JP2011253415A - Query suggestion device and method based on phrase - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a query suggestion device and method for displaying, to a user terminal, information for enabling a user to much more easily grasp association with his or her desired information including information which is not included in the past input history, retrieval log and retrieval index.SOLUTION: In this query suggestion device 20, a phrase DB generation means 250 stores a phrase extracted from an object document in a phrase DB260, and a matching DB generating means 270 calculates relevance scores between words included in a retrieval index or the like and the object document, and stores the relevance scores in association with the words and the object document in a matching DB280, and the query estimation means 212 estimates a query based on query input operation information received from a user terminal 10, and a relevant phrase extraction means 213 generates a suggestion query with the phase of the object document whose relevance with the estimated query is high, and transmits it to the user terminal 10.

Description

  The present invention relates to a phrase-based query suggestion apparatus and method.

  Conventionally, when performing a content search on the Internet, based on information related to a query input operation received by a browser or the like of a user terminal, a query input history stored by the user terminal or a query log or a search index stored in advance by a search server There is known a technique for displaying a suggestion query composed of an estimated query, a related word, a correction candidate for a variation in notation including a mistype, etc. on a terminal (for example, Patent Document 1, Non-Patent Document 1). etc).

  According to such a technique, a user can correct a requested query with reference to a suggestion query displayed in response to a query input operation, and can efficiently search for desired content.

JP 2009-104602 A

Netmarks Co., Ltd., “Google Search Appliance [Ver. 6.2..G14 Features]”, [online], Netmarks Co., Ltd. [Search April 30, 2010], Internet <URL: http: // www. netmarks-gsa-support. com / main_gsa. html>

  However, the above-described technology relies on information on the input history, query log, or search index, and it is impossible to present to the user what is not in the past input history, query log, and search index information. Furthermore, many of the past input histories, query logs, and search indexes are composed of words, morphemes, etc., and these words and morphemes can only have simple meanings. The visually recognized user may not be able to easily grasp the relationship with the desired information.

  Therefore, the present invention makes it easier for the user to grasp the relationship with the desired information, including information not included in the past input history, search log, and search index, based on the information related to the query input operation accepted by the user terminal. An object of the present invention is to provide a query suggestion apparatus and method for displaying information that can be performed on the user terminal.

  Specifically, the present invention provides the following.

  (1) A query suggestion device capable of communicating with a user terminal via a communication network, wherein a phrase is extracted from the target document received in response to receiving the target document, and is associated with the target document as a phrase DB For a word included in a stored phrase DB generating means and a query log or search index stored in advance, a higher relevance score is calculated as the relevance with the target document is higher, and the matching DB is associated with the word and the target document. Matching DB generation means stored as: means for receiving information related to query input operation from the user terminal; input by referring to the query log or the search index based on the received information related to the query input operation Query guessing means to guess the middle query, The matching DB is referred to based on the query, the target document having the same word as the query and the relevance score is extracted, and the query is referred to the phrase DB based on the extracted target document. A query suggestion device comprising: related phrase extraction means for extracting the phrase having a high degree of relevance to the user terminal; and suggestion query transmission means for transmitting the extracted phrase as a suggestion query to the user terminal.

  The query suggestion device having the configuration of (1) stores a phrase extracted from the target document as a phrase DB, further calculates a relevance score between the word included in the query log or the search index and the target document, and The word is stored as a matching DB in association with the target document. Furthermore, the query suggestion device estimates the query based on information related to the query input operation received from the user terminal, and further refers to the matching DB and the phrase DB based on the estimated query, thereby inputting the query A target document having a high degree of association with the query estimated according to the operation is extracted, and a phrase is extracted based on the extracted target document and transmitted to the user terminal.

  As a result, the query suggestion device can extract and transmit a phrase from a target document having a high degree of association with the query estimated according to the query input operation received by the user terminal. As a result, the user who visually recognizes the phrase displayed on the user terminal can easily grasp the relationship with the desired information and search for the desired document more efficiently.

  (2) The query suggestion device according to (1), wherein the matching DB generation unit calculates a higher relevance score for a characteristic word of the target document by TF-IDF.

  The query suggestion device having the configuration of (2) calculates a higher relevance score for a characteristic word of the target document by TF-IDF.

