JP4087769B2 - Server and related word proposal method - Google Patents

Description

  The present invention relates to a service for providing a document to a corresponding user when a document including a keyword (word) registered in advance by the user is collected. More specifically, the present invention relates to a technique for proposing new words, related words, and search expressions for keywords registered in advance by a user.

  In recent years, for example, a clipping service that provides a real-time document (for example, a news article) has been implemented as a service that applies document collection and document search from the web. This clipping service is a service that automatically distributes a document to a corresponding user when a document including a keyword registered in advance by the user is collected.

  In the conventional clipping service, the keyword registered by the user is continuously used unless the user himself performs a change operation. For this reason, the conventional clipping service has a problem that the registered keyword becomes obsolete. Furthermore, the keywords registered by the user are mostly words and phrases within the range of knowledge possessed by the user. For this reason, the keyword registered by the user may be insufficient as a material for extracting information requested by the user.

  In addition, as a technique for presenting information using registered keywords or the like, there is a method of presenting related words for keywords (search terms) by calculating relevance from the appearance frequency of phrases (Patent Document 1, Patent Document 2).

In addition, as a technique for presenting information using a keyword, there are techniques disclosed in Patent Document 3 and Patent Document 4.
JP-A-9-044523 JP 2000-222427 A JP-A-5-151273 Japanese Patent Laid-Open No. 10-134075

  However, conventional related word proposal methods do not take into account temporal changes related to the old and new documents. For this reason, it is not possible to suggest words or phrases that suddenly become related due to social backgrounds or events that have attracted the attention of the world, and provide real-time information like a clipping service. It was unsuitable to apply to the service.

  The present invention provides a technique capable of solving the above-described problems, presenting seasonal words that occur due to changes over time, and assisting maintenance of keywords registered by the user. With the goal.

In order to solve the above problems, the present invention has the following configuration. That is, the server of the present invention uses a classification dictionary that shows a relationship between a document, a word included in the document, and a field to which the document belongs as a relationship between the word and the field of the word. based on the words that are contained in, and means for classifying the documents in the category, document frequency is the number of documents that contain that word in the same categories for words included in the document It means for calculating for each unit interval and means for recording the estimated transfer of the document frequency in a plurality of the unit sections in the word information database, referring to the word information database, in the same category, the first word Means for extracting a second word having a similar document frequency transition as a related word of the first word.

According to the present invention, a document, a word included in the document, and a field to which the document belongs are included in the document using a classification dictionary that indicates the relationship between the word and the field of the word. based on the words that are to classify the documents in the category, in the same category for the word contained in the document, the unit sections each the document frequency is the number of documents that contain the word calculated, to record changes in the document frequency in a plurality of the unit sections in the word information database, referring to the word information database, in the same category, the having a transition of document frequency similar to the first word Two words can be extracted as related words of the first word. Here, the first word can be defined as a registered word registered in advance in the server, and the transition can be defined as an increase / decrease rate per unit period. Therefore, for example, using a registered word registered in advance as a key, a word having a document frequency transition similar to a document frequency transition per unit period of the registered word can be extracted as a related word.

Preferably, said server, said the first word or the number of documents that the related word is included, the ratio of the number of first words and documents that contain both the relevant word matching degree And a means for outputting a search expression including the first word and its related word in accordance with the degree of coincidence.

According to the present invention, for the number of documents that contain the first word or related terms thereof, the degree of coincidence is the ratio of the number of first words and documents that include both the related words The search expression including the first word and the related word can be output according to the degree of coincidence. Therefore, for example, a related word for a registered word registered in advance can be proposed in combination with the registered word.

  Preferably, the output means included in the server outputs a search expression composed of a logical product of the first word and its related word when the degree of coincidence reaches a predetermined value, and the degree of coincidence reaches the predetermined value. If not, a search expression composed of a logical sum of the first word and its related word may be output.

  According to the present invention, when the degree of coincidence reaches a predetermined value, a search expression consisting of the logical product of the first word and its related word is output, and when the degree of coincidence does not reach the predetermined value, the first It is possible to output a search expression consisting of the logical sum of the words and related words. Therefore, for example, when a related word for a registered word registered in advance is proposed in combination with the registered word, the combination can be changed according to the degree of coincidence.

