step 44 to step 48: the keyword search unit 146 searches for documents having this upper word as a keyword. The procedure of steps 44 to 48 is the same as the procedure of the second embodiment (FIG. 7) (steps 22 to 26).

In step 42, when the search mode is not the keyword search mode, step 49: the search character string conversion unit 148 obtains a synonym word group to which the search character string belongs from the synonym dictionary 149, and step 50: this Display the words contained in a word group on the screen.

Step 51: When the user viewing the screen selects a word to be used for full-text search, Step 52: The full-text search unit 147 collates the search target document read from the document storage unit 141 with the selected word. Then, all documents including the word are detected.

As described above, in the document search device of the fourth embodiment, in the case of keyword search, since the high-order word of the search condition character string is used as a keyword, efficient search is possible. Further, in the case of full-text search, the user is given the opportunity to select a search condition character string from the synonym group, and it is possible to execute a search that meets the needs of the user.

In the document retrieving apparatus of the second, third and fourth embodiments, the keyword automatic extracting unit includes a keyword candidate word cutting unit and a keyword candidate word extracting unit of the keyword automatic extracting apparatus of the first embodiment. Although it is desirable to have a configuration corresponding to the selection unit, it is also possible to employ another configuration capable of extracting an appropriate keyword from the search target document using a dictionary and a thesaurus.

[0081]

As is apparent from the above description of the embodiments, the keyword automatic extraction device of the present invention confirms the user's intention with respect to a word having a plurality of meanings or a word that is easily mistaken. Since the keyword is extracted, it is possible to extract an appropriate keyword according to the content of the document, and as a result, it is possible to improve the accuracy in keyword search.

Further, since the document search apparatus of the present invention performs the search after clarifying the user's search intention when the input search keyword has a plurality of meanings, the search is highly accurate. The result can be obtained efficiently.

Further, in the document search device capable of selecting the keyword search mode and the full-text search mode, it is possible to perform a flexible and highly accurate search in line with the user's intention.

Further, in the document search device provided with the synonym dictionary, the keyword search by the synonyms can be made efficient, and in the full text search, the synonyms used for the search character string can be selected. It is possible to perform full-text search that meets the needs of users.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a keyword automatic extraction device according to a first embodiment of the present invention,

FIG. 2 is a flowchart showing the operation of the automatic keyword extraction device of the first embodiment,

FIG. 3 is a conceptual diagram of a thesaurus in the keyword automatic extraction device according to the first embodiment,

FIG. 4 is a display example in the keyword automatic extraction device according to the first embodiment (when a word has a plurality of meanings),

FIG. 5 is a display example (when a word is a caution word) in the keyword automatic extraction device of the first embodiment,

FIG. 6 is a block diagram showing a configuration of a document search device according to a second embodiment of the present invention,

FIG. 7 is a flowchart showing the operation of the document search device of the second embodiment,

FIG. 8 is a conceptual diagram of a thesaurus in the document search device according to the second embodiment,

FIG. 9 is a conceptual diagram of a keyword extraction result storage unit in the document search device of the second embodiment,

FIG. 10 is a display example of the document search device according to the second embodiment (when the search condition has a plurality of meanings),

FIG. 11 is a block diagram showing a configuration of a document search device according to a third embodiment of the present invention,

FIG. 12 is a flowchart showing the operation of the document search device of the third embodiment,

FIG. 13 is a conceptual diagram of a keyword extraction result storage unit in the document retrieval device of the third embodiment,

FIG. 14 is a block diagram showing the configuration of a document search device according to a fourth embodiment of the present invention,

FIG. 15 is a flowchart showing the operation of the document search device of the fourth embodiment,

FIG. 16 is a conceptual diagram of a synonym dictionary in the document search device according to the fourth embodiment,

FIG. 17 is a conceptual diagram of a keyword extraction result storage unit in the document search device of the fourth embodiment,

FIG. 18 is a display example of the document search device according to the fourth embodiment (when displaying synonyms).