  Accordingly, the query suggestion device can extract a phrase related to a characteristic word in the target document with higher priority and transmit the phrase to the user terminal. As a result, the user can search the desired document more efficiently by visually recognizing the phrase related to the characteristic word.

  (3) Of the phrases stored in association with the target document extracted by the related phrase extraction unit with reference to the phrase DB, it is a word other than the query estimated by the query estimation unit, and the feature of the target document Support phrase extracting means for further extracting a phrase including a word is further provided, and the suggestion query transmitting means further adds the phrase extracted by the support phrase extracting means as the suggestion query and transmits it (1) or (2) Query suggestion device as described in.

  The query suggestion device having the configuration of (3) further extracts a phrase that is a word other than the estimated query and includes the characteristic word of the target document, and transmits it to the user terminal. Thereby, the query suggestion device can make the user of the user terminal visually recognize the support phrase including the characteristic word of the target document as the suggestion query in addition to the main phrase including the estimated query. Here, the characteristic words can be specified using various known techniques such as TF-IDF.

  As a result, the user can grasp the contents of the target document more easily by visually recognizing the support phrase including the characteristic words of the target document in addition to the main phrase, and perform an appropriate suggestion query selection operation. It is possible to easily grasp the relationship with desired information and search for a desired document more efficiently.

  (4) The query among the phrases stored in association with the target document extracted by the related phrase extraction unit with reference to the morpheme DB stored in association with the target document by dividing the target document into morpheme units The method further comprises support word extraction means for further extracting a feature word of the target document that is a word other than the query estimated by the estimation means, and the suggestion query transmission means further includes the word extracted by the support word extraction means. The query suggestion device according to (1) or (2), which is transmitted as a suggestion query.

  The query suggestion device having the configuration of (4) further extracts a feature word of the target document that is a word other than the estimated query and transmits it to the user terminal. Thereby, the query suggestion device can make the user of the user terminal visually recognize the characteristic words of the target document as a suggestion query in addition to the main phrase including the estimated query. Here, the characteristic words can be specified using various known techniques such as TF-IDF.

  As a result, in addition to the main phrase, the user can grasp the contents of the target document more easily by visually recognizing the characteristic words of the target document, perform an appropriate suggestion query selection operation, and select desired information and Thus, it is possible to easily grasp the relationship and search for a desired document more efficiently. Since feature words are shorter than phrases, they can be displayed more in the same display space, and when the main phrases and feature words are displayed in combination, the user may be able to narrow down the target documents to be linked more appropriately. is there.

  (5) A query suggestion device capable of communicating with a user terminal via a communication network performs query suggestion, wherein the query suggestion device receives a phrase from the target document received in response to receiving the target document. For a phrase DB generation step that is extracted and stored as a phrase DB in association with the target document, and a word included in a query log or search index stored in advance, a higher relevance score is calculated as the relevance with the target document is higher. , Based on the matching DB generation step of storing as a matching DB in association with the word and the target document, the step of receiving information related to a query input operation from the user terminal, and the information related to the received query input operation, The query log or the search event A query estimation step for estimating a query in the middle of input by referring to a dex, and referring to the matching DB based on the estimated query, and extracting the target document having the same word as the query and a high relevance score A phrase extracting step of extracting the phrase having a high degree of relevance to the query by referring to the phrase DB based on the extracted target document, and a suggestion query for transmitting the extracted phrase to the user terminal as a suggestion query A transmitting step.

  By performing the method described in (5), the same actions and effects as in (1) can be expected.

  ADVANTAGE OF THE INVENTION According to this invention, the query suggestion apparatus can extract and transmit the phrase containing the word highly relevant with the query estimated according to the query input operation which the user terminal received from the object document. As a result, the user who visually recognizes the phrase displayed on the user terminal can easily grasp the relationship with the desired information and search for the desired document more efficiently.