  Preferably, the output means included in the server is configured such that, even when the degree of coincidence reaches a predetermined value, the first word and its related word are equivalent to pre-registered synonyms. You may comprise so that the search formula consisting of the logical product of 1 word and its related word may not be output.

  According to the present invention, even when the degree of coincidence reaches a predetermined value, if the first word and its related word are equivalent to pre-registered synonyms, the first word and its related word It is possible not to output a search expression consisting of a logical product of words. Therefore, for example, synonyms registered in advance in the synonym dictionary or the like can be excluded even if the word is considered not to be a synonym due to the value of the degree of coincidence. For this reason, it is possible to extract not only a synonym (paraphrase) of a registered word but also a word that has become related to the registered word as time passes.

Preferably, the server further includes means for accepting registration of the first word from a user, and the means for outputting transmits means for sending a notification including a search expression to the user who has registered the first word. And means for receiving a response including adoption or non-adoption of the search formula from the user in response to the transmitted notification, and if the response includes adoption, the search formula is continued after a certain period of time. It may be configured to have a means for transmitting a notification for confirming whether or not to use.

  According to the present invention, the registration of the first word is accepted from the user, the notification including the search formula is transmitted to the user who has registered the first word, and the user is notified of the transmitted notification. When a response including adoption or non-adoption of a search expression is accepted and the response includes adoption, a notification confirming whether or not to continue using the search expression after a certain period has elapsed Can do. Therefore, for example, since a user who has once employed a search expression can confirm the will of using the search expression again after a certain period of time, maintenance of a registered word registered by the user can be assisted. .

  Further, the present invention may be a method in which a computer, other devices, machines, etc. execute any one of the processes described above. Furthermore, the present invention may be a program that causes a computer, other devices, machines, or the like to realize any of the above functions. Further, the present invention may be a program in which such a program is recorded on a computer-readable recording medium.

  According to the present invention, it is possible to present seasonal phrases that occur due to temporal changes to the user in combination with the original keywords. This makes it possible to present seasonal phrases to the user while maintaining the original keywords registered by the user.

The best mode for carrying out the present invention will be described below with reference to the drawings. The description of the present embodiment is an exemplification, and the configuration of the present invention is not limited to the following description.
<Embodiment>
In the present embodiment, the case where the present invention is applied to an information system that performs a clipping service for a user will be described.

<Overview>
The information system proposes new words / related words / search formulas to a user who uses the clipping service, thereby realizing a function that enables maintenance of keywords registered in advance by the user. Below, the points of new words, related words, and search formulas proposed to the user will be described.

  For new words, in the field (genre) in which the keywords registered by the user frequently appear, the appearance frequency is higher than a preset threshold value, and the appearance increase rate is set in advance compared to the previous day, for example, the previous day. A phrase higher than a predetermined threshold is referred to, and such a phrase is presented to the user as a new word.

  Related words are words and phrases whose appearance frequency is similar to that of the user's keyword in a field (genre) in which the keywords registered by the user frequently appear, and such words are related words to the user. Presented. When the related word is adopted by the user, it becomes a new keyword in the clipping service. Further, the transition of the appearance frequency is determined using, for example, the values of the graph shown in FIG. FIG. 12 is a diagram illustrating an example in which the appearance frequency of a keyword for each certain period is graphed. In this way, the similarity between keywords is determined by obtaining the degree of similarity of the inclination from the transition of the time-series frequency for each keyword. That is, it is determined whether or not the same inclination change is made as a whole.

  The search formula is a method of combining the keywords registered by the user in the presentation of related words, and such a combination is proposed to the user as a search formula. This is for considering the conditions under which the related word is used for the keyword registered by the user to provide the clipping service. Hereinafter, the present embodiment will be described in detail by using related words and search formulas as examples.

