JP5953248B2 - Intent classification device, method and program, service selection support device, method and program - Google Patents

Description

  The present invention relates to an intent classification device, method, and program for classifying a number of services provided by a website, for example, in a form that is easy for the user to select, and a service selection support device for presenting services to the user using the classification result , Methods and programs.

  In recent years, with the evolution of mobile devices and the Web, services are diversifying and rapidly increasing. On the other hand, it has become difficult for users to find the service they want from many services, and the problem of increasing the number of users who feel that they cannot keep up with ICT (Information and Communication Technology) services has become serious. . In response to the current state of society, there is an increasing demand for technology that presents appropriate information that fulfills the vague desires of users.

Therefore, as a first step for service discovery support, a method has been proposed in which an action is first recommended to a user and then a service that realizes what is desired to be performed is recommended (for example, see Non-Patent Document 1). ).
In addition, a technique for supporting discovery of a service presumed to be preferred by a user based on a service user's usage log has been proposed. This is, for example, recommending a service used by a friend or recommending a service used by a person who has similar hobbies and preferences (see, for example, Non-Patent Document 2).

"Expansion of task-oriented mobile service navigation system and user evaluation", Yusuke Fukasawa et al., Information Processing Society of Japan Vol.50, No.1, pp.159-170, January 2009 “On Social Networks and Collaborative Recommendation”, I. Konstas, V. Stathopoulos and J. M Jose, In 32nd international ACM SIGIR conference on Research and development in information retrieval, pp.195-202, 2009, SIGIR ’09

  In order to realize the method described in Non-Patent Document 1, it is necessary to attach an action tag to a service. However, in Non-Patent Document 1, since the action tag is manually assigned, the efficiency is poor. In addition, in order to realize automatic assignment of action tags, it is necessary to classify noise from among action tags containing a large amount of noise, and further classify the importance of action tags to support service discovery. .

  Further, the technique described in Non-Patent Document 2 has the following problems. That is, recently, there are service vendors that collect the number of times an application is used by a large number of mobile terminal users as log information, and it is conceivable to use the log information collected by the vendor. However, since the log information is personal information for the user, it is generally difficult to acquire the log information. Even if it can be acquired, there remains a cold start problem that the recommendation accuracy cannot be maintained unless there is a certain amount of service search logs.

  The present invention has been made paying attention to the above circumstances, and its purpose is to effectively classify intents of a large number of services while reducing the influence of noise without using the personal information of the user. It is an object of the present invention to provide an intent classification device, method, and program that can be used, and a service selection support device, method, and program that can effectively present a service to a user according to the classification result.

To achieve the above object, according to a first aspect of the present invention, there is provided an intent classification apparatus for classifying intents included in a service based on the function importance, and the service extracted from the explanatory text representing the service. read and Inten bets representing the intention of the provider by natural language expression, the intent service related information including accessing information for accessing the device or data when executing the service from the serial憶部. Then, based on the read information, the feature amount based on the degree of association between all services including the intent and the access information included in these services is used as the first feature amount indicating the functional importance of the intent. First calculation means for calculating as an evaluation scale is provided. Then, based on the feature amount calculated by the first calculation means, the function importance level of the intent is classified, and information representing the classification result is stored.

The first aspect of the present invention is also characterized by comprising the following aspects.
The first aspect further includes second calculation means for calculating a feature quantity based on the appearance position of the extracted intent in the service commentary as a second evaluation measure representing the function importance of the intent. . The function importance level of the intent is classified based on the feature amounts calculated by the first and second calculation means.

According to a second aspect, there is provided a third calculation unit that calculates a feature amount based on the number of categories to which the service including the extracted intent belongs as a third evaluation measure representing the functional importance of the intent. Further prepare. Then, based on the feature quantities calculated by the first and third calculation means or the feature quantities calculated by the first, second, and third calculation means, the function importance of the intent is calculated. The level is classified.

According to a second aspect of the present invention, information related to an intent extracted from an explanatory text representing the service is read from the storage unit for the intent service related information, and the function importance of the intent is read from the classification result storage means. Reads information representing the classification result of the level . First, based on the information indicating the classification result of the read function importance level , the plurality of extracted intents are sequentially selected in a certain amount from those having a high function importance level , The list is displayed step by step on the user terminal. Next, when one of the intents displayed on the user terminal is selected and operated by the user, the service information corresponding to the selected intent is read from the storage unit for the intent service related information, and The list is displayed on the user terminal.