It is a figure which shows the functional block which concerns on an example of suitable embodiment of this invention. It is a figure which shows the functional block which concerns on another example of suitable embodiment of this invention. It is a flowchart which shows the phrase DB production | generation process which concerns on an example of suitable embodiment of this invention. It is a flowchart which shows the matching DB production | generation process which concerns on an example of suitable embodiment of this invention. It is a flowchart which shows the related phrase extraction process which concerns on an example of suitable embodiment of this invention. It is a flowchart which shows the related phrase extraction process which concerns on another example of suitable embodiment of this invention. It is a figure which shows an example of phrase DB which concerns on an example of suitable embodiment of this invention. It is a figure which shows an example of matching DB which concerns on an example of suitable embodiment of this invention. It is a figure which shows the screen image in the user terminal which concerns on an example of suitable embodiment of this invention. It is a figure which shows the screen image in the user terminal which concerns on an example of suitable embodiment of this invention. It is a figure which shows the screen image in the user terminal which concerns on an example of suitable embodiment of this invention. It is a figure which shows the screen image in the user terminal which concerns on an example of suitable embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described in detail.

  Note that the components in the preferred embodiment of the present invention can be appropriately replaced with existing components and the like, and various variations including combinations with other existing components are possible, The description of the preferred embodiments of the present invention is not intended to limit the content of the invention described in the claims.

  FIG. 1 is a block diagram illustrating a functional configuration of main devices including a user terminal 10 and a query suggestion device 20 according to an example of a preferred embodiment of the present invention. Each means provided in these devices is constituted by hardware provided in a computer and its peripheral devices, and software for controlling the hardware.

  In addition to the CPU, the hardware includes a storage unit, a communication unit, a display unit, and an input unit. Examples of the storage unit include a memory (RAM, ROM, etc.), a hard disk drive (HDD), and an optical disk (CD, DVD, etc.) drive. Examples of the communication unit include various wired and wireless interface devices. Examples of the display unit include various displays such as a liquid crystal display and a plasma display. Examples of the input unit include a keyboard and a pointing device (mouse, tracking ball, etc.).

  Here, the user terminal 10 includes a query input operation accepting unit 11, a query input operation information transmitting unit 12, a suggestion query receiving unit 13, and a suggestion query display unit 14. The query input operation accepting unit 11 accepts a query input operation from the user. The query input operation information transmitting unit 12 transmits query input operation information including information on the query character string being input to the query suggestion device 20 based on the query input operation received from the user by the query input operation receiving unit 11. The suggestion query receiving unit 13 receives a suggestion query transmitted from the query suggestion device 20. The suggestion query display unit 14 displays the suggestion query received by the suggestion query reception unit 13 from the query suggestion device 20.

  The query suggestion device 20 includes a suggestion query distribution unit 210, a search page request reception unit 220, a search page transmission unit 230, a target document reception unit 240, a phrase DB generation unit 250, a phrase DB 260, a matching DB generation unit 270, and a matching DB 280. And a reference DB group 290 including a query log DB 291, a morpheme dictionary DB 292 and an index DB 293.

  Further, the suggestion query distribution unit 210 includes a query input operation information reception unit 211, a query estimation unit 212, a related phrase extraction unit 213, a support phrase extraction unit 213a, and a suggestion query transmission unit 214.

  The suggestion query delivery unit 210 delivers a suggestion query to the user terminal 10 in response to receiving query input operation information including a query character string from the user terminal 10. The query input operation information reception unit 211 receives the query input operation information transmitted by the query input operation information transmission unit 12 of the user terminal 10. Based on the query input operation information received by the query input operation information receiving unit 211, the query estimation unit 212 estimates a candidate word of the query being input by referring to the query log DB 291 or the index DB 293. The related phrase extraction unit 213 refers to the matching DB 280 and the phrase DB 260 based on the query estimated by the query estimation unit 212 and extracts a phrase including the estimated query as a suggestion query. The support phrase extraction unit 213a refers to the phrase DB 260, and is a word other than the query candidate words estimated by the query estimation unit 212 among the phrases stored in association with the target document extracted by the related phrase extraction unit 213. Phrases including feature words of the target document are further extracted. The suggestion query transmission unit 214 transmits the phrase (main phrase) and the support phrase extracted by the related phrase extraction unit 213 and the support phrase extraction unit 213a to the user terminal 10 as a suggestion query.

  In addition, the search page request receiving unit 220 receives a search page request from the user terminal 10. The search page transmission unit 230 configures a search page and transmits it to the user terminal 10 in response to the search page request reception unit 220 receiving the search page request.

  On the other hand, the target document receiving unit 240 receives the target document transmitted from the news server 30.

  Next, the phrase DB generation unit 250 extracts a phrase from the target document received by the target document reception unit 240 and stores it as the phrase DB 260 in association with the target document. FIG. 7 shows an example of the phrase DB 260. The phrase DB 260 stores a phrase extracted from the document in association with the target document ID for identifying the target document.