<System configuration>
FIG. 1 is a diagram showing a system configuration of the information system. This information system is realized by the apparatus 1 providing a service to the user 20. The device 1 is an information device such as a personal computer, a workstation, or a server. The apparatus 1 includes a content collection system 2, a content classification processing unit 3, a search index creation unit 4, a keyword / search formula suggestion processing unit 5, a keyword proposal management processing unit 6, a clipping processing system 7, and various databases. The user 20 holds a user terminal that can communicate with the device 1. Since this user terminal is assumed to be a general information terminal, detailed description thereof is omitted.

  The various databases are a content database (DB) 8, a search database 9, a document / word correspondence list 10, a category master database 11, a classification information database 12, a word information database 13, and a suggested search formula management database 14. Details of the various databases will be described later in the section <Data structure>.

  The content collection system 2 collects documents from the Internet using a web collection robot. The content classification processing unit 3 classifies the collected documents into each field (genre). This classification is performed using a generally known document classification technique. For example, a classification dictionary indicating the relationship between words contained in a document and a field may be used. The classification result is stored in the classification information database 12.

  The search index creation unit 4 creates a search database 9 from the collected documents using an indexer. At the same time, the search index creation unit 4 creates a document / word correspondence list 10 that can identify which keyword contains which keyword.

  The keyword / search formula suggestion processing unit 5 performs a search for presenting a new keyword and the keyword based on the document / word correspondence list 10 and the keyword registered by the user 20 based on the result of classifying the document. Processing for proposing an expression to the user 20 is performed.

  The keyword proposal management processing unit 6 performs processing related to the update of the proposal retrieval formula management database 14 and other services based on the response result from the user 20 to the proposal. For example, the keyword suggestion management processing unit 6 performs temporary registration or main registration of keywords and search expressions. Further, the keyword proposal management processing unit 6 notifies the user 20 of confirmation of the main registration after a certain period of time has elapsed.

  The clipping processing system 7 holds a series of information related to the clipping service, and executes the clipping service for the user 20 based on the information. The information here is, for example, a keyword registered by the user 20 in advance. The clipping processing system 7 also functions as a means for accepting keyword registration from the user 20.

<data structure>
Next, various databases included in the apparatus 1 will be described with reference to FIGS. FIG. 2 is a diagram illustrating a data structure of the document / word correspondence list 10. The document / word correspondence list 10 manages correspondence between words and documents. Each document is managed as a document ID list by adding a document ID. The document ID is a unique ID assigned to each collected document. In the example shown in FIG. 2, it is indicated that the word “mad cow disease” is included in the document having the ID “1, 105, 201”.

  Next, the data structure of the search database 9 will be described. The search database 9 holds information having the same contents as the document / word correspondence list 10. This search database 9 is different from the document / word correspondence list 10 in that it can be accessed at high speed. In the search database 9, for example, an address is generated from a word using a hash or the like, and a document ID can be acquired based on the address.

  FIG. 3 is a diagram showing the data structure of the category master database 11. The category master database 11 manages each category by associating the category ID with the category name. The category is an information type for classifying words. In the example shown in FIG. 3, industry types such as “Finance” and “Food” are shown as categories. The category ID is an ID that uniquely indicates a category that is a unit of classification. The category name is a name that identifies the category.

  FIG. 4 is a diagram showing the data structure of the classification information database 12. The classification information database 12 manages the correspondence between document IDs and category IDs. In the example shown in FIG. 4, it is indicated that the document having the document ID “1” belongs to the category having the ID of “2, 5, 10”.

  FIG. 5 is a diagram illustrating a data structure of the word information database 13. The word information database 13 has items “a (k)”, “df (1)”, “update date”, and “user use” corresponding to each combination of “word” and “category ID”. Among these, “a (k)” indicates that n times of processing results are recorded by k = 1,..., N. Here, a (k) indicates the inclination of the document frequency converted for each unit section (section before k times and before k + 1 times). Document frequency refers to the number of documents in the category in which the word is detected. Moreover, this inclination is the increase / decrease number per unit section (for example, every day). df (1) is the document frequency at the previous processing. The update date is the execution date when the keyword proposal process was executed last time. User use indicates whether or not the word is used (specified) by the user as a clipping keyword, and is indicated by a value of “0 (unused)” or “1 (used)”. With this word information database 13, it is possible to grasp the tendency and transition of the appearance frequency for each category with respect to the words existing in the collected document.

