JP5964149B2 - Apparatus and program for identifying co-occurrence words - Google Patents

Description

  The present invention relates to a technique for specifying a co-occurrence word that is a word that appears in the same sentence together with a keyword indicating information of interest to a user.

  Information retrieval is widely performed via the Internet. For example, a user who searches for information inputs a word of interest as a search keyword on a search screen of a Web browser operating on a terminal device, and transmits it to a server device having a search engine function. The server device extracts a Web page including a search keyword transmitted from the terminal device from a large number of Web pages published on the Internet according to a predetermined rule, and a Web page indicating the extraction result is stored in the terminal device. Send. The user can obtain information related to the search keyword by opening a Web page linked to the Web page transmitted from the server device to the terminal device.

  However, the extraction result transmitted from the server device in response to the transmission of the search keyword is often linked to a Web page that does not provide information requested by the user. Therefore, the user can improve the quality of the extraction result by combining a large number of search keywords, but this is troublesome.

  As a technique for reducing such effort, for example, in Patent Document 1, a word that is frequently used in the same sentence together with a search keyword specified by the user is specified as a co-occurrence word and specified by the user. Techniques have been proposed for improving the quality of extraction results by automatically adding those co-occurrence words to search keywords and extracting sentences.

JP 2003-22275 A

  One reason why unnecessary information called so-called noise is mixed in the extraction result based on the search keyword as described above is that the topic tendency at the time of search is not necessarily taken into consideration. For example, a user who wants to know the latest liquid crystal display functions inputs a search keyword “liquid crystal display” and extracts a Web page. At that time, if the Web page A that explains the basic principle of the liquid crystal display in detail has been browsed by many people for many years, the Web page A is included in the extraction result. In this case, since the user is not interested in the basic principle of the liquid crystal display, the Web page A included in the extraction result is noise.

  As described above, when the user wants to search for information that takes into account the so-called topical trend at the present time, keywords such as “3D”, “white LED”, “IPS panel”, etc., which have recently become highly topical, are being used. What is necessary is just to input as a search keyword with "liquid crystal display". In the present application, the “topic trend” means a topic in which many people are interested. Since a topic arises not from a single word but from a combination of a plurality of words, an extraction result considering a topic tendency can be obtained by combining a plurality of search keywords as described above.

  However, the user cannot always easily know the current topic trend, and therefore it is often difficult to specify the search keyword to be added. Also, as in this example, you may want to know the topic trend itself. In that case, the user must first input a general search keyword such as “liquid crystal display” and search for necessary information from the noisy extraction results obtained as a result.

  In view of the circumstances described above, an object of the present invention is to provide a technique that allows a user to easily know a topic trend within a specific period related to a keyword in which a user is interested.

  In order to solve the above-described problem, the present invention provides a text data acquisition unit that acquires text data indicating a sentence together with time data indicating a time, and a morphological analysis of the sentence indicated by the text data acquired by the text data acquisition unit The morpheme analyzing means for dividing the morpheme and generating a plurality of morpheme data each indicating the divided morpheme, and one or more of the morpheme data generated by dividing each of the plurality of text data by the morpheme analyzing means For each of the morpheme data, the text data acquisition means together with the morpheme data, the text data identification data for identifying the text data used to generate the morpheme data, and the text data used to generate the morpheme data Morphological data storage that stores the acquired time data And the keyword data indicating one keyword, the morpheme data storage means stores the time keyword indicating the time within a predetermined period and the one keyword indicated by the one keyword data. Co-occurrence extracting morpheme data stored together with morpheme data together with the same text data identification data as the text data identification data stored in the morpheme data storage means as co-occurrence word data corresponding to the one keyword data An apparatus comprising word data extraction means is proposed.

  According to such an apparatus, since co-occurrence words used together with a specific keyword are extracted in a large number of text data published within a specific period, the trend of topics in the specific period related to the specific keyword is indicated. Information is obtained.

  Further, in the above-described apparatus, the text data identification data stored together with the morpheme data indicating the one keyword indicated by the one keyword data corresponds to the one keyword data in the morpheme data storage unit. For each piece of morpheme data extracted by the co-occurrence word data extraction unit as co-occurrence word data, the same morpheme data as the morpheme data stored together with the one text data identification data is stored in the one text data. Each of the co-occurrence word data corresponding to the one keyword data extracted by the co-occurrence word data extraction means, extracted as co-occurrence word data corresponding to the one keyword data included, and the one Co-occurrence word data corresponding to the one keyword data included in the text data Of based on the number of each, the provided relevance data generating means for generating a relevance data indicating the degree of association between one of the text data and the one of keyword data, it may be configured that the adopted.

