JP2012068687A - Information processing device, method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To determine a type of a noun whose type is unclear from a given context.SOLUTION: On the basis of a context relating to a noun in a search result of a search engine, a type of even a noun whose type is unclear from a given context can be determined. Therefore, the determined type can be utilized for the optimization of an information processing result based on the intention of a user by, for example, presenting a related search word in a related search or reflecting it on the control of a listing order of a Web search result. For the prescribed number of summaries, a quantity determined according to the fact that a noun to be determined is used with a corresponding pattern and so forth is tabulated for the each type. The type with a larger quantity is then preferentially output as a determination result by being transferred for post-processing, for example. As for even nouns which are pertinent to a plurality of types, the most general type or primary types down to several ranks can be used as the determination result, therefore it is possible to optimize the information processing result matching the intention of many users.

Description

  The present invention relates to an improvement in language processing technology.

  Conventionally, in the field of language information processing such as Japanese, a device for optimizing information processing results such as automatic translation (see, for example, Non-Patent Document 1) and search according to the user's intention is known. For example, the present applicant provides a service (“related search”) that presents words that are often combined as related search words based on search terms such as nouns (also referred to as “search queries”) entered by the user in web search. (See, for example, Non-Patent Document 2). There is also known a proposal of analyzing the intention of a user's question by repeating questions with the user and adjusting the amount of search results, the presentation format, and the like (see, for example, Patent Document 1).

Yahoo Japan Corporation, "Yahoo! Translation", [online], [Search July 16, 2010], Internet <URL: http://honyaku.yahoo.co.jp/> Yahoo Japan Corporation, "Yahoo! Search Help-What is" Related Search Word "", [online], [Search July 16, 2010], Internet <URL: http://help.yahoo.co.jp/help /jp/search/web/web-17.html> Akinori Takemura et al., “Statistics of Language and Psychology”, 1st edition, Iwanami Shoten Co., Ltd., March 2003, p.74-86

JP 2002-288201 A

  However, if nouns are included in the search query entered by web search or nouns included in the original text of automatic translation using morphological analysis, the type of noun is There is a problem that optimization of information processing results is often difficult because it is unknown.

  For example, if the noun “Kiruna” is not widely known and only the context “Go to Kieluna” is given, then this context alone is the type of the noun “Kiruna” (for example, place name, person name, organization In this case, if the type of the noun can be determined, the information processing result can be optimized according to the user's intention.

  In this regard, the technique of Non-Patent Document 2 can present related search words for frequently input nouns based on a large number of input examples, but it is difficult to effectively deal with other nouns. Further, the technique of Patent Document 1 is based on questions with the user, and it has been difficult to apply to a process that provides a search result in response to an input like a web search.

  In view of the above problems, an object of the present invention is to determine a type of a noun whose type is unknown from a given context.

  Based on the above object, an information processing apparatus according to one aspect (1) of the present invention includes a noun acquisition means for acquiring a determination target noun given as a language processing target, and a web search based on the acquired noun It comprises search result acquisition means for acquiring results, and type determination means for determining the type of the noun based on the context relating to the noun in the acquired web search result.

  Another aspect (5) of the present invention is an information processing method that captures the above aspect in a method category, in which a computer acquires a noun to be determined that is given as a language processing target, A search result acquisition process in which the computer acquires a web search result based on the acquired noun, and a type determination process in which the computer determines a type of the noun based on a context regarding the noun in the acquired web search result It is characterized by including these.

  Another aspect (6) of the present invention is a computer program that captures the above aspects in the category of a computer program, and that determines the type of a noun by controlling the computer. A determination target noun given as an object is acquired, a computer is caused to acquire a web search result based on the acquired noun, and the computer is caused to acquire the noun based on a context related to the noun in the acquired web search result It is characterized in that the type of is determined.

  In this way, the noun type can be determined even for a noun whose type is unknown from the given context, based on the noun context in the search engine search results. It can be used for optimizing information processing results based on the user's intention, such as word presentation and reflection in the control of the list order of web search results.

  According to another aspect (2) of the present invention, in any one of the above aspects, the quantity for each type determined by the type determination unit for a predetermined number of elements included in the acquired web search result is aggregated. And a result output means for preferentially outputting the type having a large number of the collected quantities as a determination result.

  In this way, for a predetermined number of summary and other elements such as a snippet that is a character string in which a part including a search word is extracted for each URL, a noun to be determined is used in a corresponding pattern for each type. The most common types or numbers can be obtained for nouns that correspond to multiple types by summing up the quantities determined in (1) and outputting them in the form of giving priority to the post-processing as the determination result. Since the first major types can be used as determination results, it is possible to optimize information processing results that match the intentions of many users.