  FIG. 6 is a diagram showing the data structure of the proposed search expression management database 14. The proposal search formula management database 14 has items indicating a search formula corresponding to the user ID, a proposal date, and a status flag. The user ID is an ID for identifying a user for whom a search formula has been proposed. The search formula indicates the proposed search formula. The proposal date indicates the date on which the search formula is proposed. The status flag represents the status of the search formula proposed to the user using values of 1 to 4. Each status is set to “1” when the user is proposing, “2” when the user is evaluating, “3” when the user adopts the proposal, and the user adopting the proposal. If not (not adopted), set to “4”. Further, an item whose status flag indicates “4 (not adopted)” may be deleted in a certain time.

The content database 8 holds data including correspondence between documents and document IDs. Here, the documents collected by the collecting robot are managed by the document ID.
<Processing flow>
Next, the flow of processing executed by the keyword / search formula suggestion processing unit 5 of the device 1 will be described. First, the overall flow executed by the keyword / search formula suggestion processing unit 5 will be described with reference to FIG.

《Keyword / search formula proposal processing flow》
FIG. 7 is a flowchart showing the keyword / search formula suggestion process. As shown in FIG. 7, the keyword / search formula suggestion processing unit 5 executes the keyword / search formula proposal processing for the user 20 in the order of pre-processing, word information processing, and search formula proposal processing.

  The keyword / search formula suggestion processing unit 5 aggregates the document frequency and the increase amount for all the words in the document collected as preprocessing (S10). Then, the keyword / search formula suggestion processing unit 5 proceeds to word information processing when trying to propose a search formula to the user 20 (S20). Here, for example, a keyword (word) X that is likely to be related to the keyword W is extracted. Subsequently, the keyword / search formula suggestion processing unit 5 executes, as the search formula proposal processing, for example, processing for proposing a search formula combining the keyword W and the extracted keyword X to the user (S30). Hereinafter, each processing step will be described in more detail.

《Pre-processing flow》
FIG. 8 is a flowchart showing details of the preprocessing shown in FIG. This pre-process is an update process of the word information database 13 that is executed for all the words in the document / word correspondence list 10 mainly by the keyword / search formula suggestion processing unit 5. In the following description, “d (t, Cm, 0)” indicates a processing date.

  First, the keyword / search formula suggestion processing unit 5 totals the document frequency “df (t, Ck, 0)” for each “category Ck” of the “word t” from the document / word correspondence list 10 and the classification information database 12. (S11). Here, “0” in the document frequency indicates this processing.

  Subsequently, the keyword / search formula suggestion processing unit 5 acquires the corresponding document frequencies “df (1)” and “update date” from the word information database 13 using “word” and “category ID” as keys (S12). ). The document frequency “df (1)” and “update date” acquired here are replaced with “df (t, Ck, 1)” and “d (t, Ck, 1)” in the subsequent processing. Used. Here, “1” in the document frequency indicates that it was calculated in the process one time before this time.

  Subsequently, the keyword / search formula suggestion processing unit 5 calculates a value “a (0)” indicating the inclination of the document frequency in the current process (S13). a (0) is a difference in document frequency ((df (t, Ck, 0) −df (t, Ck, 1)) to a difference in date (update date) ((d (t, Ck, 0) −d It is obtained by dividing by (t, Ck, 1)).

  Subsequently, the keyword / search formula suggestion processing unit 5 updates the word information database 13 (S14). Here, the obtained a (0) is replaced with a (1), and the gradient of the document frequency is updated while shifting subsequent values (a (1), a (2)) one by one. In addition, the document frequencies “df (1)” and “update date” acquired in S12 are simultaneously updated to the document frequency and the current date tabulated in S11.

Through the processing as described above, the word information database 13 can always hold the latest N pieces of data for each word.
《Word information processing flow》
FIG. 9 is a flowchart showing details of the word information processing shown in FIG. This word information processing is a process of searching for a keyword to be proposed for a keyword registered by the user 20. That is, it is a process of extracting a new keyword for the keyword registered by the user 20. Here, the keyword registered by the user 20 is described as W.