  According to such a device, how many co-occurrence words used with a specific keyword in a large number of text data released within a specific period are used with the specific keyword in specific text data. Based on the information, the degree of association between specific text data and a specific keyword within a specific period is specified. Therefore, even if the user does not grasp the topic trend within the specific period, the user can obtain text data that seems to be related to the specific keyword in consideration of the topic trend.

  Further, in the above-described apparatus, the relevance level data generation unit is configured to generate the relevance level data according to a weight determined according to a predetermined rule based on the number of co-occurrence word data extracted by the co-occurrence word data extraction unit. A configuration may be employed in which the number of numbers related to each co-occurrence word data is weighted.

  According to such an apparatus, since co-occurrence words having a large number of appearances that generally characterize topic trends more strongly are considered in determining the relevance, co-occurrence with a small number of occurrences that do not characterize the topic tendency much Even in the case where there are many words, relevance data that sufficiently reflects the characteristics of the topic tendency can be obtained.

  In the above-described apparatus, the keyword data receiving means for receiving the one keyword data or the text data including the one keyword data from the terminal device, and the text data for storing the text data acquired by the text data acquiring means Each of the plurality of text data stored in the text data storage means and the one keyword data received from the terminal device by the keyword data reception means, calculated by the storage means and the relevance data generation means And a degree-of-association data transmitting unit that associates each of the plurality of text data with each of the plurality of text data and transmits the data to the terminal device.

  According to such an apparatus, since the process for specifying the degree of association is performed in an apparatus different from the terminal apparatus of the user, morphological analysis regarding a large number of text data, identification of co-occurrence words, and results of those processes There is no need for each user's terminal device to perform processing such as storage.

  Further, in the above-described apparatus, the relevance data transmission means sorts the plurality of text data in an order according to the relevance indicated by the relevance data and transmits the text data together with the relevance data to the terminal device. A configuration may be employed.

  According to such a device, the user of the terminal device can easily obtain information he / she wants to know in consideration of the topic trend within a specific period by looking at the contents of the text data sorted according to the degree of relevance. By that order, it is possible to know the contents of the topic trend of what topics are attracting more attention from the world.

Further, the present invention causes a computer, the text data indicating a text, and the text data acquisition means for acquiring together with time data indicating the time, the text indicated by the acquired text data by the text data acquisition means into morphemes by morphological analysis A morpheme analysis unit that divides and generates a plurality of morpheme data each indicating the divided morpheme, and one or more morpheme data of the morpheme data generated by dividing each of the plurality of text data by the morpheme analysis unit For each, the morpheme data, the text data identification data for identifying the text data used to generate the morpheme data, the time data acquired by the text data acquisition means together with the text data used to generate the morpheme data, and morphological data storing means for storing, one key Relates one keyword data indicating a word in the morpheme data storage means is stored together with time data indicating the time within a predetermined time period, and the morphological data storage means together with the morpheme data representing a keyword indicating the one keyword data because make function morphemes data stored together with the text data identification data of the same text data identification data stored, as co-occurrence word data extraction means for extracting as occurrence word data corresponding to the one keyword data to propose a program.

  According to such a program, the above-described apparatus is realized using a general computer.

  According to the present invention, a user can easily know a topic tendency within a specific period related to a keyword of interest.

It is the figure which showed the structure of the text search system concerning one Embodiment of this invention. It is the figure which showed the function structure of the server apparatus concerning one Embodiment of this invention. It is the figure which showed typically the data structure of text DB concerning one Embodiment of this invention. It is the figure which showed typically the data structure of morpheme DB concerning one Embodiment of this invention. It is the figure which showed the example of the result of the process of the morpheme analysis which the morpheme analysis part concerning one Embodiment of this invention performs. It is the figure which showed typically the data structure of co-occurrence word DB concerning one Embodiment of this invention. It is the figure which showed the process which the server apparatus concerning one Embodiment of this invention performs with registration of morpheme data. It is the figure which showed the process which the server apparatus concerning one Embodiment of this invention performs with the production | generation of co-occurrence word data and co-occurrence coefficient data. It is the figure which showed the process which the server apparatus concerning one Embodiment of this invention performs with the production | generation of relevance data, and transmission of relevance data and text data. It is the figure which showed typically the display screen of the extraction result displayed in the terminal device concerning one Embodiment of this invention.

