JP2007304642A - Document data sorting device and document data sorting program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve precision of sorting document data corresponding to content, and to sort documents from a point of view of correspondence between the content and the document data. <P>SOLUTION: This document data sorting device for sorting one or more document data relevant to content based on metadata corresponding to the content is provided with a metadata dividing means for dividing a document included in the metadata by predetermined units; a document data dividing means for dividing a document included in the document data by predetermined units; a corresponding expression detecting means for detecting expression sections corresponding to at least one of the plurality of content configuring the content based on one or more divided sentences to be acquired by the metadata dividing means and one or more divided sentences to be acquired by the document data dividing means; and a corresponding expression sorting means for sorting the document data based on the detection result to be acquired by the corresponding expression detecting means. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a document data classification device and a document data classification program, and more particularly to a document data classification device and a document data classification program for improving classification accuracy.

  Conventionally, a large number of document data to be classified using a computer or the like is individually digitized by keywords included in the document based on parameters such as keywords set in advance, and groups set in advance using the numerical values (See, for example, Patent Document 1 and Patent Document 2).

Further, the conventional method for digitizing a document uses the frequency of words appearing in the document and the meaning of the digitized words. Specifically, for example, when classifying document data using the appearance frequency of a word, a word representing the strength belonging to a specific group is set in advance, and the proportion of the word appearing in the document is set as the group. Can be classified as a group to which the document belongs. The group with the highest strength (likelihood) that belongs most when a particular document is focused on.
JP-A-1-188934 JP-A-5-89193

  By the way, the above-described classification method can be applied, for example, when classifying viewer's reverberation data such as opinions and requests for a program by a viewer who has watched content such as a program.

  However, in the conventional method described above, it is difficult to detect which part of the document included in the echo data is the echo for the program. In addition, it is difficult to grasp to which part of the entire program the detected echo is directed.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide a document data classification device and a document data classification program for improving classification accuracy.

  In order to solve the above problems, the present invention employs means for solving the problems having the following characteristics.

  The invention described in claim 1 is a document data classification device that classifies one or a plurality of document data related to the content based on metadata corresponding to the content, and the document included in the metadata is a predetermined document data. Metadata dividing means for dividing in units, document data dividing means for dividing a document included in the document data in predetermined units, one or a plurality of divided sentences obtained by the metadata dividing means, and the document data division Based on one or more divided sentences obtained by the means, the corresponding expression detecting means for detecting an expression part corresponding to at least one of the plurality of contents constituting the content, and the corresponding expression detecting means And corresponding expression classification means for classifying the document data based on the detected result.

  According to the first aspect of the present invention, the correspondence expression portion is detected by dividing and comparing the documents included in the metadata corresponding to the content and the document data, and classification is performed based on the detection result. As a result, the classification accuracy of the document data can be improved.

  According to a second aspect of the present invention, the correspondence expression detection unit is configured to perform a morpheme for each of one or a plurality of divided sentences obtained by the metadata dividing unit and one or a plurality of divided sentences obtained by the document data dividing unit. Analysis is performed, a tree structure is generated from dependency analysis using the analysis result, and the expression part is detected by comparing the generated tree structures.

  According to the second aspect of the present invention, a tree structure can be generated from the dependency analysis result, and the corresponding expression portion can be detected with high accuracy based on the tree structure.

  The invention described in claim 3 is characterized in that the correspondence expression detecting means sets a predetermined type for the document data based on a preset thesaurus.

  According to the third aspect of the present invention, the type can be set easily and accurately by setting in advance a thesaurus composed of characteristic expressions (phrases) as corresponding expressions.

  According to a fourth aspect of the present invention, the correspondence expression detecting unit discriminates an asserted sentence and an objective sentence for each of one or a plurality of divided sentences obtained by the document data dividing unit, and the determined asserted sentence The expression part is detected based on the above.

  According to the fourth aspect of the present invention, the expression part can be detected with higher accuracy based on the assertion sentence. Further, the type of document data can be set with high accuracy based on the assertion sentence.

  The invention described in claim 5 is characterized by comprising display screen generation means for generating a screen for displaying at least a graph or a table of the classification results obtained by the correspondence expression classification means.

  According to invention of Claim 5, a user etc. can be made to grasp | ascertain a classification result easily.

  The invention described in claim 6 is characterized in that the document data is reverberation data created by each viewer as a reverberation for the content provided to each viewer.

  According to the sixth aspect of the present invention, it is possible to classify the echo data from the viewer for the content with high accuracy.