  First, the keyword / search formula suggestion processing unit 5 uses the “word” and the “category ID” as keys from the word information database 13 and the corresponding document frequency gradients “a (1), a (2)... A ( n) "is acquired (S21). The inclinations “a (1), a (2)... A (n)” of the acquired document frequencies are “a (W, Ck, 1), a (W, Ck, 2). , Ck, n) ".

  Subsequently, the keyword / search formula suggestion processing unit 5 similarly obtains a corresponding document frequency gradient for the word t that has not been processed from the word information database 13 by using the same “category ID” as a key (S22). ). This step is performed for all words belonging to the same category. The acquired document frequency gradients “a (1), a (2)... A (n)” are “a (t, Ck, 1), a (t, Ck, 2). (T, Ck, n) "is used as a replacement.

  Subsequently, the keyword / search formula suggestion processing unit 5 calculates the similarity of the gradient of the document frequency between the keyword W and the word t (S23). This similarity is obtained by adding the absolute value of the difference in a certain section (for example, (| a (w, Ck, 1) −a (t, Ck, 1) |)).

  Subsequently, the keyword / search formula suggestion processing unit 5 determines whether or not the obtained similarity value is equal to or less than a threshold value (S24). Here, if the similarity value is equal to or smaller than the threshold value, the keyword / search formula suggestion processing unit 5 performs the search formula proposal processing as the keyword X for proposing the word t (S25).

  Then, the keyword / search formula suggestion processing unit 5 determines whether or not the process of searching for keywords for all words has been completed (S26). Here, if the processing has been completed for all the words, the processing ends (S27). On the other hand, if the processing has not been completed for all words, the process returns to S22, and the same processing is executed until the processing is completed for all words.

If the similarity value is not less than or equal to the threshold value in S24, the process proceeds to a process (S26) for determining whether or not the process of searching for keywords for all words is completed.
Through the processing as described above, it is possible to extract a word that belongs to the same field and has a similar appearance transition with an appearance frequency in a certain section as a new keyword for a keyword registered by the user 20. In this way, it is possible to extract a phrase that has become related to the keyword registered by the user 20 due to temporal changes.

《Search formula suggestion processing flow》
FIG. 10 is a flowchart showing details of the search formula suggestion process shown in FIG. This search formula proposing process is a process for proposing to the user 20 a search formula that is a combination of the keywords extracted in the word information processing and the keywords used in the search (keywords registered by the user).

  First, the keyword / search formula suggestion processing unit 5 acquires the document ID list “L (W)” of the keyword W and the document ID list “L (X)” of the keyword X from the document / word correspondence list 10 (S31). .

Subsequently, the keyword / search formula suggestion processing unit 5 obtains a matching rate between the acquired document ID lists “L (W)” and “L (X)” (S32). This matching rate is, for example, a value obtained by multiplying a value obtained by squaring the absolute value of the logical product of document IDs existing in two document ID lists and the absolute value of the number of document IDs existing in each document ID list. What is necessary is just to obtain | require by dividing. In other words, a value obtained by squaring the number of documents (document ID) including both the keyword W and the keyword X is divided by the product of the number of documents (document ID) including either the keyword W or the keyword X. For example, in FIG. 2, in the document ID list of the words “mad cow disease” and “Takebe”, the document ID “1” and “105” match, so the logical product is 2. In this way, the ratio at which the two document ID lists overlap can be grasped from the obtained value.

  Subsequently, the keyword / search formula suggestion processing unit 5 determines whether or not the obtained matching rate value is equal to or greater than a threshold value (S33). If the matching rate is equal to or greater than the threshold, the proposed search formula is determined as a search formula Y consisting of the logical product (W and X) of the keyword W and the keyword X (S35). That is, among keywords extracted by word information processing, keywords that frequently appear in a document including a keyword registered by the user 20 are considered to be related words, and therefore, “a keyword registered by the user” and A search expression “related words to propose” is proposed. Here, keywords that are regarded as related words are used as narrowing conditions.

