JPH0944523A - Relative word display device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To specify a key word and extract a relative word from text information at high speed by the relative word display device which extracts and displays a word (relative word) in close relation with a specified word. SOLUTION: This relative word display device, which extracts and displays the relative word from the text of a document, is equipped with an analytical information storage part 1 which stores analytical information generated by taking a morpheme analysis of the text, a key word input part 2 which receives a key word and takes the morpheme analysis, a retrieval process part 3 which retrieves the analytic information from the analytic information storage part 1 with the key word and extracts a noun cooccurring with the key word, a cooccurrence word registration part 4 which registers the extracted noun in a frequency table 5 together with the frequency, a representative word selection part 6 which selects representative words out of the data in the frequency table 5, an unnecessary word deletion part 7 which deletes previously specified unnecessary words from the selected representative words, and a display part 8 which displays the representative words after the unnecessary words are deleted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a related word presenting apparatus, and more particularly to a related word presenting apparatus for extracting and presenting a word (related word) having a strong relationship with a designated word.

[0002]

2. Description of the Related Art In recent years, "words having a strong relationship with a certain word" called related words, associative words, and co-occurrence words have been widely used in the technical fields of natural language processing and information retrieval. . In the following description, in particular, such a "word having a strong relationship with a certain word" is used as a related word in the above meaning.

Conventionally, as a method of utilizing this kind of related word for searching, a proposal of "a related word chain search method" described in Japanese Patent Laid-Open No. 2-56068 is known, and the related word is indexed. Japanese Patent Application Laid-Open No. 3-122
A proposal of "indexing support device" described in Japanese Patent No. 768 is known. In these proposed devices, a dictionary such as a related word dictionary or a thesaurus dictionary is used to obtain a related word.

As a method of creating a dictionary for using such related words, the first method has been as follows.
Two methods are used: a method of manually creating a dictionary by hand, and a second method of creating a dictionary by using co-occurrence frequency (frequency of simultaneous appearance) from text. .

As a method classified into the above-mentioned second method, for example, a method of utilizing the co-occurrence frequency for extraction, there is disclosed in Japanese Patent Laid-Open No. 6-274541. , "Co-occurrence word extraction method and its device" described in JP-A-7-28823, and "Co-occurrence dictionary creation device" described in JP-A-7-36883.

[0006]

However, in the above-mentioned first conventional method of manually creating the dictionary, (1) it takes a huge cost to create the dictionary, and it is biased depending on the creator. (2) It requires a large storage capacity for the dictionary. (3) The headwords are limited, and the capacity further increases as the number of words increases. There is a problem.

In the second method of creating a dictionary from the text by utilizing the co-occurrence frequency, the problem of (1) above is solved, but the problems of (2) and (3) above are solved. The problem is not solved, and as another problem, (4) the extraction accuracy of the related words that are the contents of the dictionary is low. The problem also arises.

In other words, the unit of counting the co-occurrence frequency is not taken into consideration for the compound word included in the document to be searched, and the unit is inappropriate, so the co-occurrence frequency is properly counted. There is a problem that the extraction accuracy becomes low due to not being performed or redundant information increasing.

Generally, the unit for counting the co-occurrence frequency is either a morpheme unit, the entire compound word, or a combination of morphemes, but in that case, the counted number differs depending on each unit. Come on. For example,
For the compound word "patented electronic application", if the co-occurrence frequency is counted in the morpheme unit as the first case, in this case, the unit of three morphemes of "patent", "electronic", and "application" The co-occurrence frequency is counted accordingly. In this case, although the co-occurrence frequency is properly counted, it is often difficult to understand the meaning in units of morphemes.

In the second case, if the co-occurrence frequency is counted in units of the entire compound word, the frequency is counted in "patented electronic application", and its meaning is easy to understand.
In this case, the counted compound word is "patent",
Related words such as “patent application” will be treated as different words. Therefore, the co-occurrence frequency may not be properly counted.

When the co-occurrence frequency is counted by the combination of morphemes in the third case, in this case,
From each combination of morphemes, "patent", "electronic", "application", "electronic patent", "electronic application", "electronic patent application"
Since the co-occurrence frequency is counted, the meaning is easy to understand and the frequency is properly counted.However, when the number of morphemes increases, the number of combinations increases, so the capacity is very large. Grows to. Further, since the frequency of the morphemes forming the compound word is always higher than the frequency of the compound word, there is a drawback that the information of the morphemes remains and becomes redundant.

The present invention has been made to solve such a problem, and an object of the present invention is to use a text information on the spot when a keyword is specified without using a related word dictionary. Another object of the present invention is to provide a related word presentation device capable of extracting related words at high speed and improving the extraction accuracy of the related words.

[0013]

In order to achieve the above-mentioned object, a related word presenting apparatus, which is a first feature of the present invention, is a related word presenting apparatus for extracting and presenting related words from a text of a document. The content of the storage means is searched by the storage means for storing the text, the keyword input means for receiving the keyword, and the keyword input by the keyword input means, and the keyword and the text are shared within a predetermined range. Search processing means for extracting the raised word,
Co-occurrence word registration means for registering the noun extracted by the search processing means in the frequency table together with the appearance frequency in the text, and representative word selection means for selecting a representative word from the data in the frequency table based on the appearance frequency. And a display unit for displaying a representative word in which unnecessary words are deleted by the unnecessary word deleting unit, and an unnecessary word deleting unit for deleting unnecessary words from the representative word selected by the representative word selecting unit. To do.