  The invention described in claim 7 is a document data classification program for causing a computer to execute document data classification processing for classifying one or a plurality of document data related to the content based on metadata corresponding to the content. 1 obtained by the metadata dividing process for dividing the document included in the metadata in a predetermined unit, the document data dividing process for dividing the document included in the document data in a predetermined unit, and the metadata dividing process. Or a corresponding expression for detecting an expression portion corresponding to at least one of a plurality of contents constituting the content based on a plurality of divided sentences and one or a plurality of divided sentences obtained by the document data dividing process. A pairing process for classifying the document data based on a detection process and a detection result obtained by the correspondence expression detection process. Executing the expressions classification process on the computer.

  According to the seventh aspect of the present invention, the correspondence expression portion is detected by dividing and comparing the documents included in the metadata corresponding to the content and the document data, and classification is performed based on the detection result. As a result, the classification accuracy of the document data can be improved. In addition, the document data classification process can be easily realized by installing the program.

  According to the present invention, it is possible to improve the classification accuracy of document data corresponding to content, and it is possible to realize classification from the viewpoint of which part of the content corresponds to document data.

<Features of the present invention>
The present invention classifies one or a plurality of document data related to content based on metadata corresponding to the content. Specifically, for example, there is a content composed of a plurality of contents such as a broadcast program, and one or a plurality of sentences (documents) included in reverberation data from viewers or the like with respect to the contents are converted into punctuation marks, end marks, etc. The unit is divided into predetermined units, and an expression closest to the content is detected from among them.

  In addition, not only the appearance frequency of a word but also the dependency structure of a sentence and the closeness of the word are used to detect the closest expression. In the present invention, since the closest expression is detected by the number of classification target sentences, this is classified into an appropriate number of groups by a conventional classification technique.

<Embodiment>
Next, a preferred embodiment of the document data classification apparatus and document data classification program according to the present invention having the above-described features will be described with reference to the drawings. In the embodiment described below, a broadcast program is used as an example of content, and an impression, an opinion, a request, a comment, a questionnaire result, etc. from a viewer who has watched the broadcast program as an example of document data related to the content. Use echo data. The above-described echo data classification technique will be described as an example of the document data classification technique.

  FIG. 1 is a diagram showing an example of a schematic configuration of a document data classification device according to the present invention. The document data classification device 10 shown in FIG. 1 includes a broadcast program metadata dividing unit 11, a viewer echo data dividing unit 12, a corresponding expression detecting unit 13, a corresponding expression classifying unit 14, a storage unit 15, and a display screen. And generating means 16.

  The broadcast program metadata dividing means 11 inputs the broadcast program metadata 21 that is metadata corresponding to the broadcast program as a natural sentence. In other words, the broadcast program metadata 11 is input to the broadcast program metadata dividing unit 11 with the metadata format used as the program content intact. Note that the natural sentence described above has one or more sentences. Furthermore, the text in the broadcast program metadata includes a content title, a detailed description of the content, and the like in addition to dialogue and subtitles.

  Broadcast program metadata dividing means 11 inputs broadcast program metadata 21 and divides all documents included in the input broadcast program metadata 21 into predetermined units. Specifically, the broadcast program metadata dividing unit 11 divides the program into sentence units based on, for example, a punctuation mark, a punctuation mark, and a period mark. The broadcast program metadata dividing unit 11 outputs the divided sentence of the divided metadata to the corresponding expression detecting unit 13.

  Further, the viewer echo data dividing means 12 inputs the viewer echo data 22 for the broadcast program. Note that the viewer response data are impressions, opinions, requests, comments, questionnaire results, etc. from viewers who watched the broadcast program, and these data are sent by e-mail, telephone or FAX, etc. It is data that digitizes a document. Note that one or more viewer echo data 22 (viewer echo data 1, 2,..., N in FIG. 1) exist, and the viewer echo data 22 includes one or more sentences. Shall. In addition, each of the viewer echo data 1 to N may have information (gender, age, name, etc.) regarding the viewer who created the echo data.

  The viewer echo data dividing means 12 receives one or a plurality of viewer echo data 22 and divides all the documents included in the input viewer echo data 22 into predetermined units. Specifically, the viewer reverberation data dividing unit 12 divides the sentence in units of sentences based on, for example, a punctuation mark, a punctuation mark, and a period end. Each of the viewer echo data is divided into one sentence based on a punctuation mark, a punctuation mark, an end mark, and the like. Further, the viewer echo data dividing unit 12 outputs the divided sentence of the divided viewer echo data to the corresponding expression detecting unit 13.

  Corresponding expression detection means 13 compares all the divided sentences of broadcast program metadata 21 and the divided sentence of viewer echo data 22 and measures the likelihood (likelihood) of the corresponding expression for each divided sentence. Output).