  According to another aspect (3) of the present invention, in any one of the aspects described above, based on the type determined for the noun, as a candidate for a word specified as a search query simultaneously with the noun, a predetermined according to the type The present invention is characterized in that a related presentation means for presenting the words is provided.

  In this way, the selection of a search query that is more suitable for the user's intention can be effectively performed by presenting a word or the like previously associated with a noun type as a related search word in a so-called related search based on the determined noun type. Can help.

  In another aspect (4) of the present invention, in any one of the above aspects, the information item included in the web search result based on the noun includes an expression relating to the same type as the type determined for the noun. And a list order control means for controlling the list order of the information items.

  In this way, by controlling the list order of information items such as URLs included in the web search result based on the determined noun type, the web search result suitable for the user's intention can be scrolled to the minimum. Web browsing is improved, and the usability of web search is improved.

  It should be noted that a category (method for the apparatus, program for the method, etc.) different from each of the above-described modes and more specific modes described below are also included in the present invention. For different categories, “means” shall be appropriately read as “process” or “step”. Further, the order of execution of processes and steps is not limited to the above, and can be changed as appropriate or can be processed collectively.

  According to the present invention, it is possible to determine the type of a noun whose type is unknown from a given context.

The functional block diagram which shows the structure of embodiment of this invention. The figure which illustrates the information (data) used by embodiment of this invention. The flowchart which shows the process sequence in embodiment of this invention. The figure which shows the example of a screen in embodiment of this invention. The figure which shows the example of a screen in embodiment of this invention.

  Next, modes for carrying out the present invention (referred to as “embodiments”) will be described with reference to the drawings. It should be noted that assumptions common to those already described in the background art and problems are omitted as appropriate.

[1. Constitution〕
The present embodiment relates to the information processing apparatus 1 shown in FIG. 1 (hereinafter abbreviated as “this apparatus 1” or “this apparatus”), and this apparatus 1 is a back end of an automatic translation server 2 and a web search server 3. Are received from the automatic translation server 2 and the web search server 3 and may be returned as their types. The automatic translation server 2 and the web search server 3 each receive access from a user terminal T via a communication network N (Internet, cellular phone network, LAN, etc.) and provide automatic translation and web search services. It is.

  The present apparatus 1 (same as the servers 2 and 3) has at least an arithmetic control unit 6 such as a CPU, a storage device 7 such as an external storage device (HDD or the like) or a main memory, and a communication network as a general computer configuration. N communication means 8 (LAN adapter, communication gateway device, etc.). In the apparatus 1, a predetermined computer program (not shown) stored (installed) in advance in the storage device 7 controls the arithmetic control unit 6 so that elements (10, 20, etc.) shown in FIG. Is realized. Among these elements, the information storage means can be realized in the storage device 7 in any format such as various databases (also referred to as “DB”), variables such as files and arrays, various stacks and registers, and system setting values.

  Among such storage means, the pattern storage means 25 associates determination words used for determination and associations between these determination words and nouns for each type of noun, as illustrated in FIG. 2 (1). It is means for storing data (referred to as “pattern data”). In addition, the related search word storage unit 35 of the web search server 3 associates data (referred to as “related search data”) in which words presented as related search words are associated with each type determined for the noun. It is a memorizing means. Each means other than the storage means is a processing means for realizing and executing the following information processing functions and operations.

[2. (Determination of type)
FIG. 3 is a flowchart showing a processing procedure in the operation of determining the type of noun by the apparatus 1 configured as described above.
[2-1. (Acquiring nouns)
That is, the noun acquisition unit 10 acquires the noun to be determined given as the language processing target (step S1). In the example of FIG. 1, the determination target noun is acquired by receiving the determination target noun transmitted from the automatic translation server 2 or the web search server 3. Or a noun group whose type is to be determined, such as a search query in a search log, may be prepared in a file or the like, and the determination process may be performed collectively by receiving an upload or batch processing.

[2-2. (Get search results)
Subsequently, the search result acquisition means 20 acquires a web search result based on the acquired noun (step S2). Specifically, the search keyword input unit 15 inputs the acquired noun into a predetermined search engine (that is, a web search service) realized by the web search server 3 or the like, and the search result acquisition unit 20 sets the noun as the noun. A web search result by a search engine or the like is obtained (step S2). FIG. 4 is a diagram illustrating an example of a search result display screen when a noun (for example, “Kiruna”) is input as a search query.