A second aspect of the present invention is a related word presentation device that extracts a related word from a text of a document and presents it, and stores analysis information obtained by morphological analysis of the text. An analysis information storage unit (1: FIG. 1), a keyword input unit (2: FIG. 1) that accepts a keyword and performs morpheme analysis, and analysis information in the analysis information storage unit by a keyword morphologically analyzed by the keyword input unit. A search processing unit (3: FIG. 1) for searching and extracting a noun co-occurring in a predetermined range of the keyword and the text, and the noun extracted by the search processing unit are registered in a frequency table together with the appearance frequency in the text. Co-occurrence word registration unit (4: Figure 1)
And a representative word selection unit (6: FIG. 1) that selects a representative word from the data of the frequency table based on the appearance frequency, and deletes unnecessary words specified in advance from the representative word selected by the representative word selection unit. And a display unit (8: FIG. 1) for displaying a representative word in which unnecessary words are deleted.

A third aspect of the present invention is a related word presentation device that extracts a related word from a text of a document and presents it, and a text storage unit that stores the text of the document. (71: FIG. 7) and a keyword input unit that accepts a keyword and performs morphological analysis (2: FIG. 7)
And a search analysis unit (72: FIG. 7) for searching the text in the text storage unit by the keyword, performing a morphological analysis of the search result, and extracting a noun co-occurring with the keyword, and a frequency table for the extracted noun together with the frequency. The co-occurrence word registration unit (4: FIG. 7) registered in the table, the representative word selection unit (6: FIG. 7) selecting a representative word from the data of the frequency table, and the unnecessary word specified in advance from the selected representative word. And a display unit (8: FIG. 7) for displaying a representative word in which the unnecessary words are deleted.

A fourth aspect of the present invention is a related word presenting apparatus having the above-mentioned configuration, wherein the co-occurrence word registration unit registers the appearance frequency in a frequency table. The extracted morpheme of the noun and the original compound word of the noun are registered in the table in association with each other, and the representative word selection unit selects a representative from the data of the morpheme and the original compound word according to the frequency registered in the frequency table. Characterized by selecting a word. Further, preferably, in the related word presentation device (corresponding to claims 2 to 4), the analysis information stored in the analysis information storage unit is information in which a division symbol for dividing a text into morphemes is added to the text. Is.

In the related word presentation device of the present invention having such characteristics, as a first characteristic, the analysis information storage unit stores the analysis information obtained by morphologically analyzing the text, and the keyword input unit uses the keyword. When the morphological analysis is accepted, the search processing unit searches the analysis information in the analysis information storage unit by the keyword and extracts the noun co-occurring with the keyword. With respect to the thus extracted noun, the co-occurrence word registration unit registers the extracted noun together with the frequency in the frequency table. Next, the representative word selection section
Select a representative word from the data in the frequency table. The unnecessary word deleting unit deletes, for example, an unnecessary word previously designated as meaningless from the selected representative word, and the display unit displays the representative word from which the unnecessary word is deleted. In this way,
According to the related word presentation device of the present invention, when a keyword is designated, the related word can be extracted and presented from the analysis information of the text of the document. In keyword search,
Specifically, for example, a character string search (grep or the like), a search by an index (the index may be a morpheme or a compound word unit), and the like can be used. Further, extraction of co-occurring words requires analysis, which can be dealt with by performing morphological analysis. Moreover, you may divide by a character type.

According to the related word presenting apparatus of the present invention as described above, since the related words are extracted from the information of the morpheme obtained by analyzing the text by utilizing the frequency of co-occurrence with the keyword, a dictionary such as a related word dictionary. Related words can be extracted at high speed without the need for. Further, when using a dictionary, the keywords are limited to the entry words of the dictionary, but in the present invention, since the target of extracting the related word is the information stored as the text analysis information, a large amount of information is stored. Including, you can flexibly determine the keyword. Moreover, the related word in the field can be obtained by changing the analysis information of the text from which the related word is extracted. In addition, the limitation of keywords in the dictionary is eliminated.

In the related word presentation device of the present invention, as a second feature, when the co-occurrence word registration unit registers the frequency in the frequency table, the extracted noun morpheme and the original compound word are associated with each other. And the representative word selection unit selects a representative word from the data of the morpheme and the original compound word according to the frequency registered in the frequency table. Then, the unnecessary word deleting unit deletes a previously specified unnecessary word meaningless as a related word. Therefore, redundant information and dust information that causes noise in extracting related words can be eliminated, and related words can be accurately extracted. Moreover, this can improve the accuracy of the related words.

As a third feature, in the related word presentation device of the present invention, the analysis information stored in the analysis information storage unit is subjected to morpheme analysis of text in advance, and the character string of the text is divided for each morpheme. Since the information is a division symbol added to the text, the capacity increases only about 20% of the original text, and the original text can be reproduced without the division symbol. Therefore, when it is combined with another search system, there is an advantage that it can be shared with the search target text.

