JP5542729B2 - Related word extraction device, related word extraction method, and related word extraction program - Google Patents

Description

  The present invention relates to a related word extraction technique for extracting related words for a certain word.

  There are a lot of electronic texts with various contents in the world. An information retrieval system is used to search for desired information. In this system, if a user inputs a search word related to desired information, electronic text information related to the search word can be obtained. Considering the set of search terms, the search intentions of many people are reflected, so system providers can analyze the search term set to expand the search target that people need and It can be used to improve access methods. It can also be viewed as a reflection of the common interests of people, which is useful for marketing analysis.

  As a prior art, in an information search system, a word having a strong relationship with each other is used for obtaining the same information by obtaining the strength of the degree of association between search terms used within a predetermined period. Considering, for example, an information associating device for associating search terms that are temporarily strongly related, such as “New Year's card” and “winning number” at the beginning of the year, is disclosed (see Patent Document 1).

  In addition, there is a conventional technique in which association is performed based on the co-occurrence frequency of words in a query from a query log of an information search system. For example, related words such as “lunch” and “movie theater” can be obtained by arranging words in the descending order of the co-occurrence frequency of words (co-occurrence words) that appear simultaneously with “Ginza”.

Japanese Patent No. 3547069

  However, for example, the meaning of “one piece” may be treated as a different meaning depending on the background used, such as the type of clothes or the title of an animation. When related words for such a word are extracted, related words having different meanings cannot be distinguished by the prior art.

  For example, in order to limit “one piece” to the meaning of clothes, “clothes” is added to “one piece”, and a word that co-occurs with any word can be considered as a related word. However, since all words must co-occur, even if there is a word with clothes meaning that co-occurs only in "one piece", it cannot be extracted as a related word.

  Conversely, for example, to limit "one piece" to the meaning of clothes, add "-anime" to "one piece" and co-occur with "one piece", but the word does not co-occur with "anime" Can be considered as a related term. However, even if there is a word with an anime meaning that co-occurs only in "One Piece", it cannot be excluded from related words.

  The present invention solves the above-described problems, and in order to extract related words having a specific meaning from a target word having an ambiguous meaning, the co-occurrence words of the target word are grouped and related to other words. Using the co-occurrence frequency of deep co-occurrence words, add or subtract the support level of the co-occurrence words of the target word, select a group with high support level, and target the co-occurrence words belonging to that group It is an object of the present invention to provide a related word extraction technique that outputs related words having a specific meaning that is closely related to other words by using the related words of the word.

In order to solve the above-mentioned problem, the present invention is a related word extraction device for extracting a related word for a word,
A co-occurrence word data creation means for inputting a text set, obtaining a co-occurrence frequency of words co-occurring with words in the text, and outputting the co-occurrence word data storage means;
Using the co-occurrence word data stored in the co-occurrence word data storage means, for each word of a predetermined word set, the words that co-occur with the word are obtained and grouped, and the result is output to the word group storage means Word group creation means to
Related word group extraction means for extracting group data for the input target word from the word group data stored in the word group storage means, and outputting to the related word group data storage means;
From the word group data, support word group extraction means for extracting group data for each support word described in the input support word list and outputting to the support word group data storage means;
Co-occurrence frequency of support co-occurrence words and support words, which are co-occurrence words belonging to a group closely related to the support word, from the support word list and the support word group data stored in the support word group data storage means And calculating a co-occurrence frequency for all the support words, and outputting a count result to the support co-occurrence word data storage means;
From the related word group data stored in the related word group data storage means and the support co-occurrence word data stored in the support co-occurrence word data storage means, the co-occurrence words belonging to each related word group of the target word, Obtaining a support level of matching support co-occurrence words, totaling for each related word group of the target word, selecting a co-occurrence word closely related to the support co-occurrence word as a related word, and outputting a related word extracting means; It is comprised as a related word extraction apparatus characterized by having provided.

Further, the present invention is the related word extraction device in a related word extraction system having a related word extraction device for extracting a related word for a word and an external device,
The external device includes co-occurrence word data creating means for inputting a text set, obtaining a co-occurrence frequency of a word that co-occurs with a word in the text, and outputting as co-occurrence word data,
The related word extraction device comprises:
Co-occurrence word data storage means for storing co-occurrence word data created by the external device;
Using the co-occurrence word data stored in the co-occurrence word data storage means, for each word of a predetermined word set, the words that co-occur with the word are obtained and grouped, and the result is output to the word group storage means Word group creation means to
Related word group extraction means for extracting group data for the input target word from the word group data stored in the word group storage means, and outputting to the related word group data storage means;
From the word group data, support word group extraction means for extracting group data for each support word described in the input support word list and outputting to the support word group data storage means;
Co-occurrence frequency of support co-occurrence words and support words, which are co-occurrence words belonging to a group closely related to the support word, from the support word list and the support word group data stored in the support word group data storage means And calculating a co-occurrence frequency for all the support words, and outputting a count result to the support co-occurrence word data storage means;
From the related word group data stored in the related word group data storage means and the support co-occurrence word data stored in the support co-occurrence word data storage means, the co-occurrence words belonging to each related word group of the target word, Obtaining a support level of matching support co-occurrence words, totaling for each related word group of the target word, selecting a co-occurrence word closely related to the support co-occurrence word as a related word, and outputting a related word extracting means; It can also be comprised as a related word extraction apparatus characterized by providing.