  According to the first aspect of the present invention, the function importance of an intent is calculated as a feature amount based on the degree of association between all services including the intent and access information included in these services. Then, intents are classified and presented to the user based on the calculated feature amount representing the function importance. For this reason, it is possible to provide an interface that narrows down the intent based on the function importance scale for the user who wants to do something vaguely. Also, by classifying intents based on their functional importance, it is possible to avoid the noise that was generated in automated processing using only the extraction method based on natural language processing. Generation of intent is suppressed. This makes it possible to classify intents with high accuracy. Furthermore, there is no need to collect personal information such as log data of the user, thereby preventing leakage of the personal information of the user and protecting the personal information.

According to the first aspect, the feature quantity based on the appearance position of the extracted intent in the service commentary is calculated as the second evaluation scale representing the functional importance of the intent. The function importance level of the intent is classified based on the feature amount previously calculated as the first evaluation scale and the feature amount calculated as the second evaluation scale. Therefore, it is possible to further accurately classify the level of function importance of the intent.

According to the second aspect, a feature quantity based on the number of categories to which the service including the intent belongs is further calculated as the third evaluation scale representing the functional importance of the intent. Then, the function of the intent based on the feature amounts calculated as the first and second evaluation scales or the feature amounts calculated as the first, second and third evaluation scales, respectively. The level of importance is classified. Therefore, it is possible to more accurately classify the level of function importance of intents.

According to the second aspect of the present invention, based on the information indicating the result of classifying the level of function importance, first, a plurality of intents are sequentially selected from a higher level of function importance by a certain amount. When the list is presented to the user and then one of the presented intent lists is selected and operated by the user, a list of service information corresponding to the selected intent is displayed to the user. Presented. Therefore, the user to be stepwise presented in descending order the list of intents of levels of functionality importance, because the service is presented in response to user's selection operation, the service discovery by the user It is possible to provide efficient support.

  That is, according to the present invention, an intent classification apparatus, method, and program that can effectively classify intents of a large number of services by reducing the influence of noise without using personal information of users, It is possible to provide a service selection support apparatus, method, and program capable of effectively presenting a service to a user according to the classification result.

The block diagram which shows the function structure of the system containing the intent classification | category apparatus and service selection assistance apparatus which concern on one Embodiment of this invention. The flowchart which shows the process sequence and process content by the function importance determination part of the intent classification device shown in FIG. The figure which shows the meaning of the variable used for the calculation of the feature-value showing the function importance by the intent classification device shown in FIG.

Embodiments according to the present invention will be described below with reference to the drawings.
[One Embodiment]
(Constitution)
FIG. 1 is a block diagram showing a functional configuration of a system including an intent classification device and a service selection support device according to an embodiment of the present invention. This system includes a storage device 1, an intent information processing device 2, a server 3, and a client terminal 4.

  The storage device 1 is provided, for example, in a database server, and as a storage area necessary for realizing one embodiment of the present invention, a synonym dictionary accumulation unit 11, a function importance correct answer data accumulation unit 12, an intent service A related information accumulation unit 13 and an intent importance accumulation unit 14 are provided.

  The synonym dictionary accumulation unit 11 stores synonym dictionary data used when extracting an intent from service information. The function importance correct answer data accumulating unit 12 stores function importance correct answer data used for determining the intent importance. The intent service related information accumulating unit 13 is used for storing a manifest category related to the acquired service information and information indicating one and the number of extracted intents. The intent importance accumulation unit 14 is used to store information indicating the function importance of the determined intent.

  The intent information processing apparatus 2 includes a computer. As processing functions necessary for realizing one embodiment of the present invention, a service information acquisition processing unit 21, an intent extraction processing unit 22, an intent important A degree determination processing unit 23 is provided. Each of these processing units 21 to 23 is realized by causing a computer to execute a program stored in a program memory (not shown).