[Embodiment]
Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a configuration of a text search system 1 according to the present embodiment. The text search system 1 receives from the terminal device 11 a terminal device 11 used by a user for searching for text and a search keyword input by the user, and a plurality of text data corresponding to the received search keyword is related to the search keyword. The server apparatus 12 which transmits to the terminal device 11 with the relevance degree data which shows is provided. The terminal device 11 and the server device 12 transmit / receive various data to / from each other via the network 9.

  In FIG. 1, only one terminal device 11 is shown for simplification of the drawing, but the number of terminal devices 11 is arbitrarily changed according to the number of users using the text search system 1. The number of server devices 12 can also be arbitrarily changed according to the scale of the text data search service in the text search system 1.

  The terminal device 11 is a device that can transmit data input from a user to the server device 12, can receive data transmitted from the server device 12, and can display information to the user according to the received data. Any device may be used. Therefore, for example, a mobile phone, a smart phone, a notebook PC (Personal Computer), a touch pad PC, a desktop PC, a PDA (Personal Digital Assistant), a game terminal having a communication function, a television having a communication function, etc. Any of these may be used. Since the configuration and operation of the terminal device 11 are the same as those of a general terminal device capable of data communication with the server device, description thereof is omitted.

  Since the hardware configuration of the server device 12 is the same as the hardware configuration of a general computer having a communication function, description thereof is omitted. The server device 12 operates as a device having the functional configuration shown in FIG. 2 by performing processing according to the application program according to the present invention.

The server device 12 includes the following components as functional components.
(Timer 121) The elapsed time from the reference time is continuously measured, and time data indicating the current time is generated.
(Morphological Analysis Unit 122) The server device 12 divides the text indicated by the text data acquired from another server device via the Internet into morphemes according to the morpheme analysis method, and generates morpheme data indicating these morphemes.
(Text data acquisition unit 123) The newly acquired text data is periodically acquired from another server device via the Internet.
(Search keyword data receiving unit 124) Receives search keyword data indicating a search keyword from the terminal device 11.

(Co-occurrence word data extraction unit 125) Along with the search keyword indicated by the search keyword data received by the search keyword data reception unit 124 mainly based on the morpheme data generated by the morpheme analysis unit 122, both within the same text within a predetermined period. Co-occurrence words that are used words are extracted, and co-occurrence word data indicating the extracted co-occurrence words is generated.
(Relevance data generation unit 126) Search keyword data received by the search keyword data reception unit 124 based on the morpheme data generated by the morpheme analysis unit 122 and the co-occurrence word data generated by the co-occurrence word data extraction unit 125 Relevance data indicating the relevance with the text data acquired by the text data acquisition unit 123 is generated.
(Relevance level data transmission unit 127) The text data acquired by the text data acquisition unit 123 is sorted according to the relevance level data generated by the relevance level data generation unit 126, and transmitted to the terminal device 11 together with the relevance level data.
(Storage unit 128) Stores various data.

The storage unit 128 stores the following data.
(Dictionary DB) DB (Database) that stores dictionary data used when the morphological analysis unit 122 performs morphological analysis.
(Grammar DB) A DB that stores grammar data used when the morpheme analysis unit 122 performs morpheme analysis.
(Text DB) A DB that stores text data acquired by the text data acquisition unit 123.
(Morpheme DB) A DB that stores morpheme data generated by the morpheme analyzer 122.
(Co-occurrence word DB) A DB that stores co-occurrence word data generated by the co-occurrence word data extraction unit 125.

  Since the dictionary DB and the grammar DB are known databases used in general morphological analysis, description of the data structure is omitted.

  FIG. 3 is a diagram schematically showing an example of the data structure of the text DB. The text DB is a collection of data records corresponding to each of the text data acquired by the text data acquisition unit 123, and each data record has a data field “text ID” for storing a text ID (Identifier) for identifying the text data. , A data field “time” for storing time data indicating the time when the text data was acquired, and a data field “text” for storing text data are provided.