  Next, the matching DB generation unit 270 refers to the morpheme dictionary DB 292 and divides the target document into morpheme units. Here, the morpheme dictionary DB 292 stores morphemes for morpheme analysis, and various known ones can be adopted. Furthermore, the matching DB generation unit 270 refers to the query log DB 291 or the index DB 293, calculates a higher relevance score for the words included in these DBs as the relevance with the target document is higher, and the word and the target document. And stored as a matching DB.

  Here, the query log DB 291 stores a history of past user queries and the like as a query log. The index DB 293 stores words (morphemes) as indexes for document search. Various types of query log DB 291 and index DB 293 can be employed, and their formats are not limited. The query log DB 291 and the index DB 293 preferably include words included in the target document and include many queries estimated in the related phrase extraction process described later.

  FIG. 8 shows an example of the matching DB 280. The matching DB 280 stores the relevance score in association with the target document ID that identifies the word and the target document.

  The news server 30 receives a news article and transmits it as a target document to the query suggestion device 20. Then, the target document receiving unit 240 of the query suggestion device 20 receives the target document. The query suggestion device 20 extracts a suggestion query including a phrase described in detail below and transmits it to the user terminal 10 for the target document received in this way. Note that various modes can be adopted for the reception timing of the target document, and the news server 30 may transmit the target document every time it accepts submission of a news article, or may transmit it every predetermined time. Good.

  Further, the query suggestion device 20 itself may accept an article as a target document. Further, the target document is not limited to the news article, and various other articles such as a blog article can be adopted as the target document.

  As described above, in this embodiment, various articles can be handled as target documents at various timings. However, after an article is released, it is accepted as the target document as soon as possible and a suggestion query is sent to the user terminal 10. It is desirable to be able to transmit to.

  FIG. 2 is a block diagram illustrating a functional configuration of main devices including the user terminal 10 and the query suggestion device 20 according to another example of the preferred embodiment of the present invention. Description of parts common to FIG. 1 will be omitted as appropriate.

  In this embodiment, the query suggestion device 20 includes a support word extraction unit 213b instead of the support phrase extraction unit 213a. Further, a morpheme DB 280b is provided.

  The morpheme DB 280b divides the target document into morpheme units and stores them in association with the target document. In FIG. 2, the matching DB generation unit 270 is described as being created, but the present invention is not limited to this, and may be generated separately. Based on the target document, it can be created using various known morphological analysis engines. Further, as a specific configuration example of the morpheme DB 280b, although not illustrated, for example, the morpheme DB 280b is configured to store the morpheme constituting the corresponding target document in association with the target document ID indicating the target document. Can do.

The support word extraction unit 213b refers to the morpheme DB 280b and is a word other than the query candidate words estimated by the query estimation unit 212 among the phrases stored in association with the target document extracted by the related phrase extraction unit 213. The feature words of the target document are further extracted. The suggestion query transmission unit 214 transmits the phrase (main phrase) and the support word extracted by the related phrase extraction unit 213 and the support word extraction unit 213b to the user terminal 10 as a suggestion query.
[Phrase DB generation processing]

  FIG. 3 is a flowchart showing the procedure of phrase DB generation processing by the phrase DB generation unit 250 according to an example of the preferred embodiment of the present invention.

  Various start timings for the phrase DB generation process can be used. Specifically, the processing may be started by the phrase DB generating unit 250 responding each time the target document receiving unit 240 receives the target document. The target document receiving unit 240 temporarily stores the target document, You may start at or any time. Here, the former is desirable in that the accepted target document can be the subject of a suggestion query at an earlier timing.

  First, the phrase DB generation unit 250 receives one target document from the target document reception unit 240, and divides the target document into phrases (step S11).

  Next, the phrase DB generation unit 250 stores a phrase group for one target document in the phrase DB 260 in association with the target document ID for identifying the target document (step S12).

  Furthermore, the phrase DB generation unit 250 determines whether all the target documents received by the target document reception unit 240 have been processed. If all cases have not been processed, the process proceeds to step S11. If all cases have been processed, the process ends (step S13).