Further, the present invention is the related word extraction device in a related word extraction system having a related word extraction device for extracting a related word for a word and an external device,
The external device inputs a text set, obtains a co-occurrence frequency of a word co-occurring with a word in the text, and outputs the co-occurrence word data generation means to the co-occurrence word data storage means, and the co-occurrence word data Word group creating means for using the co-occurrence word data stored in the storage means to group and search for words that co-occur with the word for each word of a predetermined word set, and outputting the result as word group data; With
The related word extraction device comprises:
Word group storage means for storing word group data created by the external device;
Related word group extraction means for extracting group data for the input target word from the word group data stored in the word group storage means, and outputting to the related word group data storage means;
From the word group data, support word group extraction means for extracting group data for each support word described in the input support word list and outputting to the support word group data storage means;
Co-occurrence frequency of support co-occurrence words and support words, which are co-occurrence words belonging to a group closely related to the support word, from the support word list and the support word group data stored in the support word group data storage means And calculating a co-occurrence frequency for all the support words, and outputting a count result to the support co-occurrence word data storage means;
From the related word group data stored in the related word group data storage means and the support co-occurrence word data stored in the support co-occurrence word data storage means, the co-occurrence words belonging to each related word group of the target word, Obtaining a support level of matching support co-occurrence words, totaling for each related word group of the target word, selecting a co-occurrence word closely related to the support co-occurrence word as a related word, and outputting a related word extracting means; You may comprise as a related word extraction apparatus characterized by providing.

  In the support co-occurrence word extraction means, information for determining a support method of selection or exclusion for the support words in the support word list is written, and on the basis of the information, for all support words When counting the co-occurrence frequencies, addition may be performed in the support method to be selected, and subtraction may be performed in the support method to be excluded.

  In addition, the target word is input in the form of a search expression, for example. In that case, the related word group extraction means extracts group data for each word included in the search expression and outputs it to the related word group data storage means. Further, in the related word extraction means, for each word included in the search formula, the support level of the support co-occurrence word that matches the co-occurrence word belonging to each related word group is obtained, and the related word group of the target word A co-occurrence word closely related to the supporting co-occurrence word is extracted, and a co-occurrence word that satisfies the search expression condition is selected as a related word.

  Further, the support word list is input in the form of, for example, a search expression. In this case, the support word group extraction unit searches for group data for each word included in the search expression and outputs the group data to the support word group data storage unit. The supporting co-occurrence word extracting unit selects a co-occurrence word that satisfies the search expression condition as a supporting co-occurrence word.

  In the related word extraction device, the processing of the word group creation means is performed in the related word group extraction means and / or the support word group extraction means, so that the target word and the support word list are sequentially input after the input. You may comprise so that a process may be performed and the word group data of the said target word and the said support word may be produced.

  Moreover, you may comprise this invention as a related word extraction method which the said related word extraction apparatus performs. Furthermore, this invention may comprise a computer as a related word extraction program for functioning as each means in the said related word extraction apparatus.

  According to the present invention, in order to extract a related word having a specific meaning from a target word having an ambiguous meaning, the co-occurrence words of the target word are grouped together and the co-occurrence words closely related to other words are co-occurred. By adding and subtracting the support level of the co-occurrence word of the target word, selecting a group with a high support level, and making the co-occurrence word belonging to that group the related word of the target word, A related word having a specific meaning closely related to another word can be extracted with high accuracy.

It is a block diagram of the related word extraction apparatus which concerns on one Embodiment of this invention. It is a flowchart showing the flow of a process of the related word extraction apparatus shown in FIG. It is an example (1) of various data. It is an example (2) of various data. It is an example (3) of various data.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings, but the present invention is not limited to the following embodiments.

(Outline of device configuration and operation)
FIG. 1 is a configuration diagram of a related word extraction device 100 according to an embodiment of the present invention, and FIG. 2 is a flowchart showing a process flow of the related word extraction device 100 of FIG. S100 to S150 in FIG. 2 indicate the steps of each process.

  As shown in FIG. 1, the related word extraction apparatus 100 according to the present embodiment includes a co-occurrence word data creation unit 110, a word group creation unit 120, a related word group extraction unit 130, a support word group extraction unit 140, and a support co-occurrence word. An extraction unit 150, a related word extraction unit 160, a co-occurrence word database 170, a word group database 180, a related word group database 190, a support word group database 200, and a support co-occurrence word database 210 are provided. The related word extraction apparatus 100 receives the text set 300, the target word 400, and the support word list 500 as inputs, and outputs the related words extracted by the related word extraction unit 160 to the related word database 600.