[2-3. (Determination of type)
Then, the type determining means 30 determines the type of the noun based on the context relating to the noun in the acquired web search result, for example, by applying a specific expression extraction technique based on morphological analysis (step S3). Here, the proper expression extraction technology (NE technology) is based on morphological analysis to determine whether a given word is a suffix or prefix, and determine the type of a noun according to the co-occurrence with the noun. For example, as described in Non-Patent Document 3, eigen-expression extraction using a hidden Markov model or a maximum entropy model is known. Here, since the target noun has already been specified, only the portion of the algorithm for determining the type of the noun may be used or not necessarily used for the existing proper expression extraction technology. .

  In addition, types of nouns to be judged include organization name (ORGANIZATION), person name (PERSON), place name (LOCATION), date expression (DATE), time expression (TIME), monetary expression (MONEY), percentage expression (PERCENT), Although the unique object name (ARTIFACT) is also handled by the unique expression extraction technique, in the present embodiment, a part or all of these types are arbitrarily selected and set as a determination target.

  As a specific example, using pattern data as illustrated in FIG. 2 (1), for example, nouns with suffixes such as “•• prefecture”, “•• departure”, “•• travel” are determined as place names. , The nouns with suffixes such as "・ ・ さ ん", "・ ・ house", "・ ・ 流" are judged as human names, or nouns with "corporation", "corporate association" etc. are judged as organization names, etc. Can be considered. For example, in the example of FIG. 4, in the summaries of so-called snippets for each URL of the web page included in the web search result, the determination words K such as “Kiruna” “Go” and “Kiruna“ Departure ”are suffixes. Based on the pattern data (FIG. 2 (1)), the type of noun can be determined as “place name”.

  Further, the web search result and its elements as the basis for the determination are not limited to the summary such as a snippet, but the contents of the linked web page included in the search result may be used.

[3. (Type output and use)
If the type determined as described above is output to the automatic translation server 2 and utilized, the translation accuracy is improved. For example, in automatic translation, when there is a word in the original text that is not in the translation dictionary, or when there is a word in the original text that cannot be categorized from the given original text, based on the determination in the apparatus 1 By performing the translation process, the accuracy of the determination result is improved. As an example, for example, when the sentence “Kiruna is famous for beauty” is translated into English, “Kiruna is famous for its beauty.” If Kiruna is a place name, “Kiruna is famous for her ( / his) beauty. ” Thus, it can be expected that the translation becomes more accurate by knowing the type of “Kiruna”.

  Moreover, although the effect | action which outputs the type determined as mentioned above to the web search server 3 and utilizes for a web search is demonstrated below, the use of the determined type is not limited to these, Various Japanese processing services Needless to say, it is optional, for example.

[3-1. (Aggregation and output)
First, as for types, it is possible to determine and output only a single type, but it is desirable to determine a plurality of possible types and output the result of aggregation. Specifically, the determination result totaling unit 40 determines the predetermined number of elements (for example, summaries such as snippets) included in the web search result acquired by the search result acquisition unit 20 by the type determination unit 30. The quantity for each type is totaled (step S4), and the result output means 50 gives priority to the type with the large total quantity and outputs it as a determination result (step S5).

  For example, among the web search results based on the noun “Kiruna”, tentatively, in the top 100 URLs whose summary includes the determination word, there are 83 examples including the determination word as “place name”, and the determination as “person name” If there are 11 examples including "6" and 6 examples including the determination word "organization name", the "place name" with the largest number of cases is output as a noun type.

[3-2. (Use for related search words)
One of the uses of the determined type is related search, that is, presentation of related search words. In this case, based on the type determined for the noun, the related presentation unit 60 presents a predetermined word corresponding to the type as a word candidate to be specified as a search query simultaneously with the noun. For example, when the type of a noun, which is a search query entered in the keyword input field of web search, is determined as a place name according to the present invention, a word relating to a hotel or a trip associated with the type in advance (for example, FIG. ), Etc.) can be selected and presented as related search words.

  In the example of FIG. 5, when the user inputs “Kiruna” and a search word in the input field F of the web search page by the web search server 3, the type determined by passing this “Kiruna” to the present apparatus 1 is used. A number of search query candidates combining the search word “Kiruna” with related search words (for example, “travel”, “accommodation”, “reservation”) are displayed in the candidate display field C, and any candidate is selected with the mouse pointer P can do.

[3-3. (Use for web search)
Further, as another example of utilizing the determined type of usage, the order of lists of information items such as URLs in web search results (also called ranking order) can be considered. In this case, the list order control means 70 may determine the information items included in the web search result based on the noun (for example, individual URLs hit as the search result and the summary sentence or extracted sentence of the web page of the URL). The list order of the information items is controlled based on whether or not an expression relating to the same type as the type determined for the noun is included.

