JPH10334119A - Network information retrieval system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To actualize the information retrieval system which enables a user to gather interesting information resources without awareness. SOLUTION: A data storage part 5 stores information resources obtained by retrieving information resources in a computer network 100 according to a profile sentence that the user specifies. A sound reproduction part 7a reproduces the information resources stored in the data storage part 5 in voice and receives the specification of the end of the reproduction from the user. A history storage part 8 holds a history of the sound reproduction by the sound reproduction part 7a. A data processing part 9 decides that the user is interested in the information resources when the sound reproduction time is longer than a predetermined threshold value and feeds it back to the profile sentence. The new fed-back profile sentence is sent to a profile sentence process part 2 and next retrieval will be carried out based on this profile sentence.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information resource which frequently changes on a computer network.
The present invention relates to a network information search system in which a collection agent collects information resources of interest to a user and outputs the collected information resources.

[0002]

2. Description of the Related Art Conventionally, there are the following documents relating to information retrieval in a network. [1] Ronald J. Vetter, Chris Spell, Charles Ward, M
osaic and the World-Wide Web, IEEE Computer, Oct
ober, 1994, pp49-57 [2] Roy T. Fielding, Maintaining distributed hyp
ertext infrostructures: Welcome to MOMspider's W
eb, Computer Networks and ISDN Systems, 27 (4), pp1
93-204, 1994 [3] Gerard Salton, Chris Buckley, Improving Retr
ieval Performance byRelevance Feedback, Journal o
f The American Society for Information Science, 41
(4), pp288-297,1990

[0003] WWW, an Internet information server
(World-Wide Web) uses hypertext as a means for connecting various resources on the Internet (shown in the above-mentioned document [1]). The language for describing this hypertext is HTML (HyperText Markup Lan).
guage), and has a function of expressing information of a link to another information resource and a format of a document. In addition, HTML
Is SGML (Standard Generalized Markup Languag)
It is a language based on e).

A method for automatically navigating information resources provided on the WWW is described in, for example, the above-mentioned document [2].
As shown by, it is called WWW. This WWW is used for the following purposes. ● Create an information retrieval system on the Internet ● Verify and maintain hypertext links ● Discover and extract information resources

[0005] Next, a vector space model and related feedback will be described as a search technique for information search on a network.

[0006] As shown in the above document [3], the vector space model is a search technique that captures words in text as vectors. Assign importance when setting words in text. The question sentence uses a natural language sentence, and is converted into a vector like a text word. Search results are
It is a list in order of similarity to a question sentence of a registered text.

A vector of a certain text is expressed by the following equation. _{Here, dw 1, dw 2, ...} , dw t is the importance of the word of the text. DW = (dw ₁ , dw ₂ ,..., Dw _t ) Similarly, the vector of the question sentence is expressed by the following equation. here,
q ₁ , q ₂ ,..., q _t are the degrees of importance of the words of the question sentence. Q = (q ₁ , q ₂ ,..., Q _t )

Further, the term importance (weight vector word T _k of the document D _i) is given in Figure 2. FIG. 2 is an explanatory diagram of each operation expression in the vector space model. Equation (1) in the figure is an arithmetic expression for the importance of a word, and in equation (1),
W _ik is a vector of word T _k of the document D _i weight, tf _ik is the number of occurrences of the word T _k of the document D _i, the total number of documents N is collected,
n _k is in the document that has been collected, the number of documents that contain the word T _k.

The similarity is represented by equation (2) in the figure. Here, there is a threshold such as θ in equation (2). The search results are displayed in descending order of similarity among those satisfying Expression (3).

A method in which a user evaluates a retrieved document and feeds back a vector of the document indicating that the user is relevant to the retrieval formula to refine the retrieval formula is described in, for example, the above document [3]. Related feedback.
Although various methods have been refined for the related feedback, it is generally confirmed by experiments that the one shown in Expression (4) is efficient and accurate. Equation (4)
In the expression, all_relevant is a set of vectors of words of all the documents of interest to the user in the retrieved documents. Also, one_nonrelevant is a vector of words having the highest importance among documents that the user is not interested in among the retrieved documents.