  The related word extraction apparatus 100 can be realized by causing a computer to execute a program describing the processing contents described in the present embodiment. That is, with respect to all or a part of the functional units of the related word extraction apparatus 100, the functions of the respective units are implemented in the respective units using hardware resources such as a CPU and a memory built in the computer constituting the apparatus. This can be realized by executing a program corresponding to the processing to be performed. Each database in the related word extraction device 100 is realized by storage means such as a memory. In addition, the program is a computer-readable recording medium such as FD (Floppy (registered trademark) Disk), MO (Magneto-Optical disk), ROM (Read Only Memory), memory card, CD (Compact Disk)- ROM, DVD (Digital Versatile Disk) -ROM, BD (Blu-ray Disk) -ROM, CD-R, CD-RW, DVD-R, DVD-RW, BD-R, BD-RE, HDD, removable disk, etc. It can be recorded and stored or distributed. It is also possible to provide the program through a network such as the Internet or electronic mail.

  Next, an outline of the operation of the related word extracting apparatus 100 will be described with reference to the flowchart of FIG.

  Step 100) The co-occurrence word data creation unit 110 inputs the text set 300, obtains the co-occurrence frequency of the words in the text and the co-occurrence words, and outputs them to the co-occurrence word database 170.

  Step 110) Using the co-occurrence words stored in the co-occurrence word database 170, the word group creation unit 120 obtains a word that co-occurs with the word for each word in a predetermined word set, and groups the results. Is output to the word group database 180.

  Step 120) The related word group extraction unit 130 searches the word group database 180 for group data for the target word 400 and outputs it to the related word group database 190.

  Step 130) The support word group extraction unit 140 searches the word group database 180 for group data for each support word described in the support word list 400, and outputs it to the support word group database 200.

  Step 140) The support co-occurrence word extraction unit 150 extracts the co-occurrence words (support co-occurrence words) and support words belonging to a group closely related to the support words from the support word list 500 and the support word group database 200. The occurrence frequency is obtained, and the co-occurrence frequencies are totaled for all the support words, and are output to the support co-occurrence word database 210.

  Step 150) The related word extraction unit 160 uses the data stored in the related word group database 190 and the support co-occurrence word database 210 to support co-occurrence words that match the co-occurrence words belonging to each related word group of the target word. The co-occurrence words closely related to the supporting co-occurrence words are selected as related words and output to the related word database 600.

(Operation details)
Next, a series of operations of the related word extraction device 100 in the present embodiment will be described in more detail using specific examples. Note that the specific example used in the following description is merely an example.

  The co-occurrence word data creation unit 110 inputs the text set 300, obtains the co-occurrence frequency of the words in the text and the co-occurrence words, and outputs them to the co-occurrence word database 170. For example, FIG. 3A shows a text set when the search query log of the information search system is a text set. The text set in FIG. 3A is a text in which one line represents one query. FIG. 3B shows a set of words that co-occur with a certain word in the text from this text set, the number of the sets (co-occurrence frequency) is totaled, and data arranged in order of frequency is created. Of co-occurrence word data. Even if the text set is a normal sentence, it can be processed in the same way by taking out independent words by morphological analysis and treating one sentence or one clause as one text.

  The word group creation unit 120 uses the co-occurrence word data stored in the co-occurrence word database 170 to find and group words that co-occur with the word for each word in a predetermined word set, and the result is a word Output to the group database 180.

  For example, first, for a predetermined word set, the top N items of the co-occurrence word count data of FIG. 3C arranged in descending order of the number of co-occurrence words for each word of the co-occurrence word data of FIG. It is determined. In addition, as another method, the top N items are determined from those having a large value calculated by the function based on the frequency of words in the text set, or the word set described in the word list prepared in advance is set as a predetermined word set. May be.

  Next, processing for each word in a predetermined word set will be described. For example, if the predetermined word is “one piece”, the co-occurrence word of “one piece” is obtained from the co-occurrence word data of FIG. 3B, and the co-occurrence word data of FIG. At this time, since the co-occurrence frequency of the word that co-occurs with “One Piece” is also known from FIG. 3B, those having a co-occurrence frequency less than a predetermined threshold are not extracted.

  Next, in order to represent the characteristics of each co-occurrence word of the co-occurrence word data in FIG. 3D, the co-occurrence word and the co-occurrence frequency for each co-occurrence word are shown in FIG. The feature vector data shown in FIG. 3E having the co-occurrence frequency for the obtained word as an element is created. Since the feature vector only needs to represent the feature of the word based on the co-occurrence frequency, a value obtained by applying an arbitrary function to the value may be used instead of the co-occurrence frequency. Finally, as a result of clustering “one-piece” co-occurrence words using the created feature vectors, group data consisting of four groups is obtained as shown in FIG. 3F and output to the word group database 170. The clustering method is not limited as long as it uses feature vectors. The number of clusters is a predetermined number. For example, FIG. 3F shows the result of processing with 4 clusters. The above is the processing for each word in the predetermined word set, and all the words in the predetermined word set are processed in the same way.