  The service information acquisition processing unit 21 acquires service information from a site on the Web. The service information includes service descriptions, service category information, and service access information. The category information is a label for a service provided by the application distribution portal, and is usually 20-30. The access information means access information to the device function of the service or access information to specific data in the terminal. Such access information is included in, for example, a manifest file which is a non-natural language, and is called ManifestXML in an application compatible with Android (registered trademark). In this XML, access information such as “PERMISSION: CAMERA” “application uses camera function” is described. The service information acquisition processing unit 21 stores the manifest file and category information of the acquired service information in the intent service related information accumulation unit 13.

The intent extraction processing unit 22 has the following processing functions.
(i) Processing for extracting intent candidates by performing morphological analysis on the service description provided from the service information acquisition processing unit 21.
Here, the intent is defined as “abstract expression in the natural language of the user intention when using the service”. For example, “photo editing”, “driving assistance”, “recipe search”, and the like. In other words, the intent is a concept that abstracts the user's behavior and purpose of “what can be done” and at the same time a concept that expresses the function provided by the service. In other words, the intent is a concept in which the intentions of the service user and the service provider are integrated, and serves as a bridge between the two.

  (ii) With respect to a plurality of extracted intent candidates, the synonym dictionary data stored in the synonym dictionary accumulating unit 11 is referred to, and intents having the same meaning are processed so as to have the same expression. And the process which memorize | stores the intent obtained in this way in the intent service related information accumulation | storage part 13 with the information showing the appearance position and appearance frequency.

The intent importance level determination processing unit 23 obtains the appearance number and appearance position of each intent and the access information and category information of each service from the intent service related information storage unit 13, and based on these information, Calculate an evaluation scale that represents the functional importance of the tent.
Here, the function importance is a level indicating the importance given to each intent. For example, a faceted search that narrows down search candidates by selecting a function is effective in finding what the user wants from many services. Facet search for service discovery support is a search method in which (1) one main function intent is selected, and (2) services are filtered by adding and selecting a plurality of sub function intents. By quantitatively calculating the function importance, it is possible to determine whether the function is a main function or a sub function. Moreover, it is possible to determine that the function importance is low as noise.

There are the following three evaluation scales representing functional importance.
(i) The first evaluation scale representing the function importance is a feature quantity considering the dependence of the intent on the access information. If common access information exists in all services including the intent Ij, the intent Ij is strongly related to the device function and the information in the device. The intent that is strongly related to the access information is judged to have high function importance.

  (ii) The second evaluation scale representing the functional importance is a feature quantity considering the appearance position and the number of appearances of the intent. The appearance position of the intent is a distance based on the number of appearances from the head of the intent in the service commentary. In order to consider the feature that the intent expressing the function of the service has a strong tendency to appear in the first half of the service commentary, the appearance frequency of the service is evaluated by weighting the appearance position.

  (iii) The third evaluation scale representing the functional importance is a feature quantity considering the co-occurrence of the intent in the category, that is, a feature quantity based on the number of categories to which the service having the intent belongs. An intent having a high function importance is considered to have a large number of services having the same intent Ij in the category C to which a certain intent Ij belongs.

  As described above, the intent importance level determination processing unit 23 creates a classifier used for classifying the three feature quantities calculated as the evaluation scale representing the functional importance level of the intent. The classifier reads correct data from the function importance correct data accumulating unit 12 and creates it by machine learning. Then, by inputting the calculated three feature quantities to the classifier, the level of function importance is classified, and information indicating the function importance level of the classified intent is stored in the intent importance accumulation unit 14. To store.

  The intent information processing unit 2 of the processing units 21 to 23 of the intent information processing device 2 described above and the intent importance accumulation unit 14 of the storage device 1 constitute an intent classification device. The

  The server 3 is composed of a Web server computer operated by a service selection support provider. As a control function necessary for realizing an embodiment of the present invention, a presentation intent processing unit 31, a presentation service processing unit 32, and a presentation are provided. A screen processing unit 33 is provided. Each of these processing units 31 to 33 is realized by causing the computer to execute a program stored in the program memory.

The presentation intent processing unit 31 reads the information indicating the function importance level of the classified intent from the intent importance accumulation unit 14, and based on this information, the intent importance level to be presented is high. Select a tent.
The presentation service processing unit 32 reads information representing a service associated with the selected intent from the intent service related information accumulation unit 13.
The presentation screen processing unit 33 is based on the intent information selected by the presentation intent processing unit 31 and the service information read from the intent service related information accumulation unit 13 by the presentation service processing unit 32. The presentation screen data for presenting to the user is generated. The presentation screen data includes a presentation intent list and a presentation service list. Then, in response to access from the client terminal 4, the presentation screen data is transmitted to the access source client terminal 4.