  FIG. 4 is a diagram schematically showing an example of the data structure of the morpheme DB. The morpheme DB is a collection of data records corresponding to each of the morpheme data generated by the morpheme analysis unit 122, and each data record is a data field “text” that stores the text ID of the text data used to generate the morpheme data. ID ”, a data field“ time ”that stores time data indicating the time when the text data was acquired, and a data field“ morpheme ”that stores morpheme data.

  The morpheme analysis performed by the morpheme analysis unit 122 is a known technique, and a detailed description thereof will be omitted. FIG. 5 shows an example of the result of the morpheme analysis process performed by the morpheme analysis unit 122. FIG. 5 shows a result when the morphological analysis unit 122 performs morphological analysis on the text data “XX tree is still expensive. XX tree town seems to be interesting”. A data field “surface shape” shown in FIG. 5 indicates a morpheme divided from the input text. The data field “part of speech” indicates the part of speech of the morpheme. The data field “original form” indicates the original form of the morpheme. For example, a morpheme whose surface shape is “interesting” has an original shape of “interesting”, and in the processing of the server device 12 described below, words are compared based on the original shape, and words having the same meaning differ as different words depending on the variation. It will not be treated. Therefore, the data stored in the data field “morpheme” of the morpheme DB is data indicating the original form of the morpheme.

  Note that the morpheme data registered in the morpheme DB by the morpheme analysis unit 122 is not limited to all the morphemes (original forms) divided from the text as shown in FIG. Specifically, nouns, verbs, adjectives, adjective verbs, and the like are registered in the morpheme DB, and particles, adverbs, auxiliary verbs, conjunctions, symbols, and the like are not registered in the morpheme DB.

  FIG. 6 is a diagram schematically showing an example of the data structure of the co-occurrence word DB. The co-occurrence word DB is a collection of data records corresponding to each of the co-occurrence word data generated by the co-occurrence word data extraction unit 125, and each data record has been transmitted from the terminal device 11 to the server device 12. A data field “search ID” for storing a search ID for identifying a search request, a data field “search keyword” for storing search keyword data transmitted from the terminal device 11 in the search request, and a source morpheme to be extracted A data field “time zone” for storing time zone data indicating a time zone to be narrowed down by the time at which the text data was acquired, and a data field “co-occurrence word” for storing co-occurrence word data indicating morphemes extracted as co-occurrence words ”, Stores co-occurrence coefficient data indicating the co-occurrence coefficient (described later) between the search keyword and the co-occurrence word It is equipped with a data field "co-occurrence factor".

  As shown in FIG. 6, the co-occurrence word data and the co-occurrence coefficient data generated by the co-occurrence word data extraction unit 125 are the co-occurrence extracted from the morphemes generated from the text acquired in a specific time zone. The words and their co-occurrence coefficients are shown. In the present embodiment, this time zone is assumed to be the time zone for the previous month from the time when the server device 12 received the search request from the terminal device 11, and is automatically set by the co-occurrence word data extraction unit 125. The

  Next, the operation of the text search system 1 will be described. First, the server device 12 periodically (for example, once a day at a predetermined time) crawls text data indicating documents newly published on the Internet after the previous day from various server devices, Morphological analysis is performed using text data, and morpheme data is registered. FIG. 7 is a diagram showing processing performed by the server device 12 in accordance with registration of the morpheme data.

  First, the text data acquisition unit 123 makes a transmission request for text data including time data indicating the date and time when a similar request was last made to each external server device, and is transmitted from each server device as a response thereto. The newly updated text data is received (S101). The text data acquisition unit 123 assigns a text ID to each of the acquired text data, and stores it in the text DB (FIG. 3) together with the time data acquired from the time measuring unit 121 (indicating the time at the time of acquisition) (S102). .

  The morpheme analysis unit 122 performs morpheme analysis using the text data stored in the text DB by the text data acquisition unit 123 (S103), and extracts the part of speech that has meaning alone from the generated morpheme data ( S104), the extracted morpheme data is stored in the morpheme DB (FIG. 4) together with the text ID and time data of the text data used for the generation (S105). The above is the process accompanying registration of morpheme data.

  A user inputs a search keyword in which he / she is interested in the terminal device 11 and transmits the search keyword to the server device 12, and as a response, the text related to the search keyword is transmitted from the server device 12 to the terminal device 11. The data can be browsed together with the relevance data with the search keyword. At that time, the relevance data used for selecting the text data transmitted from the server device 12 to the terminal device 11 is generated based on, for example, morpheme data generated from the text data acquired in the past month. It is generated based on the co-occurrence word data and the co-occurrence coefficient data. Therefore, the text data provided to the user is extracted in consideration of the topic trend in the past month.