[0011]

However, as a problem of the conventional information retrieval processing as described above, there are the following problems. That is, for information resources that are frequently changed among the information resources on the network, the user must access and navigate to a location on the network every time the information is changed.
Further, even when information resources are automatically collected like a WWW robot, the system side needs to organize and present it to the user because the association between the collected information resources is not clear. Was.

[0012] In view of the above, there has been a demand for an information retrieval system capable of collecting information resources of interest to a user without being conscious of the user.

[0013]

The present invention employs the following structure to solve the above-mentioned problems. <Structure of Claim 1> A data storage unit for storing information resources obtained by searching information resources on a network based on a profile sentence specified by a user, and reproducing the information resources stored in the data storage unit by voice. At the same time, an audio playback unit for receiving an end designation of the audio playback from the user, a history storage unit for keeping a history of the audio playback time of the information resources in the audio playback unit, and an audio playback time predetermined based on the history of the history storage unit. If the information resource is longer than the given threshold, the information resource is considered to be of interest to the user, and is fed back to the profile statement to create a new profile statement. .

<Explanation of Claim 1> The profile statement is
This is a search sentence for searching for information resources on a network. The search sentence expresses a user's interest in search for a certain length of time. The data storage unit stores information resources obtained by searching for information resources on the network based on such a profile statement. The audio reproduction unit reproduces the information resources stored in the data storage unit as audio,
When the user designates the end of reproduction, the reproduction time of the information resource is output to the history storage unit. The history storage unit,
The history of the audio playback time in the audio playback unit is obtained, and this history is output to the data processing unit. Based on the history in the history storage unit, the data processing unit determines that the information resource is of interest to the user if the audio reproduction time is longer than a predetermined threshold value, for example, 以上 or more. This is fed back to the profile sentence to create a new profile sentence. This threshold is 1
It can be set to any value other than / 2.

Therefore, the next search for information resources on the network is performed based on the new profile statement. As described above, according to the first aspect of the present invention, it is possible to collect information resources that match the interests of the user without inputting whether or not the user is interested in the collected information resources.

<Structure of Claim 2> According to claim 1, a data organizing unit for classifying the information resources stored in the data storing unit into a preset category, and a data organizing unit for each category classified by the data organizing unit. This is a network information search system including a sound reproducing unit that reproduces information resources of the organizing unit.

<Explanation of Claim 2> In the invention of claim 2, in addition to the invention of claim 1, information resources obtained by searching on a network are classified by category. Further, the audio reproduction unit reproduces the information resource by audio for each of the classified categories. With such a configuration, the user can reproduce and listen to the information resources for each category, so that the related information resources can be easily obtained.

<Structure of Claim 3> In claim 2, when a new category is specified by the user as a result of reproducing the information resource by the audio reproducing unit, the classification of the subsequent information resource is specified. A network information search system comprising a data organizing unit that includes a category.

<Explanation of Claim 3> According to the invention of claim 3, in the invention of claim 2, when the user adds a new category, the data organizing unit classifies the information resources including this category. It was made. Thereby, the information resources obtained by searching on the network can be further classified into the category intended by the user.

<Structure of Claim 4> A network information search system according to any one of claims 1 to 3, wherein the information resource reproduced by the audio reproducing unit is text data.

<Explanation of Claim 4> Since the information resources are texts, the contents of the information resources can be accurately expressed by voice, and as a result, the information resources can be grasped accurately for the user.

[0022]

Embodiments of the present invention will be described below in detail with reference to the drawings. << Specific Example 1 >><Configuration> FIG. 1 is a configuration diagram showing a specific example 1 of the network information search system of the present invention. The system shown in the figure includes a data input unit 1, a profile sentence processing unit 2, a filtering management unit 3, a collection agent 4, a data storage unit 5, a voice conversion unit 6, a voice reproduction unit 7, a history storage unit 8, a data processing unit 9, It comprises a data display unit 10.

The data input unit 1 includes a keyboard, a mouse, and the like, and is an input unit for a user to input a command for collecting and filtering information resources on a network. In addition, when a command is input from the user to the data input unit 1, the command is output to the profile sentence processing unit 2.