  The related word group extraction unit 130 searches the word group database 180 for group data for the target word 400 and outputs it to the related word group database 190. For example, when the target word is “one piece”, the group data of FIG. 3F whose group data name is “one piece” is searched from the word group data and output to the related word group database 180.

  The support word group extraction unit 140 searches the word group database 180 for group data for each support word described in the support word list 400 and outputs the group data to the support word group database 200. For example, if the support word list is as shown in FIG. 4G, this support word list is intended to select a co-occurrence word closely related to “clothes”, and the word “−” "-" With "animation" is intended to exclude co-occurrence words closely related to "anime". In this case, the group data shown in FIG. 4 (H) and FIG. 4 (I), which are the group data of the support words “clothes” and “animation”, are found from the group data name from the word group database. Output to database 200.

  The support co-occurrence word extraction unit 150 determines the co-occurrence frequency of co-occurrence words (support co-occurrence words) and support words belonging to a group closely related to the support word from the support word list 500 and the support word group database 200. The co-occurrence frequencies are calculated for all support words and output to the support co-occurrence word database 210. For example, for “clothes” described in the support word list shown in FIG. 4G, the support word group database 200 includes group data of “clothes” in FIG. 4H. Extract only co-occurrence groups that are closely related to "clothes". For each group, when the co-occurrence frequency of co-occurrence words and support words included in the group is aggregated for each group, and the cumulative value of the overall ratio is calculated in descending order of the co-occurrence frequency of each group, The corresponding group when the predetermined threshold is reached for the first time.

  For example, since the numerical value in parentheses described later of the co-occurrence word of each group in FIG. 4H is the co-occurrence frequency of the support word “clothes” and each word, the co-occurrence frequency is aggregated for each group ( When the total ratio and the accumulated value are obtained, the group total data of “clothes” in FIG. Note that the group numbers in FIGS. 4H and 4J correspond to each other. When the predetermined threshold value is 0.9, the group number 1 to 3 becomes the corresponding group because the total of the co-occurrence frequency of each group becomes equal to or greater than the threshold value when the cumulative ratio becomes 0.99. Then, when the co-occurrence frequency of the co-occurrence words belonging to the group numbers 1 to 3 and the support word is obtained from the numerical value in parentheses described later in the co-occurrence words of the group data in FIG. 4 (H), FIG. Is word data closely related to "clothes". Similarly, there is FIG. 4 (I) as the group data of the support word “anime” excluding “-” for “-animation” described in the support word list, and FIG. ), The co-occurrence word group closely related to “animation” from the group total data of “anime” becomes the threshold or more for the first time when the total of the co-occurrence frequency of each group becomes the threshold value 0.94. Therefore, group number 1 is the corresponding group. Then, when the co-occurrence frequency of the co-occurrence word belonging to the group number 1 and the support word is obtained from the group data of FIG. 4 (I), FIG. 5 (M) becomes word data closely related to “animation”. The co-occurrence words belonging to the group closely related to the support word were obtained from the co-occurrence words belonging to the corresponding group when the cumulative total of the co-occurrence frequency of each group first reached the threshold. If it is obtained, it is not limited to this method.

  The same processing is performed for all the words described in the support word list, and finally, word data closely related to the support word is totaled and output to the support word group database 210. For example, “clothes” described in the support word list is used to select words closely related to the support word, so the corresponding FIG. 4 (K) is added, and “-animation” described in the support word list is the head. Since “-” is used, it is used to exclude words closely related to the support word. Therefore, FIG. 5 (N) corresponding to “animation” excluding “-” is subtracted and tabulated. ) Is output to the supporting co-occurrence word group database 210.

  The related word extraction unit 160 obtains the support level of the support co-occurrence word that matches the co-occurrence word belonging to each related word group of the target word from the related word group database 190 and the support co-occurrence word database 210, and The words are aggregated for each word group, and co-occurrence words closely related to the support word are selected as related words and output to the related word database 600.

  For example, for the co-occurrence word “mail order” of the group number 4 of the “one piece” group data in FIG. 3F in the related word group database 180, the support in FIG. 5N in the support co-occurrence word database 210. It can be seen that word data having a deep relationship with words has a co-occurrence frequency of “mail order” of 1200. The co-occurrence frequency is added to group number 4 in FIG. The co-occurrence frequencies are obtained from FIG. 5 (N) for all co-occurrence words in the group data of “One Piece” in FIG. 3 (F), and are counted for each group in FIG. 3 (F), in descending order of support. As a result of arranging, the result of FIG. 5 (O) is obtained. Next, the corresponding group when the threshold value reaches a predetermined value or more for the first time is selected from the accumulation of the overall ratios determined in descending order of support in FIG. The overall ratio when the support level is negative is 0. When the predetermined threshold value is 0.9, the cumulative value of the whole ratio in FIG. 5 (O) becomes the threshold value or more for the first time when it is 1.0, so group numbers 3 and 4 become the corresponding groups and are included in that group. Are output to the related word database 600 as related word data.