  FIG. 8 is a diagram illustrating processing associated with generation of co-occurrence word data and co-occurrence coefficient data performed in the server device 12 when a search keyword input to the terminal device 11 by the user is transmitted to the server device 12. It is.

  First, the search keyword data transmitted from the terminal device 11 is received by the search keyword data receiving unit 124 (S201). With the reception of the search keyword data as a trigger, the co-occurrence word data extraction unit 125 stores time data indicating the time included in the time zone of the past month in the data field “time” from the morpheme DB (FIG. 4). A data record group is extracted (S202).

  Subsequently, the co-occurrence word data extraction unit 125 extracts a data record group in which the same data as the search keyword data received in step S201 is stored in the data field “morpheme” from the data record group extracted in step S202. (S203).

Subsequently, the co-occurrence word data extraction unit 125 groups the data record group extracted in step S203 by the data field “text ID”, and calculates the number of those groups, that is, the number of text data including the search keyword data as the search keyword. to identify as the appearance number of documents _{df (w i) (S204)} .

  Subsequently, the co-occurrence word data extraction unit 125 groups the data record group extracted in step S202 by the data field “morpheme” (S205). Hereinafter, in step S205, it is assumed that data record groups corresponding to the first to nth (where n is an arbitrary natural number) morphemes are generated.

  Subsequently, the co-occurrence word data extraction unit 125 selects the m-th morpheme data record group (where m is an arbitrary natural number between 1 and n) from the n data record groups generated in step S205. (S206).

Subsequently, the co-occurrence word data extraction unit 125 groups the data record group of the mth morpheme selected in step S206 by the data field “text ID”, and the number of those groups, that is, the text including the mth morpheme data. The number of data is specified as the number of appearance documents df (w _jm ) of the mth morpheme (S207).

Subsequently, the co-occurrence word data extraction unit 125 is stored in any data field “text ID” of the data record group extracted in step S203, and any one of the m-th morpheme data record group selected in step S206. The number of text IDs stored in the data field “text ID” is specified as the co-occurrence document number df (w _i , w _jm ) (S208). The co-occurrence document number df (w _i , w _jm ) specified in this way is the number of text data including both search keyword data and morpheme data indicating the mth morpheme.

Subsequently, the co-occurrence word data extraction unit 125 determines the number of appearance documents df (w _i ) specified in step 204, the number of appearance documents df (w _jm ) specified in step 207, and the number of co-occurrence documents df (w ( _m ) specified in step S208. w _i, with w _jm), in accordance with the following equation co-occurrence coefficient Dice (w _i, w _jm) calculating a (S209).

  The co-occurrence coefficient is an index indicating the frequency with which two words (in this case, the search keyword and the m-th morpheme) appear in the same document, and the above formula is a calculation of the co-occurrence coefficient known as a dice coefficient. It is a formula.

Subsequently, the co-occurrence word data extraction unit 125 receives the co-occurrence coefficient data indicating the co-occurrence coefficient Dice (w _i , w _jm ) calculated in step 209 in response to the search request when receiving the search keyword data in step S201. The search ID assigned by the search keyword data receiving unit 124, the search keyword data received in step S201, the time zone data indicating one month before the reception timing of the search keyword data, and the co-occurrence word indicating the mth morpheme The data is stored together with the data in the co-occurrence word DB (FIG. 6) (S210).

  Subsequently, the co-occurrence word data extraction unit 125 determines whether or not the registration process in step S210 has been completed for all the groups corresponding to the n morphemes grouped in step S205 (S211). If it is determined in step S211 that there is a morpheme group for which registration processing has not yet been completed (S211; No), the co-occurrence word data extraction unit 125 returns the process to step S206, and the (m + 1) th morpheme. The data record group is selected, and the processing after step S207 is performed. On the other hand, if it is determined in step S211 that the registration process has been completed for all morpheme groups (S211; Yes), the co-occurrence word data extraction unit 125 ends the series of processes.

  The above is the process accompanying the generation of co-occurrence word data and co-occurrence coefficient data related to the search keyword input by the user.