  The intent importance determination processing unit 23 of the intent information processing device 2 described above, the intent importance level accumulation unit 14 of the storage device 1, and the processing units 31 to 33 of the server 3 constitute a service selection support device. The

  The client terminal 4 is composed of, for example, a mobile phone, a smart phone, or a tablet terminal used by a user. As a control function necessary for realizing an embodiment of the present invention, in addition to a Web browser, a screen display unit 41 and , An operation input interface (operation input I / O) 42 is provided.

  The screen display unit 41 is configured by a display device such as a liquid crystal or an organic EL, for example, and displays a list of presentation intents and a list of presentation services received from the server 3 under the control of the Web browser. The operation input I / O 42 is used by the user to select and operate a desired intent and service from the displayed presentation intent and presentation service list. The Web browser transmits the intent and service selection information to the server 3.

(Operation)
Next, the procedure and processing contents of intent classification processing and service selection support processing by the system configured as described above will be described.
(1) Acquisition of Service Information In the intent information processing apparatus 2, first, service information is acquired from a Web site under the control of the service information acquisition processing unit 21. As described above, the service information includes the service description, the service category information, and the service access information. Of the acquired service information, information representing a service commentary is sent to the intent extraction processing unit 22, and access information and category information included in the manifest file are sent to the intent service related information accumulating unit 13. Remembered.

(2) Intent Extraction In the intent information processing device 2, under the control of the intent extraction processing unit 22, the service explanation sentence provided from the service information acquisition processing unit 21 is subjected to morphological analysis. Thus, intent candidates are extracted. Then, with reference to the synonym dictionary data stored in the synonym dictionary accumulating unit 11, processing is performed to make the intent candidates having the same meaning among the extracted intent candidates the same expression. The intent for which this processing has been completed is stored in the intent service related information accumulating unit 13 together with information indicating the appearance position and the number of appearances.

(3) Determination of Intent Function Importance of Intent In the intent information processing apparatus 2, the function importance of the intent extracted by the intent extraction processing unit 22 is subsequently controlled under the control of the intent extraction processing unit 22. Processing for determining the degree is performed. FIG. 2 is a flowchart showing the processing procedure and processing contents. In addition, Table 1 shows the meanings of the variables used for the function importance determination process.

  First, in step S11, from the intent service related information accumulating unit 13, information indicating the intent, its appearance position and the number of appearances, and access information and category information included in the service manifest file are read. Then, based on these pieces of information, three evaluation measures representing the functional importance of the intent are calculated as follows.

  That is, first, in step S12, the intent information and the manifest file are read from the intent service related information accumulating unit 13, and based on the read intent and the access information included in the manifest file, the function importance level is changed. As the evaluation scale of 1, the feature quantity 1 considering the dependence on the access information of the intent is calculated. If common access information exists in all services including the intent Ij, the intent Ij is strongly related to the device function and the information in the device. The intent that is strongly related to the access information is judged to have high function importance.

For example, the dependence of the intent Ij on the access information is calculated as follows. First, a set of services S {j} including an intent Ij is defined as follows.

Next, a set of services S {j, k} including the intent Ij and including the access information mk can be expressed as follows.

Finally, the dependency between the intent Ij and the access information is quantified by the maximum value of the probability of the access information that appears in the service in which the intent appears.

  Next, in step S13, a feature amount 2 that takes into account the appearance position and the number of appearances of the intent is calculated as a second evaluation scale representing the functional importance. Here, the appearance position is represented by a distance based on the number of words from the head of the service commentary as described above. Considering the feature that the intent expressing the function of the service has a strong tendency to appear in the first half of the service commentary, the appearance frequency of the service is evaluated by weighting the appearance position.