The profile sentence processing unit 2 is a functional unit for converting a profile sentence input by a user or a profile sentence newly created by the data processing unit 9 into base words. Here, the profile sentence has substantially the same meaning as the search sentence in the information search, but refers to a search sentence expressing the user's interest in the search for a certain length of time. The base words refer to the following words. That is, in the conversion process from the profile sentence to the base word in the profile sentence processing unit 2, in the case of English information resources, unnecessary words are deleted and then stemming is performed. Stemming is an operation to make the stem of the inflected form of English into a basic form, and the word resulting from performing the stemming is the base word of that word. For example, teis →
The words that have been stemmed, such as tie, hopping → hope, are the base words. A word that does not make sense even in a base word is called an unnecessary word, such as “the”, “a”, and “to”. The profile sentence can be input in a natural language. For example, a search sentence such as "Japan and United States" can be described.

The filtering management unit 3 is a functional unit that determines a collection strategy from the profile sentence of the profile sentence processing unit 2 and instructs the collection agent 4. In addition, the collection agent 4 operates in accordance with the instruction of the filtering
This is a functional unit for searching the above information resources. In general, on a distributed network, information resources include image data, moving image data, compressed files, and the like. In this specific example, the information resources are limited to text data. That is, in this specific example, the information resource indicates text data.

The data storage unit 5 is a storage unit for temporarily storing information resources searched by the collection agent 4 and filtered by the filtering management unit 3. The voice conversion unit 6 is a functional unit that converts the information resources stored in the data storage unit 5 into voice (converts text data into voice data) and sends it to the voice reproduction unit 7. The sound reproducing unit 7 is a functional unit including a speaker or the like that stores the sound converted by the sound converting unit 6 and reproduces the sound when a user requests for reproduction. In addition, the audio reproduction unit 7
When the history information is reflected in the profile, a “playback end button” that is pressed by the user is provided.

The history storage unit 8 is a function unit that records how long the user has listened to a certain information resource based on the playback state of the audio playback unit 7 and the user pressing the playback end button. If the audio reproduction time is longer than a predetermined threshold based on the history in the history storage unit 8, the data processing unit 9 determines that the information resource is of interest to the user and feeds it back to the profile sentence. Is a function unit that creates a new profile sentence and gives it to the profile sentence processing unit 2. The data display unit 10 is a functional unit that includes a display and presents the filtered information resources to the user through the display.

The portion composed of the voice conversion unit 6, the voice reproduction unit 7, the history storage unit 8, and the data display unit 10 is called a portable terminal unit 20. When the information resources are collected and filtered, the portable terminal unit 20 can be used separately from the network information resources including the filtering management unit 3 and the collection agent 4. This allows
By carrying the mobile terminal unit 20, the user can obtain the filtered information resources anywhere.

<Operation> Next, the operation of the network information search system having the above configuration will be described. FIG. 3 is an operation flowchart of the first embodiment.

[Step S1] When a start command is input from the user, the network information search system is started. The data input unit 1 waits for user data input.

[Step S2] The user inputs a profile statement, a collection start point on the computer network 100,
Enter the collection range and similarity threshold.
Here, it is assumed that the user has input the profile "Japan and United States". Also, as a starting point for collection, http://www.abc.com/english/enews/enews.html http://www.xyz.co.jp/ http://www.lmn.co.jp/index- Suppose that the start points of three collections, e.html, are input, and the depth of collection is 3. Here, http indicates a communication protocol on the computer network 100, www.abc.com indicates a certain machine on the computer network 100, and /english/enews/enews.html indicates a hierarchy of a certain directory on the machine. . The collection depth indicates how many hypertext links have been followed from the collection start point.

[Step S3] Profile sentence processing section 2
Decomposes the profile sentence input by the user into base words and calculates the weight of the base words. Here, unnecessary words are deleted, and the processing of stemming is performed. “Japan and United Sta
"japan", "unite", "stat" as base words of "tes"
"and" where "and" has been removed as an unnecessary word, and "United" has been stemmed to "un"
ite ", and" States "was stemmed to" state ".

For example, it is assumed that the first profile is as shown below. (State 0.577, unite 0.577, japan 0.577) Note that the above numbers indicate the importance of the base word.

[Step S4] Filtering management section 3
Instructs the collection agent 4 to search for a certain range from the collection start point specified by the user.

[Step S5] The collection agent 4
The collection is performed from the collection start point on the computer network 100 specified by the user. At this time, the importance of the base word is calculated for each collected information resource. This step S5
Also, in the base word, the unnecessary word is deleted, and then the processing of the stemming is performed. At this time, the importance of the base word of the information resource is obtained by using equation (1) in FIG.