  This output word becomes a related word of the target word whose meaning is specified in a word closely related to the support word. The co-occurrence words belonging to the group closely related to the support word were obtained from the co-occurrence words belonging to the corresponding group when the cumulative total of the co-occurrence frequency of each group first reached the threshold. If it is obtained, it is not limited to this method.

  In the present embodiment, for example, when it is desired to extract only related words related to clothes from the target word “one piece”, “one piece brand name in the meaning of clothes” is a co-occurrence word of “one piece”. If so, group the co-occurrence words so that the word is included in the group related to clothes. On the other hand, even if the support word “clothes” to be selected and “one-piece brand name in the meaning of clothes” do not co-occur, clothes that have a co-occurrence word “one piece” closely related to “clothes” Because it supports groups related to, "one-piece brand name in the meaning of clothes" can be extracted as a related term.

  In the present embodiment, for example, when it is desired to extract only related words related to clothes from the target word “One Piece”, there is “Character Name in Animation” which is a co-occurrence word of “One Piece”. If a co-occurrence word is grouped, the word is included in the group related to the animation. On the other hand, the co-occurrence word closely related to “Anime” is related to “One Piece” anime, even when “Anime” and “Character Name in Anime” are not co-occurring. In order not to support the group, “name of character in animation” is excluded from the related words.

  As another method different from the present embodiment, the related word group database 190 created by the related group extracting unit 130 in FIG. 1 is directly obtained for a cluster including the support word to be selected, and belongs to the cluster. A method of selecting a search term as a related term can be considered. In this method, for example, if the support word is “clothes”, and if “clothes” is included in the group number 4 of the one-piece group data in FIG. Although it can be extracted as a related word, a co-occurrence word related to clothes such as “fashion” cannot be extracted in group number 3 as well. On the other hand, according to the present embodiment, it is possible to extract both group numbers 3 and 4 by using co-occurrence words of support words, instead of directly including or not including support words.

  In the above embodiment, the target word 400 is one word in the related word extraction apparatus 100 of FIG. 1, but the target word 400 may be expressed by a search expression, for example, “(one piece OR suit)” If "AND NOT Spring" is specified, a co-occurrence word of "one piece" or a co-occurrence word of "suit", and excluding the co-occurrence word of "spring" is extracted as a related word. Specifically, the related word group extraction unit 130 in FIG. 1 searches the word group database 180 for “One Piece”, “suit”, and “spring” group data, stores them in the related word group data 190, and stores related words. In the extraction unit 160, group total data of “one piece”, “suit”, and “spring” is created, and the co-occurrence words of the corresponding group when the accumulation of the whole ratio exceeds the threshold for the first time are respectively extracted. Co-occurrence words that are included in either “One Piece” or “Suit” and are not included in “Spring” may be extracted as related words.

  In the above embodiment, in the related word extraction apparatus 100 of FIG. 1, the support word list 500 is a list of words, but the support word list 500 may be expressed by a search expression. For example, if "(clothes OR clothes) AND NOT anime" is specified, it is a co-occurrence word of "clothes" or "clothes", and excluding the co-occurrence word of "anime" is extracted as a supporting co-occurrence word Is done. Specifically, the support word group extraction unit 140 of FIG. 1 searches the word group database 180 for “clothes”, “clothes”, and “animation” group data, stores them in the support word group data 200, and supports co-occurrence. The word extraction unit 150 creates group total data of “clothes”, “clothes”, and “animation”, extracts the co-occurrence words of the corresponding group when the total proportion exceeds the threshold for the first time, and finally Co-occurrence words that are included in either “clothes” or “clothes” and that are not included in “animation” may be processed as support words.

  In the above embodiment, the word group creation unit 120 in FIG. 1 creates word group data for an arbitrary word in advance using the co-occurrence word database 170. Is performed in the related word group extraction unit 130 and the support word group extraction unit 140 (or any one of the related word group extraction unit 130 and the support word group extraction unit 140). It is also possible to create word group data of the target word and supporting word by sequentially processing after input. At this time, the number of clusters at the time of clustering may be changed for each word or the difference between the target words and the support words.

  In the above embodiment, the related word extraction apparatus 100 in FIG. 1 performs processing by inputting the text set 300. However, the co-occurrence word data creation unit 110 and the word group creation unit 120 are connected to the external device. It may be realized as a function, and the data of the co-occurrence word database 170 and the data of the word group database 180 may be created by processing of an external device and processed as input.