により算出する。ただし、ｐi,k,α は、図３に示すとおり、インテントj の出現位置を表している。式(6) は、式(7) のＳＦi の値を全てのサービスに対して合計した値を表している。式(7)のＳＦi とは、サービスi においてインテントＩj が出現した場合、サービスＳi 内で出現する他のインテントに対して出現位置をもとにした相対的な重み付けを表している。 For example, if the feature quantity that evaluates the function importance of a certain intent Ij is Sj2, the feature quantity Sj2 has a smaller value of its relative appearance position every time the intent Ij appears in all services. It is determined that the function importance is higher. That is,

Calculated by However, pi, k, α represents the appearance position of the intent j as shown in FIG. Equation (6) represents the sum of the values of SFi in equation (7) for all services. In the equation (7), SFi represents a relative weighting based on the appearance position for other intents appearing in the service Si when the intent Ij appears in the service i.

すなわち、第３の評価尺度Ｓj3はスカラ値として計算される。但し、|{Ｓ}| は集合Ｓに含まれる元の数を示す。 Finally, in step S14, the feature quantity 3 considering the co-occurrence of intents in the category is calculated as the third evaluation measure of the function importance. An intent having a high function importance is considered to have a large number of services having the same intent Ij in the category C to which a certain intent Ij belongs. Based on this viewpoint, the third evaluation scale Sj3 is calculated as follows, for example.

That is, the third evaluation scale Sj3 is calculated as a scalar value. However, | {S} | indicates the number of elements included in the set S.

  As described above, when the three feature amounts representing the function importance of the intent are calculated, it is next determined in step S15 whether or not there is a classifier for classifying the function importance. If the classification is not created as a result of this determination, the correct answer data is read from the function importance correct answer data storage unit 112 in step S18. The correct answer data is obtained by, for example, performing a five-level evaluation on an intent appearing in a service description by manual subjective evaluation and taking an average value for each intent. In step S19, a classifier is created by machine learning based on the read correct answer data. The learning algorithm performs ternary classification using, for example, a support vector machine. The priority is determined using this classifier. For example, when a classifier of ternary classification is configured, the priority level is determined by determining whether the intent function importance is high, medium, or low.

  In step S16, the calculated three feature quantities are input to the created classifier. As a result, the functional importance level is classified by the classifier. Finally, in step S <b> 17, information indicating the function importance level of the classified intent is stored in the intent importance accumulation unit 14.

  As a result of the determination of the presence or absence of the classifier in step S15, if the classifier has been created, the process proceeds to step S16 as it is, where the level classification of function importance is performed by the created classifier. In S <b> 17, information representing the function importance level classification result is stored in the intent importance level accumulation unit 14.

(4) Creation of Intent Presentation Screen and Service Presentation Screen The server 3 collects and stores intents of the service from the service commentary, and displays the determination result of the intent importance by the intent importance determination processing unit 23. Utilizing this, the accumulated intents are presented in a step-by-step manner, from the one with the highest function importance to the one with the lower function importance. Then, the service corresponding to the selected intent is presented to the user. The specific processing contents are shown below.

  First, the presentation intent processing unit 31 selects a certain number of intents in descending order of function importance based on the function importance determination result of the intent stored in the intent importance accumulation unit 13. Then, the presentation screen processing unit 33 creates intent presentation screen data for presenting the selected list of intents.

  In this state, it is assumed that the user inputs a recommended service presentation request using, for example, a Web browser at the client terminal 4. Then, the service presentation request is sent from the client terminal 4 to the server 3, and the created intent presentation screen data is first sent from the server 3 to the client terminal 4 in response to this request. As a result, the list of intents is displayed on the screen display unit 41 of the client terminal 4 as an intent selection screen. 3 shows an example of the intent selection screen displayed at this time.

  In this state, it is assumed that the user selects one of the displayed intent selection screens, for example, “photo editing”. Then, the selection information is sent from the client terminal 4 to the server 3. In contrast, in the server 3, under the control of the presentation service processing unit 32, service information associated with the intent “photo editing” selected by the user is read from the intent service related information accumulation unit 13. . Then, a service presentation screen for presenting the read list of services is created and transmitted to the client terminal 4. As a result, a list of services related to the “photo editing” is displayed on the screen display unit 41 of the client terminal 4 as a service selection screen. 3 shows an example of the service selection screen displayed at this time.

  In this state, if a user has a desired service on the displayed service selection screen and selects this service, for example, “Service B”, the selection information is sent from the client terminal 4 to the server 3. On the other hand, the server 3 downloads the screen data for using the selected “Service B” to the client terminal 4.