Therefore, the related word presenting apparatus of the present invention extracts the related words on the spot from the text information, so that the cost of constructing the dictionary is lower than that of the conventional method using the related word dictionary. It is unnecessary, does not require a capacity for the dictionary, and has an advantage that the keyword can be determined more flexibly as compared with the dictionary. Further, by changing the content of the text, it is possible to obtain a related word in the field of the text.

As a fourth feature of the related word presentation device of the present invention, the text storage unit stores the text information of the document, and when the keyword input unit receives a keyword and performs morphological analysis, The search analysis unit searches the text information in the text storage unit by the keyword, performs the morphological analysis of the search result, and extracts the noun co-occurring with the keyword.

For the noun thus extracted,
The co-occurrence word registration unit registers the extracted noun in the frequency table together with the frequency, and then the representative word selection unit selects a representative word from the data in the frequency table. The unnecessary word deletion part is
From the selected representative word, for example, an unnecessary word previously designated as meaningless as a representative word is deleted, and the display unit displays the representative word from which the unnecessary word is deleted. Thus, according to the related word presentation device of the present invention, since the related word is directly extracted from the text of the document and presented, it is possible to easily change the text information of the document stored in the text storage unit. By changing to text information of a different field, the related word of the field can be easily obtained from it.

[0024]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be specifically described below with reference to the drawings. FIG.
FIG. 1 is a block diagram showing a configuration of a related word presentation device according to a first exemplary embodiment of the present invention. In FIG. 1, 1 is an analysis information storage unit, 2 is a keyword input unit, 3 is a search processing unit, 4
Is a co-occurrence word registration unit, 5 is a frequency table, 6 is a representative word selection unit, 7 is an unnecessary word deletion unit, and 8 is a display unit. Also, 10
Indicates an input keyword, and 11 indicates an output related word.

When the user wants to present the related words, the user performs morphological analysis on the text of the document to be searched in advance and stores the analysis information in the analysis information storage unit 1.
Therefore, as will be described later, the analysis information stored in the analysis information storage unit 1 is, for example, information obtained by morphologically analyzing the text of the target document and inserting a division symbol between each analyzed morpheme. Has become.

A keyword (word) to be searched by the user 1
When 0 is input from the keyword input unit 1, the keyword input unit 2 obtains the keyword 10, performs morphological analysis of the keyword, and sends the morphological information of the analysis result to the search processing unit 3. When the search processing unit 3 obtains the morpheme information of the keyword, the search processing unit 3 searches the analysis information of the text stored in the analysis information storage unit 1, extracts the noun that co-occurs with the keyword, and shares the extracted noun information. It is sent to the word registration unit 4.
The range of co-occurrence can be sentences, sections, items, documents, etc.,
Here, it explains as a sentence. That is, a noun that co-occurs with the keyword in one sentence (a noun included in the sentence in which the keyword appears) is extracted and sent to the co-occurrence word registration unit 4.

When the co-occurrence word registration unit 4 obtains the information of the corresponding noun co-occurring with the keyword from the information of the noun extracted from the search processing unit 3, the co-occurrence word registration unit 4 refers to the frequency table 5, Regarding the noun, the noun is registered in the frequency table 5 together with the frequency. That is, new nouns are registered, nouns that have appeared so far are updated in frequency, and registered in the frequency table 5. In this way, the frequency is registered in the frequency table 5 together with the noun.

The representative word selecting unit 6 refers to the frequency information registered in the frequency table 5, selects a representative word to be displayed as a related word by a predetermined algorithm described later, and does not need information on the selected representative word. It is sent to the word deletion unit 7. In the unnecessary word deleting unit 7, a meaningless unnecessary word is designated in advance as a representative word, and when the information of the representative word is obtained, the unnecessary word deleting unit 7
The unnecessary word is deleted from the information of the representative word, and the result information is sent to the display unit 8. When the display unit 8 obtains the information of the representative word resulting from the unnecessary word deletion, the information of the representative word is presented to the user as a related word of the keyword.

Next, using a specific text and a keyword (an example of a sentence and a word), the related word presentation device of the present embodiment specifies a keyword from the analysis information of the text to be searched, and sets the keyword as a keyword. An example of processing of registering a co-occurring noun in the frequency table, extracting related words from the information in the frequency table, and presenting the related words will be described.

FIG. 2 is a diagram for explaining an example of analysis information stored in the analysis information storage section. As shown in FIG. 2, analysis information which is data to be searched for a related word from a keyword is morphological analysis processing 21 performed on text data 20 of document data, and analysis information is analyzed in units of morphemes. 22. The analysis information 22 is information created by inserting a division symbol (“/”) between each morpheme obtained by morpheme analysis of text.

That is, in the analysis information 22 stored in the analysis information storage unit, a text sentence is morphologically analyzed and divided into morpheme units. The analysis level of the morpheme analysis here is that at least the sentence is analyzed so that each morpheme is distinguished, so as shown in the example sentence of FIG. 2, in the kana-kanji conversion of a word processor (word processor) ... ”, The result of morphological analysis,
Each morpheme, such as "Word processor", "(", "word processor", ")", "no", "kana", "kanji", "conversion", "de", "ha", ... " It is divided into units of and is expressed as "word processor / (/ word processor /) / no / kana / kanji / conversion / de / wa / ..." Using the division symbol "/". That is, it is stored as information of a state in which the unit of morpheme can be identified. The division symbol “/” for indicating the unit of the morpheme obtained as a result of the morpheme analysis can be obtained as it is by changing the symbol used for each part of speech.