  That is, for example, an external device implemented by a computer inputs a text set, obtains the co-occurrence frequency of words that co-occur with words in the text, and outputs the co-occurrence word data creation unit 110 that outputs the co-occurrence word data. Is provided. Then, the related word extraction device 100 uses co-occurrence word data storage means (database) for storing the co-occurrence word data created by the external device, and the co-occurrence word data stored in the co-occurrence word data storage means. Then, for each word of a predetermined word set, a word group creation unit 120 that obtains a word that co-occurs with the word and groups the result and outputs the result to the word group storage means, and the word group storage means The related word group extraction unit 130 extracts group data for the input target word from the word group data, and outputs the extracted group data to the related word group data storage unit. The support described in the support word list input from the word group data A support word group extraction unit 140 that extracts group data for each word and outputs it to the support word group data storage means; From the support word list and the support word group data stored in the support word group data storage means, the co-occurrence frequency of a support co-occurrence word and a support word that are co-occurrence words belonging to a group closely related to the support word is determined. The co-occurrence frequency is calculated for all support words, and the co-occurrence word extraction unit 150 outputs the total result to the support co-occurrence word data storage means, and is stored in the related word group data storage means. The support co-occurrence word matching the co-occurrence word belonging to each related word group of the target word is determined from the related word group data and the support co-occurrence word data stored in the support co-occurrence word data storage means. A related word extraction unit 160 that calculates and collects the related word groups of the target words, selects the co-occurrence words closely related to the supporting co-occurrence words, and outputs them as related words;

  As another example, the external device inputs a text set, obtains the co-occurrence frequency of the words in the text and the words that co-occur, and outputs them to the co-occurrence word data storage unit 110. And using the co-occurrence word data stored in the co-occurrence word data storage means, for each word in a predetermined word set, the words that co-occur with the word are obtained and grouped, and the result is used as word group data. And an output word group creation unit 120. Then, the related word extracting apparatus 100 extracts the group data for the input target word from the word group storage means for storing the word group data and the word group data stored in the word group storage means, and the related words Related word group extraction unit 130 for outputting to group data storage means, and support for extracting group data for each support word described in the input support word list from the word group data and outputting to support word group data storage means A word group extraction unit 140, the support word list, and a support co-occurrence word that is a co-occurrence word belonging to a group closely related to the support word from the support word group data stored in the support word group data storage unit; The co-occurrence frequency with support words is obtained, the co-occurrence frequencies are totaled for all support words, and the result of the count is the support co-occurrence word Support co-occurrence word extraction unit 150 for outputting to the data storage means, related word group data stored in the related word group data storage means, and support co-occurrence word data stored in the support co-occurrence word data storage means The support co-occurrence words that match the co-occurrence words belonging to each related word group of the target word are obtained, and the support word of the co-occurrence word that is closely related to the support co-occurrence word is calculated for each related word group of the target word. And a related word extraction unit 160 that selects and outputs the related words.

  In these cases, for example, an external device and the related word extraction device 100 are connected via a communication network, and data created by the external device is input to the related word extraction device 100 via the communication network. You may do it.

  The present invention is not limited to the above-described embodiments, and various modifications and applications are possible within the scope of the claims.

DESCRIPTION OF SYMBOLS 100 ... Related word extraction apparatus 110 ... Co-occurrence word data creation part 120 ... Word group creation part 130 ... Related word group extraction part 140 ... Support word group extraction part 150 ... Related word extraction part 160 ... Co-occurrence word database 170 ... Word group Database 180 ... Related word group database 190 ... Supported word group database 300 ... Text set 400 ... Target word 500 ... Supported word list 600 ... Related word database

Claims (15)