FIG. 6 is a diagram illustrating an example of a phrase list stored in the phrase DB 260 according to the present embodiment.
In the phrase DB 260, a target document ID for identifying the divided target document and a phrase included in the target document are stored in association with each other by separating them with “/” when there are one or more phrases. Yes.
In addition, it may be configured such that one phrase is provided for each phrase, instead of separating with “/” or the like and making one phrase with a plurality of phrases.
[Matching DB generation process]

  FIG. 4 is a flowchart showing a matching DB generation process performed by the matching DB generation unit 270 according to an example of the preferred embodiment of the present invention.

  The matching DB generation process may be started by transmitting the target document submitted from the news server 30 and receiving it by the target document receiving unit 240, and the matching DB generating unit 270 responding thereto. The document reception unit 240 may temporarily start the target document and then start at a predetermined or arbitrary timing. The matching DB generation unit 270 may perform the matching DB generation process at various timings as described above, but uses the same target document ID as the target document ID assigned to the target document in the phrase DB generation process described above. Is a requirement.

  Further, the query suggestion device 20 itself may accept an article as a target document. Further, the target document is not limited to the news article, and various other articles such as a blog article can be adopted as the target document.

  First, the matching DB generation unit 270 receives one target document from the target document reception unit 240, and divides the target document into morpheme units with reference to the morpheme dictionary DB 292 (step S21). At this time, as described above, the divided morpheme may be stored in association with the target document including the morpheme to create the morpheme DB 280b.

  Next, the matching DB generation unit 270 refers to the query log DB 291 or the index DB 293, extracts the words included in these from the divided morphemes, and determines the degree of association between the extracted words and the target document. A higher relevance score is calculated as the value is higher (steps S22 and S23).

  Furthermore, the matching DB generation unit 270 stores the relevance score in the matching DB 280 in association with the target document ID that identifies the word and the target document (step S24).

  Here, the matching DB generation unit 270 can calculate the relevance score by various known methods. However, in step S23, the matching DB generation unit 270 can further calculate the relevance score with respect to characteristic words of the target document by TF-IDF. A high relevance score may be calculated. In this way, when there are a plurality of words having a relevance score for a certain target document, the matching DB generation unit 270 calculates a higher relevance score for the characteristic word of the target document. And memorize it. Thereby, in the related phrase extraction process described later, the query suggestion device 20 can preferentially transmit a phrase including a characteristic word of the target document to the user terminal 10.

  Further, the matching DB generation unit 270 determines whether all the target documents received from the target document reception unit 240 have been processed. If all cases have not been processed, the process proceeds to step S21. If all cases have been processed, the process ends (step S25).

FIG. 8 is a diagram illustrating an example of the matching table stored in the matching DB 280 according to the present embodiment. In the matching DB, the words included in the target document, the target document ID of the target document, and the degree of association thereof are associated and stored as a matching table. In the example of FIG. 8, the relevance scores are “69” and “75” in the target document with the same target document ID “2256” having the same word “La * s” and “cell Δoh * la * s”. It is remembered that In addition, when the word “cell △ o ra * su” is a characteristic word of the target document with the target document ID “2256” rather than “La * su”, the relevance score is calculated and stored higher. May be.
[Related phrase extraction processing]

  FIG. 5 is a flowchart showing a procedure of related phrase extraction processing according to an example of the preferred embodiment of the present invention.

  In the related phrase extraction process, when the user performs a query input operation on the search page displayed on the user terminal 10, the query input operation reception unit 11 receives the query input operation, and the query input operation information transmission unit 12 A query character string related to the query input operation is transmitted to the query suggestion device 20 as query input operation information, and the query input operation information reception unit 211 receives the query input operation information, and the query estimation unit 212 responds accordingly to the query log. The query is entered by referring to the DB 291 or the index DB 293, and the related phrase extracting unit 213 responds to this to start.

The related phrase extraction unit 213 first determines whether the number of query candidate words estimated by the query estimation unit 212 is one or more. If it is one or more, the process proceeds to step S32, and if it is zero, the process is terminated (step S31).
In addition, when the number of candidate words of the query exceeds the maximum display number of the suggestion query display unit 14 of the user terminal 10, the query suggestion device 20 appropriately selects the query candidate words so as to be included in the maximum display number. You may narrow down.

  Next, the related phrase extraction unit 213 refers to the matching DB 280 based on the query candidate word, and acquires a target document ID having a high degree of relevance to the candidate word (step S32).