  When the process of generating the co-occurrence word data and the co-occurrence coefficient data by the co-occurrence word data extraction unit 125 is completed, the generation of the relevance data by the relevance data generation unit 126 and the text data by the relevance data transmission unit 127 And the process of transmission of the relevance data to the terminal device 11 is performed. FIG. 9 is a diagram showing these processes.

  First, the degree-of-association data generation unit 126 makes a search request by the search keyword data receiving unit 124 with the reception of the search keyword data transmitted from the terminal device 11 in step S201 from the co-occurrence word DB (FIG. 6). A data record including the search ID assigned to the data field “search ID” is extracted (S301).

  Subsequently, the relevance data generation unit 126 extracts, from the morpheme DB (FIG. 4), a data record in which time data indicating time included in the past month is stored in the data field “time” (S302).

  Subsequently, the relevance data generation unit 126 groups the data record group extracted in step S302 by the data field “text ID” (step S303). Hereinafter, in step S303, it is assumed that data record groups corresponding to text data from the first to xth (where x is an arbitrary natural number) are generated.

  Subsequently, the relevance data generation unit 126 selects a data record group of text data of y-th (where y is an arbitrary natural number between 1 and x) from the x data record groups generated in step S303. (S304).

Subsequently, the relevance data generation unit 126 groups the data records of the y-th text data selected in step S304 by the data field “morpheme”, and counts the number of data records included in each of these groups ( S305). Now, it is assumed that the data record group corresponding to the first to pth (where p is an arbitrary natural number) morpheme is generated by the grouping in step S305, and the qth (where q is 1 to r) It is assumed that the number of data records C _q is counted as the number of data record groups according to the morpheme of any natural number).

Subsequently, the relevance data generation unit 126 assigns each of the first to p-th morphemes (hereinafter referred to as q-th) from the co-occurrence word DB (FIG. 6) by the search keyword data reception unit 124 in step S201. The retrieved search ID is stored in the data field “search ID”, the data record in which the co-occurrence word data indicating the q-th morpheme is stored in the data field “co-occurrence word” is retrieved, and the data of the retrieved data record The co-occurrence coefficient data stored in the field “co-occurrence coefficient” is read as the co-occurrence coefficient D _q related to the q-th morpheme (S306).

Subsequently, the relevance data generation unit 126 calculates the relevance R _y between the y-th text data and the search keyword data according to the following formula, and generates relevance data indicating the calculated relevance R _y. (S307).

  Subsequently, the relevance data generation unit 126 determines whether or not the relevance data generation processing in step S307 has been completed for all the groups corresponding to the x pieces of text data grouped in step S303 (S308). . If it is determined in step S308 that there is a group of text data for which the relevance data generation processing has not yet been completed (S308; No), the relevance data generation unit 126 returns the processing to step S304, and the ( The data record group of the text data of y + 1) is selected, and the processes after step S305 are performed.

  On the other hand, in the determination of step S308, when it is determined that the relevance data generation processing has been completed for all text data groups (S308; Yes), the relevance data generation unit 126 finishes a series of processing, and then continues to be related. Processing of transmission of text data and relevance data by the degree data transmission unit 127 is performed.

  The relevance data transmission unit 127 uses the text ID stored in the data field “text ID” as a search key for each of the data records grouped in step S303 for the first to x-th text data. First to xth text data are extracted from the text DB (FIG. 3) (S309).

  Subsequently, the association degree data transmission unit 127 associates the first to xth text data extracted in step S309 with the association degree data generated for the first to xth text data in step S307, respectively. The text data is sorted in descending order of the degree of association indicated by the degree of association data (S310).

  Subsequently, the relevance data transmission unit 127 transmits the text data sorted in step S310 to the terminal device 11 together with the relevance data associated with them (S311). The above is the process associated with the generation of relevance data by the relevance data generation unit 126 and the transmission of text data and relevance data by the relevance data transmission unit 127.

  When the terminal device 11 receives the text data with the relevance data transmitted from the server device 12, the terminal device 11 displays the content. FIG. 10 is a diagram schematically illustrating a display screen of extraction results displayed on the terminal device 11. By viewing the extraction result display screen, the user can know the contents of the document related to the search keyword previously input by the user. In this case, the displayed text data is text data selected in consideration of the appearance frequency of the co-occurrence words used in the same document together with the search keyword in the past month, and the display order is the co-occurrence words. Since the degree of relevance calculated in consideration of the appearance frequency of the items is in descending order, the user can know the trend of topics related to the search keyword in the past month by reading the contents.