Regarding the capacity of the analysis information, since the division symbol is added to the morphologically analyzed text here, it is increased by about 20% of the original text. The text can be reproduced. For this reason,
When combined with a search system, there is an advantage that search target texts can be easily shared. In addition, in the search processing for searching for related words here, the frequency count targets only nouns. Therefore, by not registering words other than nouns in analysis information, the capacity of analysis information is It can be smaller than the text capacity.

FIG. 3 is a diagram for explaining a process of obtaining a noun co-occurring in a sentence with a keyword in the analysis information by the search processing unit. In this case, when the user inputs the keyword "word processor", the noun that co-occurs with the keyword in the analysis information 22 in the one sentence with respect to the analysis information "word processor" 23 that is the result of morphological analysis of the keyword "word processor". A search process 31 for searching for is performed.

In the search processing 31, the analysis information 22 is searched by the keyword analysis information "word processor" 23, and the noun co-occurring in one sentence is extracted. That is, the index table 33 is searched for a sentence including the keyword analysis information "word processor" 23, and all nouns in the sentence are extracted. In this example,
The noun included in the sentence 32 in which the word processor appears is extracted, and the information (word processor, kana / kanji / conversion) of the extracted noun is stored in the next co-occurrence word registration unit (4: FIG. 1). Sent out.

In the process of extracting the co-occurring noun, the information of the appearance position is obtained by using the index table 33 storing the information of the word (morpheme) and the appearance position. From the sentence containing analysis information,
The noun in the sentence can be searched at high speed. When the keyword is a compound word and is divided into a plurality of morphemes, a product set including sentences of each morpheme is taken to perform a high-speed search.

FIG. 4 is a diagram for explaining the process of registering co-occurring nouns in the frequency table. In the frequency table, a keyword and a noun that co-occurs in one sentence are registered together with the frequency. The data structure of the frequency table here is, as shown in FIG. 4, a morpheme table 41 storing morphemes of nouns.
Has a structure in which the compound word table 42 to which frequency information is added is linked by a link pointer, and the morpheme of the extracted noun and the original compound word from which the morpheme is cut out are associated with each other, and It is registered with the frequency.

In this case, as shown in FIG. 4, a morpheme and a pointer to the second row (pointer to the information of the original compound word) are stored in the morpheme table 41 in the first row, and the morpheme in the second row is stored. The word table 42 stores the original compound word 42a and the frequency information 42b. The sum of the frequency information of the compound words in the second-stage compound word table 42 is the morpheme frequency in the first-stage morpheme table 41. For example, in the case of the noun "word processor", "word processor" is registered in the first stage morpheme table 41, and further, the noun "word processor" is registered in the second stage compound word table 42 pointing to this, and the frequency Add information. If already registered, only frequency information is added.

When the noun registered here is a compound word, for example, the noun "kana / kanji / conversion", "kana", "kanji", and "conversion" are stored in the first morpheme table 41. Is registered, the original compound word “Kana-Kanji conversion” is registered in the second-stage compound word table 42 pointed by each morpheme, and the frequency information is added.

As described above, the representative word selection unit (6: FIG. 1) selects a representative word from the information in the frequency table when the registration of the co-occurring noun in the frequency table (FIG. 4) is completed. The algorithm for selecting the representative word in this case is, for example, selecting one representative word from a group of one morpheme in the first morpheme table 41 of the frequency table and the second stage compound word table 42 pointed to by the morpheme. The maximum frequency of a compound word (noun in the compound word table 42) is the morpheme (morpheme table 4
When (1/2) of 1 noun) or more, the compound word (noun of the compound word table 42) is used as a representative word, and the other words are morphemes (noun of the morpheme table 41) as a representative word.

Explaining in a concrete example, in this example, the frequency of the word processor “word processor” in the “word processor” group is “8 (= 7 + 1)” in the first morpheme table 41, and the compound word table in the second row. The frequency of the compound word "word processor" of 42 is "7", and the frequency of this second step is (1/2) or more of the frequency of the first step. Therefore, the compound word "word processor" of the second step is Become a representative language.

In the "conversion" group, 1
The frequency of the morpheme “conversion” in the second morpheme table 41 is “7 (= 4 + 3)”, and the frequency of the compound word “kana-kanji conversion” in the second stage compound word table 42 is “3”. Since the eye frequency is less than (1/2) of the first stage frequency, 1
The morpheme "conversion" in the second row is the representative word.

Similarly, in the "kana" group, the frequency of the morpheme "kana" in the morpheme table 41 in the first row is "5 (= 3 + 2)" and the compound word table 4 in the second row.
The frequency of "Kana-Kanji conversion" in 2 is "3", and the frequency of the second stage is (1/2) or more of the frequency of the first stage. Is the representative word.