Furthermore, a phrase is extracted by referring to the phrase DB 260 from a highly relevant word-target document ID pair (step S33).
The related phrase extraction unit 213 may appropriately adjust the number of phrase extractions so as to reach the maximum display number of the suggestion query display unit 14 of the user terminal 10. Moreover, the related phrase extraction means 213 may unconditionally exclude those whose relevance score is equal to or lower than a predetermined score regardless of the maximum number of display cases.

  Here, the suggestion query delivery unit 210 refers to the phrase DB 260, and among the phrases stored in association with the target document extracted by the related phrase extraction unit 213, words other than the query candidate words estimated by the query estimation unit 212 In addition, support phrase extracting means 213a for further extracting a phrase including the characteristic word of the target document may be further provided (see FIG. 1) (step S34). The feature words can be extracted by appropriately adopting a known technique such as TF-IDF.

  The related phrase extracted by the related phrase extraction process described above is transmitted to the user terminal 10 as a suggestion query when the suggestion query transmission unit 214 responds to the end of the related phrase extraction process, and a suggestion query of the user terminal 10 is added thereto. The reception means 13 receives the response, and the suggestion query display means 14 of the user terminal 10 responds to display the suggestion query on the user terminal 10.

  FIG. 6 is a flowchart showing a procedure of related phrase extraction processing according to another example of the preferred embodiment of the present invention.

  The description of the same part as the related phrase extraction process described in FIG. 5 is omitted as appropriate. The processing from step S41 to step S43 in FIG. 6 is the same as the processing from step S31 to step S33 in FIG.

  Here, the suggestion query distribution unit 210 refers to the morpheme DB 280b, and among the phrases stored in association with the target document extracted by the related phrase extraction unit 213, words other than the query candidate words estimated by the query estimation unit 212 In addition, a support word extraction unit 213b that further extracts feature words of the target document may be further provided (see FIG. 2) (step S44). The feature words can be extracted by appropriately adopting a known technique such as TF-IDF.

  The related phrase extracted by the related phrase extraction process described above is transmitted to the user terminal 10 as a suggestion query when the suggestion query transmission unit 214 responds to the end of the related phrase extraction process, and a suggestion query of the user terminal 10 is added thereto. The reception means 13 receives the response, and the suggestion query display means 14 of the user terminal 10 responds to display the suggestion query on the user terminal 10. In this case, in the case of FIG. 5, the suggestion query is composed of phrases, whereas in the case of FIG. 6, the suggestion query is composed of phrases and feature words.

  Here, the feature word is shorter than the phrase, so that it can be displayed in the same display space, and when the main phrase and the feature word are displayed in combination, the user can more appropriately narrow down the target document of the link destination. There is also. A specific display example will be described later.

9, 10 and 11 are screen images on the user terminal 10 according to an example of the preferred embodiment of the present invention.
FIG. 9 is a screen image when the user performs a query input operation and inputs up to “La ○”, and FIG. 10 is a screen image when one character is input and input up to “La ○”. is there.

  First, FIG. 9 will be described. The user performs a query input operation using the query input operation accepting unit 11, and the query input operation information transmitting unit 12 transmits information input “R” to the query suggestion device 20.

  This query input information is received by the query input operation information accepting means 211 of the suggestion query delivery means 210 of the query suggestion device 20, and as a result, the query guessing means 212 accepts “La * s”, “La * s President”, “La Four words “○ -Z” and “La- ー” are extracted as query candidate words.

  As a result, the related phrase extracting unit 213 refers to the matching DB 280 using the above four candidate words as DB reference keys, and the target document IDs “0123”, “0124”, “2256”, “3598” regarding “Las”. , “8996”, “9151”, etc., corresponding relevance scores “89”, “92”, “69”, “57”, “59”, “44”, etc. IDs “2773”, etc., corresponding relevance scores “64”, etc., “L. President” related document IDs “6621”, “7344” etc., corresponding relevance scores “51”, “47”, etc. To get.

  The related phrase extraction unit 213 further refers to the phrase DB 260 using “0123”, “0124”, “2256”, and “2773” having high relevance among the acquired target document IDs as DB reference keys, and the phrase list “DF cell Δ "O La * s / SB / Defensive Ability / CB / ...", "La * s △ Wei / official site / profile / video / ...", "Sell △ Oh La * s / with autograph / Acquired “Photo / Sales / ...” “La-Zones / Quad Punk Rock Band / 1974 Formation / ...”.