[Modification]
The embodiment described above is an embodiment of the present invention, and can be variously modified within the scope of the technical idea of the present invention. Examples of these modifications are shown below.

  In the embodiment described above, the text search system 1 is configured by the terminal device 11 and the server device 12, and all processes except the input of search keyword data and the result display are performed in the server device 12. Instead, a configuration in which all or part of the processing performed by the server device 12 is performed in the terminal device 11 may be employed. For example, when the terminal device 11 has sufficient processing speed, communication speed, storage capacity, and the like like a desktop PC, the terminal device 11 can also serve as the server device 12.

  In the embodiment described above, the Dice coefficient is employed as the co-occurrence coefficient. However, any index other than the Dice coefficient may be employed as the co-occurrence coefficient as long as the index indicates the degree of co-occurrence. .

  In the above-described embodiment, the text data presented to the user is extracted from the text data acquired in the server device 12 during the past month. The range of the acquisition time can be arbitrarily changed, and a configuration that does not perform extraction based on the acquisition time may be employed. That is, the acquisition time of the morpheme data used in the generation of the co-occurrence word data and the co-occurrence coefficient data is limited to a predetermined period such as the past month, but the text data to be extracted is acquired outside the predetermined period It may be.

  In the above-described embodiment, the predetermined period used when narrowing down the morpheme data used in the generation of the co-occurrence word data and the co-occurrence coefficient data based on the acquisition time is the past one month. For example, when the user inputs a search keyword, the predetermined period can be specified. Based on the morpheme data acquired in the predetermined period specified by the user, the co-occurrence word data and the co-occurrence coefficient data A configuration in which generation is performed may be adopted.

  In the above-described embodiment, all morphemes are uniformly used for generating relevance data regardless of the number of appearances within a predetermined period. For example, for morphemes with a large number of appearances within a predetermined period, A morpheme with a smaller number of occurrences appears as a co-occurrence word by giving a larger weight than a morpheme with a smaller number of occurrences, and weighting according to the weight in the generation of relevance data. A configuration may be employed in which relevance data indicating a higher relevance level than when it appears. For example, by assigning “0” as a weight to a morpheme with an appearance count less than a predetermined threshold and giving “1” as a weight to a morpheme with an appearance count greater than or equal to a predetermined threshold, A configuration may be adopted in which only morphemes are considered in generating relevance data.

  In the above-described embodiment, a configuration is employed in which time data indicating the time at which text data as the source is acquired is used as time data corresponding to morpheme data. The time data corresponding to the morpheme data is not limited to this. For example, when the text data as the source is accompanied by the time data indicating the publication time of the document, the time data is used as the time data corresponding to the morpheme data. A configuration may be employed.

  In the above-described embodiment, the user inputs a single search keyword, and the relevance data corresponding to the single search keyword data is generated. However, the user inputs a plurality of search keywords. And the structure by which the production | generation of the relevance data according to the some search keyword data which respectively show those some search keywords may be employ | adopted. In that case, for example, a configuration may be employed in which the relevance data generation unit 126 generates relevance data indicating an addition value of relevance calculated for each search keyword data.

  Furthermore, even if the user designates text data instead of inputting a search keyword and the server device 12 uses a plurality of morpheme data generated by dividing the text data by the morpheme analysis unit 122 as a search keyword. Good. In that case, the user can read a document related to the document by designating the document in which the user is interested. Also in that case, the text data presented to the user is selected in consideration of the topic tendency in a predetermined period.

  In the above-described embodiment, the relevance data indicates a numerical value obtained by adding the weights of the co-occurrence words weighted by the co-occurrence coefficient. For example, the co-occurrence coefficient is not used and the co-occurrence coefficient is not used. Data indicating the number of occurrences of words may be used as relevance data as it is, or a coefficient different from the co-occurrence coefficient may be used.

  In the embodiment described above, the text data transmitted from the server device 12 to the terminal device 11 is sorted in descending order of the corresponding relevance. For example, from the server device 12 to the terminal device. 11, the text data arranged in an arbitrary order may be transmitted together with the relevance data, and the terminal device 11 may sort the text data. Further, the sort order can be arbitrarily changed. For example, the sorting may be performed in the order of the length of the text indicated by the text data or the order of the acquired time. In addition, text data is posted on a web page, and is an index of public interest in the text data, such as the number of links pasted from other web pages to the text data and the number of browsing web pages. When data is obtained, a configuration may be employed in which sorting is performed according to those indexes (such as the number of links and the number of browsing).