In this way, a representative word is determined in each group, and the result of eliminating duplication is transmitted to the unnecessary word deleting unit (7: FIG. 1) as a representative word (word processor, conversion, kana-kanji conversion). As described above, the unnecessary word deleting unit has previously designated a meaningless unnecessary word as a representative word, and when the representative word is obtained, the unnecessary word is deleted from the representative word.

FIG. 5 is a diagram showing an example of an unnecessary word list table used by the unnecessary word deleting unit. In the unnecessary word deletion part,
A word that has no meaning in itself and a word that is important as an element of a compound word but is not important by itself is designated as an unnecessary word in advance. For example, “conversion” is not important by itself, but it is important in the compound word “kana-kanji conversion”. Such unnecessary words are preliminarily classified and designated as the unnecessary word list table 50 as shown in FIG.

In this embodiment, since the unnecessary word deleting unit stores unnecessary words in advance as the unnecessary word list table 50 as shown in FIG. 5, the representative word selecting unit selects the representative words (word processor, conversion). , Kana-Kanji conversion), the representative word that matches the unnecessary word stored in the unnecessary word list table 50 is deleted. That is, in this example, the word "conversion" is deleted. As a result, “conversion” is deleted from the representative word (word processor, conversion, kana-kanji conversion) as an unnecessary word, and the resulting representative word (word processor, kana-kanji conversion) is transmitted to the display unit, which is the keyword “word processor”. It is displayed as a related word of. At this time, if the number of words is large, a fixed number of words (for example, 10 words) may be displayed.

FIG. 6 is a block diagram showing the flow of a series of processes of the overall operation of the related word presentation device according to the first embodiment of the present invention. This will be described with reference to FIG. For example, the keyword "word processor" for which the user requests a related word 6
When 1 is input from the keyword input unit 2, the keyword input unit 2 performs a morpheme analysis of the keyword “word processor” 61 and sends the morpheme information “word processor” 62 of the analysis result to the search processing unit 3. Search processing unit 3
When the morpheme information “word processor” 62 is obtained, the text analysis information 6 stored in the analysis information storage unit 1
3 is extracted, a noun that co-occurs with the keyword morpheme information "word processor" 62 in one sentence (a noun included in the sentence in which "word processor" appears) is extracted from the text analysis information 63, and the extracted noun is extracted. Information "Word Processor,
The kana / kanji / conversion ”64 is sent to the co-occurrence word registration unit 4.

When the co-occurrence word registration unit 4 obtains the noun information “word processor, kana / kanji / conversion” 64 co-occurring with the keyword from the search processing unit 3, the co-occurrence word registration unit 4 registers it in the frequency table 5 together with the frequency in association with the morpheme. . In this way, the frequency information 65 is registered in the frequency table 5 together with the noun information.

The representative word selecting unit 6 refers to the frequency information 65 registered in the frequency table 5, selects a representative word, and sets the representative word information “word processor, conversion, kana-kanji conversion” 66 to the unnecessary word deleting unit 7. Send out. The unnecessary word deletion unit 7 deletes unnecessary words that have been designated in advance from the representative word information “word processor, conversion, kana-kanji conversion” 66, and the result information “word processor, kana-kanji conversion” 67 is displayed on the display part 8 Send to. When the display unit 8 obtains the representative word information "word processor, kana-kanji conversion" 67 in which unnecessary words are deleted, the result information is presented to the user as the related word "word processor, kana-kanji conversion" 68.

With the flow of such processing, in the related word presentation device of the first embodiment, when a keyword is input, the related word is presented. That is, as is apparent from the flow of this process, the related word presentation device of this embodiment can directly extract the related words from the text analysis information on the spot. Therefore, as compared with the method using a dictionary, the cost of creating a dictionary (for example, a co-occurrence word dictionary) relating to related words is unnecessary, and the memory capacity of the dictionary is unnecessary. Therefore, when used in a document retrieval system, there is an advantage that the keyword can be determined more flexibly as compared with the method of preparing a dictionary in advance. In the related word presentation device of this embodiment, the related information in the field can be easily obtained by changing the analysis information stored in the analysis information storage unit 1 into the analysis information of documents in different fields.

In this case, for example, in the processing of the related word presentation device of this embodiment, if the analysis information is directly generated from the texts of documents in different fields and the analysis information can be used, the difference can be easily obtained. It can be modified so that related words in the field can be obtained. A related word presentation device of such a modified example will be described as a second embodiment.

FIG. 7 is a block diagram showing the configuration of a related word presentation device according to the second embodiment of the present invention. In FIG. 7, 2 is a keyword input unit, 4 is a co-occurrence word registration unit, 5 is a frequency table, 6 is a representative word selection unit, 7 is an unnecessary word deletion unit,
Reference numeral 8 is a display unit, 10 is a keyword to be input, and 11 is a presented related word. Reference numeral 71 is a text storage unit, and 72 is a search analysis unit. Among these elements, the keyword input unit 2, the co-occurrence word registration unit 4, the frequency table 5, the representative word selection unit 6, the unnecessary word deletion unit 7, and the display unit 8 are the related word presentation device of the first embodiment. (FIG. 1), each designated by the same reference numeral. In the related word presentation device of the second embodiment, a text storage unit 71 is provided in place of the analysis information storage unit (1: FIG. 1), and a search processing unit (3: FIG. 1) is provided. Thus, a search analysis unit 72 is provided.