  The target document ID, the degree of association, and the phrase list may be acquired at once by combining the matching DB 280 and the phrase DB 260 with the target document ID.

  The related phrase extraction unit 213 extracts a phrase as described above, and the suggestion query transmission unit 214 transmits the phrase extracted by the related phrase extraction unit 213 to the user terminal 10 as a suggestion query.

  Here, when the support phrase extraction unit 213a further extracts a support phrase or when the support word extraction unit further extracts a support word, the suggestion query transmission unit 214 adds to the phrase extracted by the related phrase extraction unit 213. The support phrase extracted by the support phrase extraction unit 213a or the support word extracted by the support word extraction unit may be further added and transmitted to the user terminal 10 as a suggestion query.

  FIG. 9 is an example of a screen image when the suggestion query receiving unit 13 of the user terminal 10 receives the suggestion query and the suggestion query display unit 14 displays the suggestion query. In the example of FIG. 9, the phrase “main phrase” extracted by the related phrase extraction unit 213 from the target document with the target document ID “0124” based on the query candidate word “La * s”. The other “official site”, “profile”, and “moving image” are the support phrases extracted by the support phrase extracting means 213a. Similarly, “cell △ o ra * su” is a phrase (main phrase) extracted by the related phrase extraction means 213 based on the query candidate word “La * su” from the target document with the target document ID “2256”. Yes, “official site”, “profile” and “moving image” other than that are the support phrases extracted by the support phrase extraction means 213a. As described above, it is desirable to appropriately adjust the phrase display mode so that the user can easily understand the phrase. Hereinafter, other display modes will be described.

  Next, FIG. 10 will be described. FIG. 10 further shows the display mode of the suggestion query when the user further performs a one-character query input operation using the query input operation accepting unit 11 and inputs “La ○ su” by proceeding further from FIG. 9.

  Similarly to the case of FIG. 9, the query guessing unit 212 guesses “La * s” and “La * s President” as the query candidate words.

  The related phrase extracting unit 213 refers to the matching DB 280 by using the two candidate words as DB reference keys, and the target document IDs “0123”, “0124”, “2256”, “3598”, “8996” regarding “Las”. ”,“ 9151 ”, etc., the corresponding relevance scores“ 89 ”,“ 92 ”,“ 69 ”,“ 57 ”,“ 59 ”,“ 44 ”, etc. Correspondence degree scores “51”, “47” and the like such as “6621” and “7344” are acquired. The related phrase extracting unit 213 further refers to the phrase DB 260 by using “0123”, “0124”, “2256”, “6621”, “7344”, and “8996”, which have a high degree of association, among the acquired target document IDs. Phrase list "DF cell △ o ra * su / SB / defense ability / CB / ...", "La * s △ wei / official site / profile / video / ...", "cell △ o * "La * su / autographed / photo / sales / ...", "La * President / Phi △ Le La * / Former President of the Philippines / ...", "La * President / Jo △ La ○ Shorta / East Timor / ... ”,“ La ○ s △ Wei / Japan National Soccer Team Director ... ”etc.

  FIG. 10 is an example of a screen image when the suggestion query receiving unit 13 of the user terminal 10 receives the suggestion query and the suggestion query display unit 14 displays the suggestion query. Suggestion generated based on the phrase list of the phrase DB 260 for the target document ID having low relevance in the matching DB 280, which is not displayed in FIG. Indicates that the query is displayed. In this example, the suggestion query display means 14 also displays the query candidate words themselves in a table of contents added to the head of each phrase. In this case, it is needless to say that the suggestion query transmitting unit 214 needs to transmit the candidate words of the query to the user terminal 10 as a suggestion query.

  Next, FIG. 11 will be described. FIG. 11 is an example of a screen image when the suggestion query display means 14 displays a suggestion query when the user inputs “Rakusu” from the query input operation acceptance means 11 as in FIG. 10.

  In the case of FIG. 10, suggestion queries for only the estimated candidate words “La * s” and “La * s President” are displayed in the first and sixth cases, but FIG. 11 is a screen image when this is omitted. It is an example.