  In the above-described embodiment, the server apparatus 12 is realized by causing a general computer to execute processing according to the program of the present invention. For example, the functional configuration shown in FIG. The server device 12 may be configured as a so-called dedicated machine realized by hardware.

  Note that the formulas and the contents and order of the processing flows shown in the above-described embodiments are examples for explanation, and can be variously changed.

DESCRIPTION OF SYMBOLS 1 ... Text search system, 9 ... Network, 11 ... Terminal device, 12 ... Server apparatus, 121 ... Timekeeping part, 122 ... Morphological analysis part, 123 ... Text data acquisition part, 124 ... Search keyword data receiving part, 125 ... Co-occurrence Word data extraction unit, 126 ... relevance data generation unit, 127 ... relevance data transmission unit, 128 ... storage unit

Claims (6)

Text data acquisition means for acquiring text data indicating a sentence together with time data indicating a time;
A morpheme analysis unit that divides a sentence indicated by the text data acquired by the text data acquisition unit into a morpheme by morpheme analysis and generates a plurality of morpheme data each indicating the divided morpheme;
For each of one or more morpheme data generated by dividing each of a plurality of text data by the morpheme analysis unit, the morpheme data and the text data used to generate the morpheme data are identified. Morpheme data storage means for storing the text data identification data and the time data acquired by the text data acquisition means together with the text data used for generating the morpheme data;
With respect to one keyword data indicating one keyword, the morpheme data storage means stores the time data indicating the time within a predetermined period and the morpheme data indicating the one keyword indicated by the one keyword data. Co-occurrence word data extraction for extracting morpheme data stored together with the same text data identification data as the text data identification data stored in the morpheme data storage means as co-occurrence word data corresponding to the one keyword data A device comprising: With respect to one text data identification data stored together with morpheme data indicating the one keyword indicated by the one keyword data, the co-occurrence word data corresponding to the one keyword data is stored in the morpheme data storage unit as the co-occurrence word data. For each of the morpheme data extracted by the word data extraction means, the one keyword data included in the one text data is the same morpheme data as the morpheme data stored together with the one text data identification data. And the number of co-occurrence word data corresponding to the one keyword data extracted by the co-occurrence word data extracting means and the one text data included in the one text data And the number of co-occurrence word data corresponding to the keyword data of The apparatus of claim 1, further comprising a relevance data generating means for generating a relevance data indicating the degree of association between the text data and the one of the keyword data. The relevance data generation means is a number related to each co-occurrence word data in the generation of the relevance data according to a weight determined according to a predetermined rule based on the number of co-occurrence word data extracted by the co-occurrence word data extraction means. The apparatus according to claim 2, wherein weighting is performed. Keyword data receiving means for receiving the one keyword data or text data including the one keyword data from a terminal device;
Text data storage means for storing text data acquired by the text data acquisition means;
Between each of the plurality of text data stored in the text data storage means and the one keyword data received from the terminal device by the keyword data receiving means, calculated by the relevance level data generating means. The device according to claim 2, further comprising: relevance data transmission means for transmitting relevance data to the terminal device in association with each of the plurality of text data. 5. The relevance data transmitting means transmits the plurality of text data to the terminal device together with the relevance data after sorting the plurality of text data in an order according to the relevance indicated by the relevance data. apparatus. The computer,
Text data acquisition means for acquiring text data indicating a sentence together with time data indicating a time;
A morpheme analysis unit that divides a sentence indicated by the text data acquired by the text data acquisition unit into a morpheme by morpheme analysis and generates a plurality of morpheme data each indicating the divided morpheme;
Text identifying the morpheme data and the text data used to generate the morpheme data for each of one or more morpheme data generated by dividing each of the plurality of text data by the morpheme analysis unit Morpheme data storage means for storing data identification data and time data acquired by the text data acquisition means together with text data used for generating the morpheme data ;
Relates one keyword data showing one keyword, the morpheme in the data storage means is stored together with time data indicating the time within a predetermined time period, and the morphological data with morphological data indicating a keyword indicating the one keyword data occurrence word data extracting means for extracting morphemes data stored with the same text data identification data and text data identification data stored in the storage means, as the occurrence word data corresponding to the one keyword data
Program for make function as.

Images