Since the basic processing flow of the operation of the related word presentation device of the second embodiment having such a configuration is the same as the operation of the related word presentation device of the first embodiment, detailed description will be omitted. However, only different parts will be described. In the related word presentation device according to the second embodiment, the text storage unit 71 stores the text information of the document to be searched for the related words in a state in which the morphological analysis is not performed, and the search analysis unit 72.
In, a search process of the text storage unit 71 is performed with a keyword, a sentence including the keyword is morpheme-analyzed, and is used as analysis information of a text divided for each morpheme, that is, as information for extracting a noun.

In that case, for the text information, first,
The search process is performed using the keyword, and then the morphological analysis process is performed only on the sentence targeted for extracting the noun co-occurring in one sentence.

As described above, the search analysis section 72 uses the text information stored in the text storage section 71 as follows.
The keyword search is performed, and the morphological analysis is performed on the spot for the text information of the sentence including the keyword. Therefore, the source information for extracting the related word can be easily changed.

In other words, the related word presentation device of the first embodiment uses the analysis information obtained by performing morpheme analysis of the text in advance to search by the keyword, while the related word presentation device of the second embodiment does. In the presentation device, the text information is directly searched by a keyword, a sentence including the keyword is obtained, morphological analysis is performed on the obtained sentence on the spot, and a noun in the sentence is extracted. I try to extract it. For this reason, the text information for extracting the related words can be easily changed, and the frequency of co-occurrence with the keyword also differs depending on the field, and thus different arbitrary text information can be used.

FIG. 8 is a block diagram showing the flow of a series of processes of the overall operation of the related word presentation device according to the second embodiment of the present invention. Similar to the case of the related word presentation device of the first embodiment, the flow of processing here will be described by an operation example when the user inputs “word processor” as a keyword. Referring to FIG. 8, when the user inputs a keyword “word processor” 81 for which a related word is requested from the keyword input unit 2, the keyword input unit 2 is
Morphological analysis of the keyword “word processor” 81 is performed, and morphological information of the keyword “word processor” 82
Is sent to the search analysis unit 72. Upon obtaining the keyword morpheme information “word processor” 82, the search analysis unit 72 searches the text information 83 stored in the text storage unit 71 to obtain a sentence including the keyword morpheme information “word processor” 82. Then, morphological analysis is performed on the obtained sentence “word processor (word processor) kana-kanji conversion ...”, and noun information “word processor, kana / kanji / conversion” 84 other than the keyword “word processor” 84 in the sentence is obtained. It is extracted and sent to the co-occurrence word registration unit 4.

When the co-occurrence word registration unit 4 obtains the keyword and the noun information "word processor, kana / kanji / conversion" 84 that co-occurs in one sentence from the search analysis unit 72, the morpheme and the original compound word are associated with each other. The frequency table 5 together with the frequency. In this way, the frequency information 85 is registered in the frequency table 5 together with the noun information.

The representative word selecting unit 6 refers to the frequency information 85 registered in the frequency table 5, selects a representative word, and sets the representative word information “word processor, conversion, kana-kanji conversion” 86 to the unnecessary word deleting unit 7 Send to. The unnecessary word deletion unit 7 deletes unnecessary words that have been designated in advance from the representative word information “word processor, conversion, kana-kanji conversion” 86, and the result information “word processor, kana-kanji conversion” 87 is displayed on the display unit 8 Send to. When the display unit 8 obtains the representative word information “word processor, kana-kanji conversion” 87 from which unnecessary words are deleted, the result information is presented to the user as the related word “word processor, kana-kanji conversion” 88.

With the above processing flow, in the related word presentation device of the second embodiment, if a keyword is input, morphological analysis of a sentence containing the keyword in the text to be searched is performed on the spot. , Related words are presented.
As is clear from the flow of this processing, the text storage unit 71 of the second embodiment is a storage unit that stores the text of the document to be searched, so it can be shared with other search systems, and the text can be shared. As the search method for the text in the storage unit 71, as described above, a general search method such as an index search method in which a search key and position information are registered in the index table or a full-text search method in which text is directly searched. Can be used as is.

In the related word presentation device of the second embodiment, the text is searched from the morpheme of the keyword and the morpheme analysis is performed on the spot. Although the processing time is increased as compared with the first embodiment in which the text is searched for, the text does not need to be previously analyzed information and the text can be easily exchanged. Therefore, integration with other search systems is easy.

Next, as a third embodiment, a related word presentation device of a modified example which is integrated with the search system will be described. The related word presentation device of the third exemplary embodiment further has a configuration in which a text search mechanism is added and analysis information and search target text are made common. With such a configuration, when the keyword to be searched is not clear, the user can search for and search for the target keyword from the related words presented by the related word presentation device.