[Step S6] Filtering management section 3
Calculates the similarity using the base word vector of the profile sentence and the base word vector of the information resource. Equation (2) in FIG. 2 is used to calculate the similarity. Those whose similarity is larger than the threshold value, that is, those that satisfy Expression (3), are stored in the data storage unit 5 as filtered information resources.

Next, presentation of the collected information resources by voice and presentation on a computer screen will be described separately. First, a method for presenting a voice will be described.

FIG. 4 is a flowchart of presentation by voice. [Step S7] The voice conversion unit 6 converts the collected information resources into voice and stores it in the voice reproduction unit 7.

[Step S8] The user sets the audio playback unit 7
Play the audio of and listen to the collected information resources.

[Step S9] The history storage unit 8 keeps a history of how much the user has heard each information resource.
Here, the history is information such that the user has heard only one half of the information resource, or the user has heard all the information resources. The parameters of this history can be freely determined by the user.

[Step S10] The history storage unit 8 determines whether the audio reproducing unit 7 has reproduced all the collected information resources. If the reproduction of the information resources has not ended, the process returns to step S8, and the audio reproduction unit 7 reproduces the remaining information resources. If the reproduction has ended, the process proceeds to step S11.

In the presentation on the screen on the computer (refer to the flowchart in FIG. 3), the collected information resources are displayed to the user, and the user inputs information as to whether the user is interested in the information resources. Done by

[Step S11] The data processing section 9 acquires feedback information from the history storage section 8. In the case of audio presentation, if the user hears the information resource for more than 1/2, it is assumed that the user is interested in the information resource. In the display on the computer screen, the user explicitly inputs whether or not the user is interested in the information resource. At this time, the feedback information is calculated by equation (4) in FIG.

The data processing section 9 reflects the feedback information thus obtained on the profile sentence processing section 2. For example, the profile obtained in the above step S3 is as follows. (State 0.200, unite 0.200, japan 0.100, base 0.12
3, suspect 0.104, hand 0.079, censorship 0.078, ot
a 0.072, agreement 0.072, okinawa 0.067, maruyama
0.066) As described above, when the feedback information is reflected in the profile sentence, the profile usually has an increased form.

[Step S12] If the user wants to end the filtering, the system inputs the fact through the data input unit 1, and the system performs the end processing. Otherwise, the process proceeds to the next step S13.

[Step S13] If the collection time has elapsed, the information resources that have been changed are collected within the range specified by the user in step S2. Here, the collection time is a time at which the user inputs at what time interval the information resources on the computer network 100 are to be collected.

<Effects> As described above, according to the specific example 1, the result of whether or not the user is interested in the information resource is fed back to the profile. It is possible to collect information resources that match the interests of the user without inputting whether or not the user is interested in the resources.

<< Specific Example 2 >> In specific example 2, information resources are displayed for each category.

<Structure> FIG. 5 is a diagram showing the structure of a second embodiment of the network information search system according to the present invention. The specific example 2 differs from the specific example 1 in that a data organizing unit 11 is added and that the audio reproducing unit 7a has a function of reproducing audio for each category. The other configurations are the same as those of the first embodiment, and the corresponding parts are denoted by the same reference numerals and description thereof will be omitted.

The data organizing unit 11 calculates the similarity for each category first classified by the user, and calculates the similarity between the newly collected information resource and each category. Then, the newly collected information resources are compared with the similarity of each category, and classified into the category having the closest similarity. The audio reproduction unit 7a has the function of the audio reproduction unit 7 in the specific example 1 and the function of reproducing audio for each category.

<Operation> FIG. 6 is an operation flowchart of the second embodiment. The operations of Steps S21 to S26 in the specific example 2 are the same as the operations of Steps S1 to S6 of the specific example 1 described above, and thus description thereof will be omitted.

[Step S27] When some information resources are collected, the user explicitly classifies the articles. FIG.
FIG. 5 is an explanatory diagram showing an example in which articles collected by a user are classified. As shown in the figure, the user creates a category in which the collected information resources are to be classified, and performs the classification. Here, the user uses the data input unit 1 to drag and classify the collected information resources with a mouse. In the example of FIG. 7, the user has three categories for the information resources collected from the electronic newspaper, namely, “Okinawa” and “Japanese Politi”.
cs ”and“ Else ”, where“ Okin
“awa” is “Article about US military base in Okinawa”, “Japanese
"Politics" refers to "articles on Japanese politics on US-Japan issues" and "Else" refers to "other articles on US-Japan relations." Also, the headlines of article1 to article7 are displayed in each article.