  On the other hand, it is assumed that the desired service is not found in the displayed service selection screen and the user selects “Add sub function”. Then, the selection information is sent from the client terminal 4 to the server 3. In contrast, in the server 3, under the control of the presentation intent processing unit 31, intents that can be added to the intent previously selected by the user are selected in descending order of function importance. Then, an intent selection screen for presenting the selected list of intents is created and transmitted to the client terminal 4. As a result, the sent intent addition list is displayed as a sub-function addition screen on the screen display unit 41 of the client terminal 4. An example 53 is shown in FIG.

  When the user selects a desired intent from the displayed intent addition list, for example, “add comic effect”, the selection information is sent from the client terminal 4 to the server 3. In the server 3, under the control of the presentation service processing unit 32, service information associated with the intent “Add comic effect” selected by the user is read from the intent service related information accumulation unit 13. Then, a service presentation screen for presenting the read service “Service A” is created and transmitted to the client terminal 4. As a result, the service “Service A” related to the “add comic effect” is displayed as a service selection screen on the screen display unit 41 of the client terminal 4. 3 shows an example of the service selection screen displayed at this time.

When the user selects the displayed service “Service A”, the server 3 downloads screen data for using the selected “Service A” to the client terminal 4.
Thus, the user can select the main function desired by the user and the service that realizes the sub function.

(effect)
As described above in detail, in one embodiment, the intent information processing apparatus 2 extracts intents from a plurality of services collected from a website, and calculates three feature amounts representing the functional importance from these intents. Then, the intents are classified based on the calculated three function importance levels, and the results are stored in the intent importance accumulation unit 14. Then, the server 3 generates an intent presentation screen based on the information indicating the function importance level of the intent stored in the intent importance level accumulation unit 14 and transmits it to the client terminal 4 to display it. When a desired intent is selected from the displayed intents, a list of services related to the intent is created and transmitted to the client terminal 4 for display.

  Therefore, the function importance level of the intent is calculated from three viewpoints based on the service information collected from the website, and the intent is classified based on the feature amount representing the calculated three function importance levels. Because it is presented to the user, it is possible to provide an interface that narrows down the intent based on the function importance measure for the user who wants to do something vaguely. In addition, it is possible to suppress the occurrence of noise that has occurred in automation using only the extraction method based on natural language processing, thereby suppressing the generation of unnecessary intents. This makes it possible to classify intents with high accuracy. Furthermore, it is not necessary to collect personal information such as log data of other users, thereby preventing leakage of personal information of users and protecting the personal information.

  Furthermore, when presenting the intent list to the user, the main function selection screen is first presented based on the function importance of the intent, and when the intent in the main function selection screen is selected, the intent is displayed. Generate and present a list of services related to. Then, a button for requesting the addition of a subfunction is provided in this service list, and when the user selects this subfunction addition button, a list of intents related to the first selected intent is displayed. It is generated according to the degree of function importance and presented as a sub-function addition screen. For this reason, it is possible to realize facet search according to a user's request, thereby making it possible to discover services more efficiently.

  That is, it is possible to automatically extract a large number of service intents with high accuracy while reducing the influence of noise without using personal information of the user. Furthermore, by using the classification result, it is possible to provide support so that a service desired by the user can be found efficiently.

[Other Embodiments]
In the one embodiment, the storage device 1, the intent information processing device 2, and the server 3 are configured as separate devices. However, the server 3 is provided with the functions of the storage device 1 and the intent information processing device 2. Also good. That is, the functions of the intent classification device and the service selection support device may be provided in one device.

  In the embodiment, the intent importance determination processing unit 23 calculates three evaluation scales representing the functional importance of the intent, and classifies the intent based on the calculated three evaluation scales. I made it. However, the present invention is not limited to this, and the intent may be classified using any one of the first, second, and third evaluation scales, and the first, second, and second The intent may be classified using at least two of the three evaluation scales.