FIG. 9 is a block diagram showing the configuration of the related word presentation device according to the third embodiment of the present invention. In FIG.
1 is an analysis information storage unit, 2 is a keyword input unit, 3 is a search processing unit, 4 is a co-occurrence word registration unit, 5 is a frequency table, 6 is a representative word selection unit, 7 is an unnecessary word deletion unit, and 8 is a display unit. Is. 1
0 is a keyword to be input, and 11 is a presented related word. Further, 90 is a keyword selected from the presented related words, 91 is a search keyword input section, 92 is a text search section, 93 is a search result display section, and 94 is a search result.
Among these elements, analysis information storage unit 1, keyword input unit 2, search processing unit 3, co-occurrence word registration unit 4, frequency table 5,
The representative word selection unit 6, the unnecessary word deletion unit 7, and the display unit 8 are
It is the same as the related word presentation device (FIG. 1) of the first embodiment, and is indicated by the same reference numeral. In the related word presentation device of the third embodiment here, in addition to the configuration of the related word presentation device of the first embodiment (FIG. 1), a search keyword input unit 91, a text search unit 92, and a search result. A display unit 93 is provided. That is, the text search mechanism is added to the first embodiment.

When the user inputs the keyword 10 from the keyword input unit 2, the same processing as that performed by the related word presentation device of the first embodiment is performed, and the related word 11 for the keyword can be obtained from the display unit 8. The user refers to the related word 11 presented on the display unit 8 to determine a new keyword 90, and inputs the keyword 90 from the search keyword input unit 91 to perform a search. Alternatively, the obtained related word 11 may be merged with the original keyword 10 to search as a new keyword 90.

Upon obtaining the keyword (word) 90, the search keyword input unit 91 performs morphological analysis and sends the information to the text search unit 92. Text search unit 92
When the keyword analysis information is obtained from the search keyword input unit 91. The analysis information storage unit 1 is searched, and only the text information is sent to the search result display unit 93 from the sentence of the obtained analysis information excluding the division symbols. Search result display section 93
Displays the obtained text information as a search result 94.

As described above, according to the related word presentation device of the third embodiment, the related information presentation device can be integrated with the search system by sharing the analysis information and the search target text. The text search mechanism is the only additional mechanism for this purpose. In the related word presentation device of the third embodiment shown in FIG. 9, the search keyword input unit 91 and the search result display unit 93 are separately provided in order to clearly show the flow of processing. It is obvious that any of these functional units may be omitted by commonly using a part of the functions of the keyword input unit 2 and the display unit 8. Next, a configuration example of the related word presentation device that has undergone such a modification will be described as a fourth embodiment.

FIG. 10 is a block diagram showing the configuration of the related word presentation device according to the fourth embodiment of the present invention. In FIG. 10, 1 is an analysis information storage unit, 2 is a keyword input unit, 3 is a search processing unit, 4 is a co-occurrence word registration unit, 5 is a frequency table, and 6
Is a representative word selection unit, 7 is an unnecessary word deletion unit, 10 is a keyword to be input, 95 is a text search unit, 96 is a search result display unit, and 97 is a search result. Among these elements, the analysis information storage unit 1, the keyword input unit 2, the search processing unit 3, the co-occurrence word registration unit 4, the frequency table 5, the representative word selection unit 6, and the unnecessary word deletion unit 7 are the first embodiment. The elements are the same as those of the related word presentation device (FIG. 1) in the example, and are respectively denoted by the same reference numerals. The text search unit 95, the search result display unit 96, and the search result 97 in the configuration of the related word presentation device of the fourth embodiment are the text search unit in the configuration of the related word presentation device of the third embodiment. A search result display unit 92, a search result display unit 93, and a search result 94 correspond to each other and are similar.

That is, in the related word presentation device of the fourth embodiment, the display unit 8 and the search keyword input unit 91 are removed from the configuration of the related word presentation device of the third embodiment (FIG. 9). The text search unit 95 (92) is configured to obtain the search keyword information directly from the unnecessary word deletion unit 7. In other words, the search keyword input to the text search unit 95 is once displayed as a related word to the user as in the configuration of the related word presentation device (FIG. 9) of the third embodiment, and the user arbitrarily selects from the related words. The related word and the keyword information obtained by the processing step of the related word presenting apparatus of the present invention are merged into a new keyword without the operation process of selecting the text search unit 95 directly.
The text search unit 95 is configured to search the analysis information in the analysis information storage unit 1.

The flow of processing in the related word presentation device of the fourth embodiment will be described with reference to FIG. When the user inputs the keyword 10 from the keyword input unit 2, the same processing as that performed by the related word presenting apparatus of the first embodiment is performed, and the unnecessary word deleting unit 7 finally obtains information about the related word for the keyword. Since it is obtained, the unnecessary word deleting unit 7 directly sends the related word information to the text searching unit 95.
When the text search unit 95 obtains the information of the related word, the text search unit 95 searches the analysis information storage unit 1 as new keyword information by merging with the keyword 10, removes the division symbol from the obtained analysis information sentence, and extracts the text information. As search result display section 96
To send to. The search result display unit 96 displays text information as the search result 97.

As described above, according to the related word presentation device of the fourth embodiment, the related information presentation device is integrated with the search system by sharing the analysis information and the search target text, and the functionally overlapping elements are provided. Omitted. As a result, the search system can search for related items that cannot be searched with a keyword, and the recall is improved. Even if the user does not know what to search for (keyword) and cannot input a more appropriate keyword, he / she only needs to input one keyword to perform the target search operation from the related words of the input keyword. It can be performed.

In the third and fourth embodiments, the text search mechanism is added to the first embodiment, but similarly, the text search mechanism may be added to the second embodiment.