  Specifically, the type of the search term input when the user performs a web search is determined by the apparatus 1 and the result is reflected in the ranking order. For example, when the type of a noun that is a search query is determined to be a place name according to the present invention, when outputting a web search result based on the search query, a web page including a description of a pattern that matches the type of place name is displayed as a search result. It is possible to reflect the user's search intention by placing it at the upper level. In addition, about the web search server 3, since the structure regarding storage of search index data, a search process using the same, etc. is common, it abbreviate | omits.

[4. effect〕
As described above, based on the context regarding the noun in the search result of the search engine (for example, FIG. 3 and FIG. 4), the type of the noun can be determined even for a noun whose type is unknown from the given context. The type can be utilized for optimizing the information processing result based on the user's intention, for example, by reflecting the related search word in the related search (for example, FIG. 5) or controlling the list order of the web search result.

  Further, in the present embodiment, the nouns to be determined in the corresponding patterns for each type with respect to a predetermined number of summaries (for example, FIG. 4) such as snippets that are character strings in which a part including a search word is extracted for each URL. The most commonly used nouns that correspond to multiple types are collected by counting the quantities that have been used, etc. Therefore, it is possible to optimize information processing results that match the intentions of many users.

  In particular, in the present embodiment, by presenting a word or the like previously associated with a noun type as a related search word in a so-called related search based on the determined noun type (for example, FIG. 5), The selection of a suitable search query can be effectively supported.

  In addition, in the present embodiment, by controlling the list order of information items such as URLs included in the web search result based on the determined noun type, the web search result suitable for the user's intention is minimized. It becomes possible to see by scrolling and page transition, improving the usability of web search.

[5. Other embodiments]
In addition, said each embodiment is only an illustration, and this invention includes what is illustrated below and other embodiment other than that. For example, the present invention can be applied to Chinese, Korean, French and any other foreign language other than Japanese.

  In addition, each element such as means relating to the present invention may be realized by other information processing mechanisms such as a physical electronic circuit without being limited to the arithmetic control unit of the computer. Further, each configuration diagram, data diagram, flowchart diagram, and the like are merely examples, and the presence / absence of each element, the order of arrangement and processing execution, specific contents, and the like can be changed as appropriate. For example, the apparatus of the present invention may be realized by using a plurality of apparatuses such as servers, and a configuration in which each storage unit is realized by a separate and independent server apparatus or system is also common. Depending on the function, the configuration can be flexibly changed, for example, by calling an external platform or the like with an API (application program interface) or network computing (so-called cloud or the like).

1. Information processing device (this device)
2 automatic translation server 3 web search server 6 arithmetic control unit 7 storage device 8 communication means 10 noun acquisition means 15 search keyword input means 20 search result acquisition means 25 pattern storage means 30 type determination means 35 related search word storage means 40 determination result totalization Means 50 Result output means 60 Related presentation means 70 List order control means C Candidate display field F Input field N Communication network P Mouse pointer T Terminal

Claims (6)

A noun acquisition means for acquiring a determination target noun given as an object of language processing;
Search result acquisition means for acquiring a web search result based on the acquired noun;
A type determining means for determining a type of the noun based on a context regarding the noun in the acquired web search result;
An information processing apparatus comprising: A determination result aggregating unit for aggregating the quantity for each of the types determined by the type determining unit for a predetermined number of elements included in the acquired web search result;
A result output means for preferentially outputting the type with a large number of the aggregated quantities as a determination result;
The information processing apparatus according to claim 1, further comprising:   The related presentation means for presenting a predetermined word according to the type as a candidate for a word specified as a search query simultaneously with the noun based on the type determined for the noun. 3. The information processing apparatus according to 1 or 2. List order control means for controlling the list order of information items based on whether or not the information item included in the web search result based on the noun includes an expression related to the same type as the type determined for the noun. The information processing apparatus according to any one of claims 1 to 3, further comprising: A noun acquisition process in which a computer acquires a determination target noun given as a language processing target;
A search result acquisition process in which a computer acquires a web search result based on the acquired noun;
A type determining process in which a computer determines a type of the noun based on a context regarding the noun in the acquired web search result;
An information processing method comprising: A computer program for determining a type of noun by controlling a computer,
Let the computer get the nouns to be judged given as language processing targets,
Let the computer acquire web search results based on the acquired nouns,
An information processing program for causing a computer to determine a type of the noun based on a context regarding the noun in the acquired web search result.

Images