A related word extraction device for extracting related words for a word,
A co-occurrence word data creation means for inputting a text set, obtaining a co-occurrence frequency of words co-occurring with words in the text, and outputting the co-occurrence word data storage means;
Using the co-occurrence word data stored in the co-occurrence word data storage means, for each word of a predetermined word set, the words that co-occur with the word are obtained and grouped, and the result is output to the word group storage means Word group creation means to
Related word group extraction means for extracting group data for the input target word from the word group data stored in the word group storage means, and outputting to the related word group data storage means;
From the word group data, support word group extraction means for extracting group data for each support word described in the input support word list and outputting to the support word group data storage means;
Co-occurrence frequency of support co-occurrence words and support words, which are co-occurrence words belonging to a group closely related to the support word, from the support word list and the support word group data stored in the support word group data storage means And calculating a co-occurrence frequency for all the support words, and outputting a count result to the support co-occurrence word data storage means;
From the related word group data stored in the related word group data storage means and the support co-occurrence word data stored in the support co-occurrence word data storage means, the co-occurrence words belonging to each related word group of the target word, It is provided with a related word extraction means that obtains the support level of matching co-occurrence words, aggregates them for each related word group of the target word, selects a co-occurrence word closely related to the support co-occurrence word as a related word, and outputs it. A related word extraction device characterized by that. The related word extraction device in a related word extraction system having a related word extraction device for extracting a related word for a word and an external device,
The external device is
A co-occurrence word data creation means for inputting a text set, obtaining a co-occurrence frequency of a word in the text and a co-occurrence word, and outputting the co-occurrence word data,
The related word extraction device comprises:
Co-occurrence word data storage means for storing co-occurrence word data created by the external device;
Using the co-occurrence word data stored in the co-occurrence word data storage means, for each word of a predetermined word set, the words that co-occur with the word are obtained and grouped, and the result is output to the word group storage means Word group creation means to
Related word group extraction means for extracting group data for the input target word from the word group data stored in the word group storage means, and outputting to the related word group data storage means;
From the word group data, support word group extraction means for extracting group data for each support word described in the input support word list and outputting to the support word group data storage means;
Co-occurrence frequency of support co-occurrence words and support words, which are co-occurrence words belonging to a group closely related to the support word, from the support word list and the support word group data stored in the support word group data storage means And calculating a co-occurrence frequency for all the support words, and outputting a count result to the support co-occurrence word data storage means;
From the related word group data stored in the related word group data storage means and the support co-occurrence word data stored in the support co-occurrence word data storage means, the co-occurrence words belonging to each related word group of the target word, A related word extraction unit that obtains support levels of matching co-occurrence words, aggregates them for each related word group of the target word, selects a co-occurrence word closely related to the support co-occurrence word as a related word, and outputs it. The related word extraction device characterized by this. The related word extraction device in a related word extraction system having a related word extraction device for extracting a related word for a word and an external device,
The external device is
A co-occurrence word data creation means for inputting a text set, obtaining a co-occurrence frequency of words co-occurring with words in the text, and outputting the co-occurrence word data storage means;
Using the co-occurrence word data stored in the co-occurrence word data storage means, for each word in a predetermined word set, a word that co-occurs with the word is obtained and grouped, and the result is output as word group data. A word group creation means,
The related word extraction device comprises:
Word group storage means for storing word group data created by the external device;
Related word group extraction means for extracting group data for the input target word from the word group data stored in the word group storage means, and outputting to the related word group data storage means;
From the word group data, support word group extraction means for extracting group data for each support word described in the input support word list and outputting to the support word group data storage means;
Co-occurrence frequency of support co-occurrence words and support words, which are co-occurrence words belonging to a group closely related to the support word, from the support word list and the support word group data stored in the support word group data storage means And calculating a co-occurrence frequency for all the support words, and outputting a count result to the support co-occurrence word data storage means;
From the related word group data stored in the related word group data storage means and the support co-occurrence word data stored in the support co-occurrence word data storage means, the co-occurrence words belonging to each related word group of the target word, A related word extraction unit that obtains support levels of matching co-occurrence words, aggregates them for each related word group of the target word, selects a co-occurrence word closely related to the support co-occurrence word as a related word, and outputs it. The related word extraction device characterized by this. In the support co-occurrence word extraction means, information for determining a support method of selection or exclusion for the support words in the support word list is written, and on the basis of the information, for all support words The related word according to any one of claims 1 to 3, wherein, when the co-occurrence frequencies are aggregated, addition is performed in the support method to be selected and subtraction is performed in the support method to be excluded. Extraction device. The target word is input in the form of a search expression, and the related word group extraction unit extracts group data for each word included in the search expression, and outputs the group data to the related word group data storage unit. In the word extraction means, for each word included in the search formula, the support level of the support co-occurrence word that matches the co-occurrence word belonging to each related word group is obtained, and the support level is calculated for each related word group of the target word and supported. 5. The co-occurrence word closely related to the co-occurrence word is extracted, and the co-occurrence word satisfying the condition of the search expression is selected as a related word. 5. Related word extraction device. The support word list is input in the form of a search expression, and the support word group extraction unit searches for group data for each word included in the search expression, and outputs the group data to the support word group data storage unit. 6. The related word extraction device according to claim 1, wherein a co-occurrence word extracting unit selects a co-occurrence word that satisfies the search expression condition as a support co-occurrence word. By performing the processing of the word group creating means in the related word group extracting means and / or the supporting word group extracting means, the processing is sequentially performed after inputting the target word and the supporting word list, and the target word The word group data of the said support word is created. The related word extraction apparatus of any one of Claim 1 thru | or 6 characterized by the above-mentioned. A related word extraction method executed by a related word extraction device that extracts related words for a word,