  Finally, FIG. 12 will be described. FIG. 12 is an example of a screen image in the case where the suggestion query display means 14 displays a suggestion query when the user inputs “Rakusou” by the query input operation acceptance means 11 as in FIG. 10 and FIG. 11. is there.

  In the case of FIG. 10, a main phrase including the estimated query and a support phrase including characteristic words are displayed following the estimated candidate words “La * s” and “President La * s”. Is an example of a screen image when a characteristic word itself is displayed as a support word instead of a support phrase. In the case of FIG. 10, “Las △ △ Wei” is displayed as the main phrase and “official site”, “profile” and “video” are displayed as the support phrase, whereas in FIG. “Las △ Wei” and “official”, “profile”, “video”, “win” and “final” are displayed as support words. As described above, the feature word is shorter than the phrase, so that it can be displayed in the same display space, and when the main phrase and the feature word are displayed in combination, the user can more appropriately narrow down the target document of the link destination. In some cases.

DESCRIPTION OF SYMBOLS 10 User terminal 11 Query input operation reception means 12 Query input operation information transmission means 13 Suggestion query reception means 14 Suggestion query display means 20 Query suggestion apparatus 30 News server 210 Suggestion query distribution means 211 Query input operation information reception means 212 Query estimation means 213 Related phrase extraction means 213a Support phrase extraction means 213b Support word extraction means 214 Suggestion query transmission means 220 Search page request reception means 230 Search page transmission means 240 Target document reception means 250 Phrase DB generation means 260 Phrase DB
270 Matching DB generation means 280 Matching DB
280b Morphological DB
290 Reference DB group 291 Query log DB
292 Morphological Dictionary DB
293 Index DB

Claims (5)

A query suggestion device capable of communicating with a user terminal via a communication network,
Phrase DB generating means for extracting a phrase from the received target document in response to receiving the target document and storing it as a phrase DB in association with the target document;
A matching DB generation unit that calculates a higher relevance score for a word included in a query log or a search index stored in advance and that has a higher relevance with the target document and stores it as a matching DB in association with the word and the target document; ,
Means for receiving information relating to a query input operation from the user terminal;
Query estimation means for estimating a query in the middle of input by referring to the query log or the search index based on the received information related to the query input operation;
The matching DB is referred to based on the estimated query, the target document having the same word as the query and the relevance score is extracted, and the phrase DB is referred to based on the extracted target document. A related phrase extracting means for extracting the phrase highly relevant to the query;
A query suggestion device comprising: a suggestion query transmission unit that transmits the extracted phrase as a suggestion query to the user terminal.   The query suggestion device according to claim 1, wherein the matching DB generation unit calculates a higher relevance score for a characteristic word of the target document by TF-IDF. Among phrases stored in association with the target document extracted by the related phrase extraction unit with reference to the phrase DB, the phrase is a word other than the query estimated by the query estimation unit and includes a characteristic word of the target document Further comprising a support phrase extracting means for further extracting a phrase;
The query suggestion device according to claim 1, wherein the suggestion query transmission unit further adds and transmits the phrase extracted by the support phrase extraction unit as the suggestion query. Of the phrases stored in association with the target document extracted by the related phrase extraction unit with reference to the morpheme DB stored in association with the target document by dividing the target document into morpheme units, the query estimation unit A support word extracting unit that further extracts a feature word of the target document that is a word other than the estimated query;
The query suggestion device according to claim 1, wherein the suggestion query transmission unit further adds the word extracted by the support word extraction unit as the suggestion query and transmits the query. A query suggestion device capable of communicating with a user terminal via a communication network performs a query suggestion, wherein the query suggestion device comprises:
A phrase DB generation step of extracting a phrase from the received target document in response to receiving the target document and storing it as a phrase DB in association with the target document;
A matching DB generation step of calculating a higher relevance score for a word included in a query log or a search index stored in advance and a higher relevance score with respect to the target document, and storing it as a matching DB in association with the word and the target document; ,
Receiving information related to a query input operation from the user terminal;
Based on the received information related to the query input operation, a query estimation step of estimating a query in the middle of input by referring to the query log or the search index;
The matching DB is referred to based on the estimated query, the target document having the same word as the query and the relevance score is extracted, and the phrase DB is referred to based on the extracted target document. A related phrase extracting step of extracting the phrase that is highly relevant to the query;
A suggestion query transmission step of transmitting the extracted phrase as a suggestion query to the user terminal.

Images