FIG. 19 is a block diagram showing the configuration of a conventional keyword automatic extraction device,

FIG. 20 is an example of a search target document (document number 20);

FIG. 21 is an example in which a search target document (document number 20) is divided into words,

FIG. 22 is an example of keywords extracted from a search target document (document number 20);

FIG. 23 is a conceptual diagram of a keyword extraction result storage unit in a conventional keyword automatic extraction device,

FIG. 24 is an example of a search target document (document number 24),

FIG. 25 is an example of keywords extracted from a search target document (document number 24);

FIG. 26 is an example of a search target document (document number 26),

FIG. 27 is an example of keywords extracted from a search target document (document number 26);

FIG. 28 is an example of a search target document (document number 28),

FIG. 29 is an example of keywords extracted from a search target document (document number 28);

FIG. 30 is an example of a document to be searched (document number 30),

FIG. 31 is an example of keywords extracted from a search target document (document number 30);

FIG. 32 is an example of a search target document (document number 32),

FIG. 33 is an example of keywords extracted from a search target document (document number 32).

[Explanation of symbols]

 11, 61, 111, 141, 191 Document storage unit 12, 62, 112, 142, 192 Dictionary 13, 63, 113, 143, 193 Thesaurus 14 Keyword candidate word cutout unit 15 Keyword candidate word selection unit 16, 67, 118, 150 Input / output unit 17, 65, 115, 145, 195 Keyword extraction result storage unit 64, 114, 144, 194 Automatic keyword extraction unit 66, 116, 146 Keyword search unit 117, 147 Full text search unit 148 Search character string conversion unit 149 Synonym dictionary

Front page continued (72) Inventor Chuichi Kikuchi 1006 Kadoma, Kadoma City, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. (72) Teruichi Kiryu 1-1-1, Hitotsubashi, Chiyoda-ku, Tokyo Stock Company Mainichi Newspaper (72) Inventor Tetsuya Otsuka 1-1-1, Hitotsubashi, Chiyoda-ku, Tokyo Stock company Mainichi Newspaper

Claims (7)

[Claims] 1. An automatic keyword extracting device for automatically extracting a keyword from a document to be searched using a dictionary or a thesaurus in which upper-lower relations among keyword candidate words are written, in which the keyword candidate word is cut from the document to be searched. A keyword candidate word cutting-out means for outputting and a keyword candidate word for selecting the keyword candidate word as a keyword after confirming the user's intention when the cut-out keyword candidate word is described in a plurality of locations in the thesaurus. An automatic keyword extraction device comprising: a selection unit and an input / output unit that presents information to the user and receives selection input from the user. 2. Among the keyword candidate words described in the thesaurus, a keyword candidate word that needs attention in selecting a keyword is marked in advance with a caution word, and the keyword candidate word cut out by the keyword candidate word cutting means is the above-mentioned keyword candidate word. The automatic keyword extraction device according to claim 1, wherein, when the keyword is selected, the keyword selection unit selects the keyword candidate word as a keyword after confirming the intention of the user. 3. A keyword is automatically extracted from a document to be searched using a dictionary or a thesaurus in which upper-lower relationships between keyword candidate words are described, and the extraction result is stored in the keyword extraction result storage means and used. In a document retrieval device for retrieving a document having this keyword from the keyword extraction result storage means using a retrieval character string input by a user as a keyword, when the retrieval character string is described in a plurality of locations in the thesaurus, the user The keyword search means for confirming the intention of the user and searching the document having the keyword intended by the user from the keyword extraction result storage means, and the input / output means for presenting information to the user and accepting the selection input from the user A document retrieval device characterized in that and are provided. 4. A keyword candidate word cutout unit for cutting out a keyword candidate word from the search target document as a unit for automatically extracting a keyword from the search target document, and the cut out keyword candidate words at a plurality of locations in the thesaurus. 4. The document search device according to claim 3, further comprising a keyword candidate word selection unit that selects the keyword candidate word as a keyword after confirming a user's intention. 5. A full-text search unit for searching a search target document including the search character string by character string matching is provided, and a search by the keyword search unit or the full-text search unit can be selected from the input / output unit. Claim 3
Document retrieval device described in. 6. A synonym dictionary describing a correspondence between a word group having a synonym relationship and a keyword candidate word of the thesaurus, and a search character string input by a user are converted using the synonym dictionary. A search character string conversion means is provided,
The document search device according to claim 3 or 5, wherein the keyword search means supplies the keyword candidate word of the thesaurus converted by the search character string conversion means as a search character string. 7. The search character string conversion means converts a search character string input by a user into a word group of synonyms using the synonym dictionary, and the full text search means performs the word group conversion. 7. The document search device according to claim 6, wherein a character string selected by the user from among the above is given as a search character string.