  On the other hand, if the matching rate is not equal to or higher than the threshold, the proposed search formula is determined as a search formula Y consisting of the logical sum (W or X) of the keyword W and the keyword X (S36). That is, among keywords extracted by word information processing, a keyword that frequently appears in a document that does not include the keyword registered by the user 20 may be a synonym of the word registered by the user. A search expression of a keyword “or“ proposed keyword ”registered by the user is proposed.

  Further, before proceeding to S35, it is determined whether or not the proposed keyword X is a synonym (paraphrase) of the keyword W registered in advance (S34). For this determination, for example, a synonym is registered in advance in a database or the like (for example, a synonym dictionary), and it is determined whether or not a keyword to be proposed is registered. Here, when the keyword X to be proposed is not a synonym (paraphrase) of the keyword W registered in advance, the process of proposing the search expression Y composed of the logical product (W and X) is continued. On the other hand, when the keyword X to be proposed is a synonym (paraphrase) of the keyword W registered in advance, the process of proposing a search expression is not executed.

  When suggesting any search formula to the user, the keyword / search formula proposal processing unit 5 registers the proposed search formula Y in the proposal search formula management database 14 and sets the status flag to “1 (proposed)”. Set (S37).

  Through the processing as described above, it is possible to propose a combination for a keyword registered in advance by the user 20 as a search expression, instead of simply proposing the extracted keyword as a related word.

  In addition, by suggesting synonyms (paraphrases) before proposing a search expression composed of logical products (S34), the keywords are not simply paraphrased but become related to the keywords as time passes. Can be proposed as related words.

《Search formula suggestion management processing flow》
FIG. 11 is a flowchart showing details of the search formula suggestion management process shown in FIG. This search formula proposal management process is a process executed mainly by the keyword proposal management processing unit 6 when a search formula to be newly proposed is registered in the proposal search formula management database 14.

First, the keyword proposal management processing unit 6 proposes the search formula Y to the corresponding user 20 (S41). That is, the keyword proposal management processing unit 6 identifies the user 20 having the user ID corresponding to the search formula Y from the proposal search formula management database 14 and transmits a proposal notification including the search formula Y to the user 20 (output) ) Hereinafter, the identified user 20 will be described as a user A.

  Upon receiving the proposal notification, the user A determines whether or not to adopt the search expression Y included in the notification, and inputs adoption or non-adoption by an input operation to the user terminal (S42). The input result of the user A is transmitted to the device 1 as a response notification for the proposal notification.

  Upon receiving the response notification, the keyword proposal management processing unit 6 of the device 1 determines whether or not the proposal result is adopted from the content (S43). Here, if the proposal result is adopted, the keyword proposal management processing unit 6 sets the status flag of the corresponding record (the record of the user A) in the proposal retrieval formula management database 14 to “2 (under evaluation)”. Update (S44).

  Further, the keyword proposal management processing unit 6 adds the search formula Y to the keyword condition for executing the clipping service for the user A (S45). That is, the keyword proposal management processing unit 6 registers the search expression Y in the clipping processing system 7.

  After a certain period of time has elapsed since the search expression Y was registered, the keyword proposal management processing unit 6 confirms with the user A whether or not to continue using the search expression Y (S46). Here, a confirmation notification for confirming whether or not to continue using is transmitted to the user A.

  Receiving the confirmation notification, user A determines whether or not to continue using the search expression Y included in the notification, and inputs end or continuation by an input operation to the user terminal (S47). The input result of the user A is transmitted to the device 1 as a response notification to the confirmation notification.

  Upon receiving the response notification, the keyword proposal management processing unit 6 of the device 1 determines whether or not the confirmation result is continued from the content (S48). Here, if the confirmation result is continuation, the keyword proposal management processing unit 6 updates the status flag of the corresponding record in the proposal retrieval formula management database 14 to “3 (adopted)” (S49). On the other hand, if the confirmation result is not continuation, the keyword proposal management processing unit 6 deletes the search expression Y registered in the clipping processing system (S50). At the same time, the keyword proposal management processing unit 6 sets the status flag of the corresponding record in the proposal retrieval formula management database 14 to “4 (not adopted)” (S51).

  In S43, even if the proposal result is not adopted, the keyword proposal management processing unit 6 sets the status flag of the corresponding record in the proposal retrieval formula management database 14 to “4 (not adopted)” (S44). .