[Step S28] With respect to the category classified by the user, "importance of the base word for each category" is calculated. Here, "importance of base words for each category"
Is the average of the importance of the base word of each information resource for each category.

[Step S29] The collected information resources are classified according to the categories classified in step S28. Specifically, the similarity between the collected information resource and the “importance of the base word for each category” is calculated, and the collected information resource is classified into the category having the highest similarity.

FIG. 8 is an explanatory diagram showing an example of classification by the system side. As shown in the figure, article11 and article12 belong to category 1, and article13 and article14 belong to category 2.
In addition, article15 and article16 are classified into category 3.

As a method of presenting the collected information resources to the user, a case where the information resources are presented by voice and a case where the information resources are presented on a computer screen will be described in the following steps S30 and S31.

[Step S30] When presenting by voice,
The classified information resources are reproduced for each category in descending order of similarity with the user's profile.

[Step S31] When presented on a computer screen, the collected information resources can be viewed in descending order of user profile similarity for each category.

<Effects> As described above, according to the specific example 2,
When the user classifies the collected information resources into categories in advance, the system can classify the newly collected information resources into the category of the user.
Also, by converting the collected information resources into speech, the user can listen to the information by reproducing the category.

<Explanation of Usage Form> In the specific example 2, if the user creates a new category by looking at some collected information resources, the data organizing unit 11
The category may be used for classification at the time of the next collection. For example, the following is performed. In the example shown in FIG. 8, “Prosecutors pla”
n to quiz Yamaguchi ”, article18 as“ Yamaguch
i served as witness in golf deal ”, article 19,“ Yamaguchi fails to repay university project
At this time, the user creates a new category as “Yamaguchi” from article15, article17, article18, and article19. At the stage of the classification, the classification is performed in consideration of the newly generated category.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a specific example 1 in the network information search system of the present invention.

FIG. 2 is an explanatory diagram of each operation expression in a vector space model.

FIG. 3 is a flowchart showing an operation of a specific example 1 in the network information search system of the present invention.

FIG. 4 is a flowchart of audio presentation in the first specific example in the network information search system of the present invention.

FIG. 5 is a configuration diagram of a specific example 2 in the network information search system of the present invention.

FIG. 6 is a flowchart showing an operation of a specific example 2 in the network information search system of the present invention.

FIG. 7 is an explanatory diagram showing an example of classifying articles collected by a user of the specific example 2 in the network information search system of the present invention.

FIG. 8 is an explanatory diagram showing an example in which the system side of the specific example 2 in the network information search system of the present invention is classified.

[Explanation of symbols]

 2 Profile sentence processing unit 3 Filtering management unit 4 Collection agent 5 Data storage unit 7, 7a Voice reproduction unit 8 History storage unit 9 Data processing unit 11 Data reduction unit 100 Computer network

Claims (4)

[Claims] 1. A data storage unit for storing an information resource obtained by searching for an information resource on a network based on a profile sentence specified by a user, and reproducing the information resource stored in the data storage unit by voice. An audio playback unit that receives from the user an end designation of the audio playback, a history storage unit that takes a history of the audio playback time of the information resource in the audio playback unit, and a voice playback time based on the history of the history storage unit. A network information search system comprising: a data processing unit that, when longer than a predetermined threshold value, considers the information resource to be of interest to the user and feeds it back to a profile statement to create a new profile statement. . 2. The data sorting unit according to claim 1, wherein the data sorting unit sorts the information resources stored in the data storage unit into categories set in advance, and for each of the categories sorted by the data sorting unit. A network information retrieval system comprising: an audio reproduction unit that reproduces information resources. 3. The information reproducing apparatus according to claim 2, wherein when the user reproduces the information resource and the user designates a new category, the classification of the information resource thereafter is performed including the designated category. A network information search system comprising a data organizing unit for performing a search. 4. The network information search system according to claim 1, wherein the information resource reproduced by the audio reproduction unit is text data.