A co-occurrence word data creation step of inputting a text set, obtaining a co-occurrence frequency of a word co-occurring with a word in the text, and outputting the co-occurrence word data storage means;
Using the co-occurrence word data stored in the co-occurrence word data storage means, for each word of a predetermined word set, the words that co-occur with the word are obtained and grouped, and the result is output to the word group storage means A word group creation step,
A related word group extracting step of extracting group data for the input target word from the word group data stored in the word group storage means, and outputting to the related word group data storage means;
Extracting group data for each support word described in the input support word list from the word group data and outputting to the support word group data storage means, a support word group extraction step;
Co-occurrence frequency of support co-occurrence words and support words, which are co-occurrence words belonging to a group closely related to the support word, from the support word list and the support word group data stored in the support word group data storage means And calculating a co-occurrence frequency for all the support words, and outputting a count result to the support co-occurrence word data storage means;
From the related word group data stored in the related word group data storage means and the support co-occurrence word data stored in the support co-occurrence word data storage means, the co-occurrence words belonging to each related word group of the target word, A support word extraction step for obtaining a support level of matching support co-occurrence words, totaling each related word group of the target word, selecting a co-occurrence word closely related to the support co-occurrence word as a related word, and outputting it. A related word extraction method characterized by that. A related word extraction method executed by the related word extraction device in a related word extraction system having a related word extraction device for extracting a related word for a word and an external device,
The external device is
A co-occurrence word data creation means for inputting a text set, obtaining a co-occurrence frequency of a word in the text and a co-occurrence word, and outputting the co-occurrence word data,
The related word extraction method is:
Using the co-occurrence word data created by the external device, for each word in a predetermined word set, a word that co-occurs with the word is obtained and grouped, and the result is output to the word group storage means Steps,
A related word group extracting step of extracting group data for the input target word from the word group data stored in the word group storage means, and outputting to the related word group data storage means;
Extracting group data for each support word described in the input support word list from the word group data and outputting to the support word group data storage means, a support word group extraction step;
Co-occurrence frequency of support co-occurrence words and support words, which are co-occurrence words belonging to a group closely related to the support word, from the support word list and the support word group data stored in the support word group data storage means And calculating a co-occurrence frequency for all the support words, and outputting a count result to the support co-occurrence word data storage means;
From the related word group data stored in the related word group data storage means and the support co-occurrence word data stored in the support co-occurrence word data storage means, the co-occurrence words belonging to each related word group of the target word, A support word extraction step for obtaining a support level of matching support co-occurrence words, totaling by each related word group of the target word, selecting a co-occurrence word closely related to the support co-occurrence word as a related word, and outputting it; The related word extraction method characterized by this. A related word extraction method executed by the related word extraction device in a related word extraction system having a related word extraction device for extracting a related word for a word and an external device,
The external device is
A co-occurrence word data creation means for inputting a text set, obtaining a co-occurrence frequency of words co-occurring with words in the text, and outputting the co-occurrence word data storage means;
Using the co-occurrence word data stored in the co-occurrence word data storage means, for each word in a predetermined word set, a word that co-occurs with the word is obtained and grouped, and the result is output as word group data. A word group creation means,
The related word extraction method is:
A related word group extracting step of extracting group data for the input target word from the word group data created by the external device and outputting it to the related word group data storage means;
Extracting group data for each support word described in the input support word list from the word group data and outputting to the support word group data storage means, a support word group extraction step;
Co-occurrence frequency of support co-occurrence words and support words, which are co-occurrence words belonging to a group closely related to the support word, from the support word list and the support word group data stored in the support word group data storage means And calculating a co-occurrence frequency for all the support words, and outputting a count result to the support co-occurrence word data storage means;
From the related word group data stored in the related word group data storage means and the support co-occurrence word data stored in the support co-occurrence word data storage means, the co-occurrence words belonging to each related word group of the target word, A support word extraction step for obtaining a support level of matching support co-occurrence words, totaling by each related word group of the target word, selecting a co-occurrence word closely related to the support co-occurrence word as a related word, and outputting it; The related word extraction method characterized by this. In the support co-occurrence word extraction step, information for determining a support method of selection or exclusion with respect to the support words of the support word list is written, and the related word extraction device is based on the information, 11. The total number of co-occurrence frequencies for the support words is added when the support method is selected, and is subtracted when the support method is excluded. The related word extraction method according to Item 1. The target word is input in the form of a search expression, and in the related word group extraction step, the related word extraction device extracts group data for each word included in the search expression and supplies the related word group data storage means. And in the related word extracting step, the related word extracting device determines the support level of the support co-occurrence word that matches the co-occurrence word belonging to each related word group for each word included in the search expression. It is obtained by calculating for each related word group of target words, extracting co-occurrence words closely related to supporting co-occurrence words, and selecting co-occurrence words satisfying the conditions of the search expression as related words. The related word extraction method according to any one of claims 8 to 11. The support word list is input in the form of a search expression, and in the support word group extraction step, the related word extraction device searches for group data for each word included in the search expression and sends it to the support word group data storage means. 13. The output of the supporting co-occurrence word, the related word extraction device selects a co-occurrence word satisfying the search expression condition as a supporting co-occurrence word. The related word extraction method of any one of these. By performing the process of creating the word group in the related word group extraction step and / or the support word group extraction step, the processing is sequentially performed after inputting the target word and the support word list, and the target word or The related word extraction method according to any one of claims 8 to 13, wherein word group data of the support word is created.   The related word extraction program for functioning a computer as each means of the related word extraction apparatus of any one of Claims 1 thru | or 7.

Images