In addition, the configuration of the intent classification device and service selection support device, the procedure and processing contents of the intent classification processing and service selection support processing, the types of services and intents, and the like are variously modified without departing from the scope of the present invention. Can be implemented.
In short, the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

  DESCRIPTION OF SYMBOLS 1 ... Web, 2 ... Intent information processing part, 3 ... Server, 4 ... Client terminal, 11 ... Synonym dictionary accumulation part, 12 ... Function importance correct answer data accumulation part, 13 ... Intent service related information accumulation part, 14 ... Intent importance level accumulating unit, 21 ... Service information acquisition processing unit, 22 ... Intent extraction processing unit, 23 ... Intent importance level determining processing unit, 31 ... Presentation intent processing unit, 32 ... Presentation service processing unit, 33 ... presentation screen processing unit, 41 ... screen display unit, 42 ... operation input interface (operation input I / O).

Claims (7)

It extracted from commentary representing the service, intent service related information including a Inten bets expressed by natural language expressions intentions of the service provider, the access information for accessing the device or data when executing the service and means for reading from the serial憶部,
A feature amount based on the degree of association between all the services including the extracted intent and access information included in these services is calculated as a first evaluation scale representing the functional importance of the intent. Means for calculating
And classifying means based on Risan out feature quantity by the first calculating means, for classifying the level of functional importance of the intent,
An intent classification apparatus comprising: a classification result storage unit that stores information representing a classification result of the level of functional importance of the intent. A second calculating means for calculating a feature quantity based on the appearance position in the service description of the extracted intent as a second evaluation scale representing the function importance of the intent;
2. The intent classification apparatus according to claim 1, wherein the classification unit classifies the level of function importance of the intent based on the feature amounts calculated by the first and second calculation units. . A third calculation means for calculating a feature quantity based on the number of categories to which the service including the extracted intent belongs as a third evaluation measure representing the function importance of the intent;
The classification unit is configured to determine the intent of the intent based on the feature amounts calculated by the first and third calculation units or the feature amounts calculated by the first, second, and third calculation units. 3. The intent classification apparatus according to claim 1 , wherein the level of function importance is classified. A storage unit for storing intent service related information;
Classification result storage means for storing information indicating the classification result of the level of functional importance of the intent;
From the storage unit of the intent service-related information, which is extracted from the commentary representing the service, a means for reading the Inten bets expressed by natural language expressions intentions of the service provider,
Means for reading from the classification result storage means information representing the classification result of the functional importance level of the intent;
Based on the information indicating the classification result of the read level of function importance, the extracted plurality of intents are sequentially selected in a certain amount from those having the highest level of function importance and the list is used. Means for displaying in stages on the user terminal,
When a user selects and operates one of the intents displayed on the user terminal, the operation information is received from the user terminal, and the service corresponding to the intent represented by the received operation information A service selection support apparatus comprising: means for reading information from the storage unit for intent service related information and displaying the list on the user terminal. An intent classification method executed by an intent classification device including a computer,
The intent classifier was extracted from commentary representing the service, Inten Doo and access information for accessing the device or data when executing the service intended service providers represented by natural language expression A process of reading from the storage unit storing intent service related information including
The intent classifying device represents a feature amount based on a degree of association between all the services including the extracted intent and access information included in these services. A process of calculating as an evaluation scale;
The intent classification device classifying the level of functional importance of the intent based on the feature amount calculated as the first evaluation scale;
The intent classification apparatus includes a step of storing information representing a classification result of the function importance level of the intent in a storage unit. A service selection support method executed by a service selection support device including a computer,
Said service selection support device, comprising the steps of reading from a storage unit for storing the intent-related service information, extracted from the commentary representing the service, a Inten bets expressed by natural language expressions intentions of the service provider,
Said service selection support apparatus, from the storage unit for storing the classification result of the level of functional importance of intent, comprising the steps of: reading information indicating the classification result of the level,
The service selection support device sequentially extracts a plurality of extracted intents in a certain amount from the one with the highest function importance level based on the information indicating the classification result of the read function importance level. Selecting and displaying the list step by step on the user terminal,
When the service selection support device selects and operates one of the intents displayed on the user terminal , the service selection support device receives the operation information from the user terminal and is represented by the received operation information. A service selection support method, comprising: reading service information corresponding to an intent from the first storage unit and displaying the list on the user terminal. The program which makes the computer with which the said intent classification apparatus or service selection support apparatus performs the process performed by each means with which the intent classification apparatus or service selection support apparatus in any one of Claims 1 thru | or 4 is provided.

Images