[0071]

As described above, according to the related word presentation device of the present invention, the related words are extracted on the spot using the co-occurrence frequency from the information of the morpheme obtained by analyzing the text. Related words can be extracted at high speed without the need for a dictionary such as an origination dictionary. Further, when a dictionary is used, the keywords from which the related words are extracted are limited to the headwords of the dictionary, but in the present invention, the target for extracting the related words holds information in the form of text. So it contains a lot of information,
You can flexibly determine keywords. Further, there is an effect that the related word in the field can be obtained by changing the text of the target for extracting the related word.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a related word presentation device according to a first embodiment of the present invention,

FIG. 2 is a diagram illustrating an example of analysis information stored in an analysis information storage unit,

FIG. 3 is a diagram illustrating a process of obtaining a noun co-occurring with a keyword in the analysis information by the search processing unit,

FIG. 4 is a diagram illustrating a process of registering a co-occurring noun in a frequency table,

FIG. 5 is a diagram showing an example of an unnecessary word list table used by an unnecessary word deleting unit;

FIG. 6 is a block diagram showing the flow of a series of processes of the overall operation of the related word presentation device according to the first embodiment of the present invention;

FIG. 7 is a block diagram showing a configuration of a related word presentation device according to a second embodiment of the present invention,

FIG. 8 is a block diagram showing the flow of a series of processes of the overall operation of the related word presentation device according to the second embodiment of the present invention;

FIG. 9 is a block diagram showing a configuration of a related word presentation device according to a third embodiment of the present invention,

FIG. 10 is a block diagram showing a configuration of a related word presentation device according to a fourth exemplary embodiment of the present invention.

[Explanation of symbols]

1 ... Analysis information storage unit, 2 ... Keyword input unit, 3 ... Search processing unit, 4 ... Co-occurrence word registration unit, 5 ... Frequency table, 6 ... Representative word selection unit, 7 ... Unnecessary word deletion unit, 8 ... Display unit 10 ... Input keyword, 11 ... Output related word, 20 ...
Text data, 21 ... Morphological analysis processing, 22 ... Analysis information, 23 ... Keyword morphological information, 31 ... Search processing,
32 ... 1 sentence of analysis information, 33 ... Index table, 4
1 ... Morphological table, 42 ... Compound word table, 50 ... Unnecessary word list table, 71 ... Text storage section, 72 ... Search analysis section, 90 ... Keywords, 91 ... Search keyword input section, 92 ... Text search section, 93 ... Search Result display,
94 ... Search result, 95 ... Text search section, 96 ... Search result display section, 97 ... Search result.

Claims (4)

[Claims] 1. A related word presentation device for extracting a related word from a text of a document and presenting the related word, the storage means storing the text, a keyword input means for receiving a keyword, and a keyword input means for inputting the keyword. A search processing unit that searches the contents of the storage unit by a keyword and extracts a word that co-occurs in a predetermined range of the keyword and the text; and a noun extracted by the search processing unit together with the frequency of appearance in the text. Co-occurrence word registering means for registering in the table, representative word selecting means for selecting a representative word from the data of the frequency table based on the appearance frequency, and deleting unnecessary words from the representative word selected by the representative word selecting means. A related word characterized by comprising unnecessary word deleting means and display means for displaying a representative word from which unnecessary words are deleted by the unnecessary word deleting means. Display devices. 2. A related word presentation device for extracting a related word from a text of a document and presenting the related word, wherein an analysis information storage section for storing analysis information obtained by performing a morpheme analysis of the text, and a keyword input for accepting a keyword and performing a morpheme analysis. Unit, a search processing unit that searches the analysis information in the analysis information storage unit by a keyword morphologically analyzed by the keyword input unit, and extracts a noun co-occurring in a predetermined range of the keyword and the text; A co-occurrence word registration unit that registers the noun extracted in step 1 in the frequency table together with the appearance frequency in the text, a representative word selection unit that selects a representative word from the data in the frequency table based on the appearance frequency, and the representative word An unnecessary word deletion unit that deletes unnecessary words that have been specified in advance from the representative words selected in the selection unit, and a display unit that displays the representative words from which unnecessary words have been deleted Related word presentation device, characterized in that it comprises a. 3. A related word presentation device for extracting a related word from a text of a document and presenting the related word, wherein a text storage unit for storing the text of the document, a keyword input unit for receiving a keyword and performing a morphological analysis, and a keyword Search the text in the text storage area with
A morphological analysis of the search results is performed to extract nouns that co-occur with the keyword, a co-occurrence word registration unit that registers the extracted nouns in the frequency table together with the frequency, and a representative word is selected from the data in the frequency table. Related word selection unit, an unnecessary word deletion unit that deletes unnecessary words specified in advance from the selected representative word, and a display unit that displays a representative word from which unnecessary words have been deleted. Word presentation device. 4. The related word presentation device according to claim 2, wherein the co-occurrence word registration unit registers the extracted noun morpheme and the morpheme in the appearance frequency table when the appearance frequency is registered in the frequency table. A relation characterized by registering the original compound word of the noun in association with each other, and the representative word selecting unit selecting the representative word from the data of the morpheme and the original compound word according to the frequency registered in the frequency table. Word presentation device.