  Through the processing as described above, keywords extracted by the user 20 can be combined with the extracted keywords and proposed as a search expression. In addition, when the proposed search formula is adopted by the user 20, the clipping service can be implemented using a keyword that takes into account temporal changes obtained from the search formula.

  Even if the proposed search formula is adopted, after a certain period of time, the service provider encourages the maintenance of keywords by asking the user whether to continue using the search formula. Can be implemented.

According to the present invention, new words / related words / search formulas can be proposed, so that the user's knowledge can be complemented and the keywords / search formulas can be updated to better ones. For this reason,
A clipping service can be implemented using keywords outside the range of knowledge the user has.

  According to the present invention, it is possible to propose a phrase having a new relevance using information generated in a time series as a related word, and thus it is possible to prevent the keyword from becoming obsolete. For this reason, for example, it is possible to quickly clip a new topic to be noticed such as news by using a word having a new relationship.

<Modification>
In the embodiment described above, it is determined whether or not the proposed keyword X is a synonym (paraphrase) of the keyword W registered in advance before determining the search expression to be proposed in S35. However, this system does not need to determine whether or not it is a synonym before determining the search expression Y proposed in S35.

  By adopting such a configuration, in this system, even if a phrase is originally treated as a synonym, a phrase that is not considered as a synonym by a user can be prevented from being excluded as a synonym. For this reason, in this system, it can respond to the difference in recognition of a synonym different for every user.

<Others>
The present invention can be specified as follows.
And means for classifying (Appendix 1) documents into categories,
Means for calculating a document frequency is the number of documents that contain that word in the same categories for words included in the document,
Means for recording a transition of the document frequency over a predetermined period;
Means for extracting a second word having a document frequency transition similar to the first word in the same category as a related word of the first word;
A server comprising
(Supplementary Note 2) The first word or for the number of documents that the related word is included, the first word and the associated word means both seeking coincidence degree is the number ratio of the document that contains the When,
Means for outputting a search expression including the first word and its related word according to the degree of coincidence;
The server according to appendix 1, comprising:
(Supplementary Note 3) When the degree of coincidence reaches a predetermined value, the outputting means outputs a search expression composed of a logical product of the first word and its related word, and the degree of coincidence does not reach the predetermined value. The server according to appendix 2, which outputs a search expression comprising a logical sum of the first word and its related word.
(Additional remark 4) Even if it is a case where the said coincidence reaches a predetermined value, when the said means to output corresponds to a synonym with which the 1st word and its related word are registered beforehand, the 1st word The server according to supplementary note 3, which does not output a search expression composed of a logical product of a word and its related word.
(Supplementary Note 5) Further comprising means for receiving registration of the first word from the user,
The means for outputting means for transmitting a notification including a search expression to a user who has registered the first word;
Means for accepting a response including adoption or non-adoption of a search expression from the user in response to the transmitted notification;
Means for transmitting a notification for confirming whether or not to continue using the search formula after elapse of a fixed period when adoption is included in the response;
The server according to appendix 1, wherein
(Supplementary note 6) The server according to any one of supplementary notes 1 to 5, wherein the first word is a registered word registered in advance in the server.
(Supplementary note 7) The server according to supplementary note 1, wherein the transition is an increase / decrease rate in a unit period.
(Appendix 8) The computer
The method comprising the steps of: classifying documents into categories,
Calculating a document frequency is the number of documents that contain that word in the same categories for words included in the document,
Recording the transition of the document frequency over a predetermined period;
Extracting a second word having a document frequency transition similar to the first word in the same category as a related word of the first word;
The related word suggestion method to execute.
(Supplementary note 9)
And determining the relative first word or the number of documents that related word is contained, the degree of coincidence is the ratio of the number of first words and documents that contain both the relevant language,
The related word suggestion method according to appendix 8, wherein a step of outputting a search expression including the first word and its related word according to the degree of coincidence is executed.
(Appendix 10) The computer
In the outputting step, when the degree of coincidence reaches a predetermined value, a search expression consisting of a logical product of the first word and its related word is output, and when the degree of coincidence does not reach the predetermined value, the first The related word proposal method according to supplementary note 9, wherein a search expression including a logical sum of the word and the related word is output.
(Supplementary Note 11) The computer
In the outputting step, even when the degree of coincidence reaches a predetermined value, if the first word and its related word are equivalent to pre-registered synonyms, the first word and its related word The related word suggestion method according to supplementary note 10, wherein a search expression consisting of a logical product of words is not output.
(Additional remark 12) The said computer receives registration of the said 1st word from a user,
In the step of outputting, the step of transmitting a notification including a search expression to a user who has registered the first word;
In response to the transmitted notification, receiving a response including adoption or non-adoption of a search expression from a user;
A step of sending a notification for confirming whether or not to continue to use the search formula after a certain period of time when adoption is included in the response;
The related word suggestion method according to attachment 11, further executing:

  The present invention is applicable to a service industry using various information devices such as a service that provides real-time information.

FIG. 1 is a diagram showing a system configuration of the information system. FIG. 2 is a diagram illustrating a data structure of the document / word correspondence list 10. FIG. 3 is a diagram showing the data structure of the category master database 11. FIG. 4 is a diagram showing the data structure of the classification information database 12. FIG. 5 is a diagram illustrating a data structure of the word information database 13. FIG. 6 is a diagram showing the data structure of the proposed search expression management database 14. FIG. 7 is a flowchart showing the keyword / search formula suggestion process. FIG. 8 is a flowchart showing details of the preprocessing shown in FIG. FIG. 9 is a flowchart showing details of the word information processing shown in FIG. FIG. 10 is a flowchart showing details of the search formula suggestion process shown in FIG. FIG. 11 is a flowchart showing details of the search formula suggestion management process shown in FIG. FIG. 12 is a diagram illustrating an example in which the appearance frequency of a keyword for each certain period is graphed.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Apparatus 2 Content collection system 3 Content classification processing part 4 Search index creation part 5 Keyword / search formula proposal processing part 6 Keyword proposal management processing part 7 Clipping processing system 8 Content database 9 Search database 10 Document / word correspondence list 11 Category master database 12 Classification information database 13 Word information database 14 Proposed search expression management database 20 User

Claims (5)

Based on the word contained in the document, using a classification dictionary that shows the relationship between the document, the word contained in the document, and the field to which the document belongs, by the relationship between the word and the field of the word. , means for classifying the documents in the category,
Means for calculating a document frequency for each unit interval is the number of documents that contain that word in the same categories for words included in the document,
And means for recording the estimated transfer of the document frequency in a plurality of the unit sections in the word information database,
Means for referring to the word information database and extracting a second word having a document frequency transition similar to the first word in the same category as a related word of the first word;
A server comprising For the number of the first word or document that related word is contained, it means for determining the degree of matching is the number ratio of the document that contains both the first word and its related terms,
Means for outputting a search expression including the first word and its related word according to the degree of coincidence;
The server according to claim 1, comprising: Means for accepting registration of the first word from a user;
The means for outputting means for transmitting a notification including a search expression to a user who has registered the first word;
Means for accepting a response including adoption or non-adoption of a search expression from the user in response to the transmitted notification;
Means for transmitting a notification for confirming whether or not to continue using the search formula after elapse of a fixed period when adoption is included in the response;
The server according to claim 1. Computer
Based on the word contained in the document, using a classification dictionary that shows the relationship between the document, the word contained in the document, and the field to which the document belongs, by the relationship between the word and the field of the word. , the method comprising the steps of: classifying the documents in the category,
Calculating a document frequency is the number of documents that contain that word in the same category for each unit interval for the word contained in the document,
And recording the estimated transfer of the document frequency in a plurality of the unit sections in the word information database,
Referring to the word information database and extracting a second word having a document frequency transition similar to the first word in the same category as a related word of the first word;
The related word suggestion method to execute. The computer
And determining the relative first word or the number of documents that related word is contained, the degree of coincidence is the ratio of the number of first words and documents that contain both the relevant language,
Outputting a search expression including the first word and its related word according to the degree of coincidence;
The related word suggestion method according to claim 4, wherein:

Images