  Here, FIG. 2 is a diagram showing an example for explaining a divided sentence comparison method according to the present invention. As shown in FIG. 2, the corresponding expression detection means 13 receives broadcast program metadata divided sentences 1 to N and the divided sentences 1 to M in a certain viewer echo data k, and the corresponding expressions are divided between the divided sentences. A sentence with a predetermined likelihood among combinations of the number of divided sentences obtained from broadcast program metadata and the number of divided sentences obtained from viewer echo data. Is the viewer's echo data for the broadcast program.

  Note that the predetermined likelihood is, for example, the one with the highest likelihood of the corresponding expression in the comparison of the number of divided sentences obtained from the broadcast program metadata and the divided sentences obtained from the viewer echo data, or a preset likelihood. It can be set to detect more than likelihood. Therefore, if the likelihood is highest, one corresponding expression can always be detected for the viewer echo data 1 to N from each viewer. Thereby, it is possible to make an overall evaluation of the broadcast program such as what kind of opinions the viewer has.

  Further, when detecting a corresponding expression part having a likelihood equal to or higher than a preset likelihood, one or a plurality of corresponding expressions can be detected with respect to a certain viewer echo data. Further, if the likelihood set in advance is not satisfied, the corresponding expression is not detected for the viewer echo data.

  Therefore, when a plurality of reverberations (documents) exist in a certain viewer reverberation data, each corresponding expression can be detected, so that a classification without omission can be realized. In addition, since the viewer echo data that does not satisfy the preset likelihood is data unrelated to the program, it can be excluded from the classification target. As a result, highly accurate document data classification can be realized.

  In this way, by comparing the segmented sentence obtained from the broadcast program metadata and the segmented sentence obtained from the viewer echo data, by detecting the corresponding expression, which part of the document included in the viewer echo data is It is possible to detect whether the response is for a program. In addition, it is possible to detect with high accuracy which part of the plurality of contents (for example, the beginning, the beginning, the middle, the end, etc.) of the program corresponds to the response.

  Corresponding expression detection means 13 sets the type of reverberation as one parameter for classifying the detected corresponding expression by corresponding expression classification means 14 described later. A specific description of correspondence expression detection and type setting in the correspondence expression detection unit 13 will be described later. Corresponding expression detection means 13 outputs the corresponding expression detection result obtained for the above-mentioned viewer echo data to corresponding expression classification means 14. The detection result includes, for example, the likelihood of the corresponding expression described above, the broadcast program metadata corresponding part where the corresponding expression is detected, the corresponding part of the viewer echo data, identification information for identifying a plurality of viewer echo data, And at least one piece of information arbitrarily selected from the type of echo and the like.

  Next, the correspondence expression classification unit 14 classifies the viewer echo data based on the detection result obtained from the correspondence expression detection unit 13 and outputs the classification result 23. Here, as the classification result 23, for example, as shown in FIG. 1, at least one classification is assigned to each viewer echo data. Note that the output method of the classification result in the correspondence expression classification unit 14 is not limited to this, and for example, a list of viewer echo data may be output for each classification. FIG. 3 is a diagram illustrating an example of outputting a viewer echo data list for each classification. In addition, as a classification method, when classifying viewer echo data according to each preset classification (group), for example, the likelihood given to each viewer echo data can be applied as a reference. Moreover, when classifying without setting each classification (group) in advance, for example, a centroid method, a k-NN method, or the like can be applied.

  Therefore, the corresponding expression classification unit 14 performs the above-described classification for all the viewer echo data, and outputs the viewer echo data to which some sort of grouping information is added. The corresponding expression classification unit 14 can also output the obtained classification result to the storage unit 15 and / or the display screen generation unit 16.

  The storage unit 15 stores the classified viewer echo data obtained by the corresponding expression classification unit 14. The storage unit 15 can also store the input broadcast program metadata 21 and viewer echo data 22, and can also store detection results obtained by the corresponding expression detection unit 13. Various data stored in the storage unit 15 can be read out as necessary. The broadcast program metadata dividing unit 11, the viewer echo data dividing unit 12, the corresponding expression detecting unit 13, or the corresponding expression classification described above. It can be used by means 14 or the like.

  Further, the display screen generation unit 16 edits the classification result obtained by the correspondence expression classification unit 14 into a display format that is easy to see for a user or the like such as a table or a graph, and presents the generated display screen to the user or the like using a display device or the like. . Thereby, a user etc. can be made to grasp | ascertain a classification result easily. An example of the generated display screen will be described later.

  As described above, according to the document data classification apparatus 10, the means for inputting and dividing the broadcast program metadata, the means for inputting and dividing the viewer echo data, and the corresponding expressions by comparing the respective divided data And a means for classifying a corresponding portion according to a scale (likelihood) indicating the proximity of the corresponding expression, so that when the metadata is given to the broadcast program, the metadata 1 or a plurality of viewer echo data can be classified with high accuracy.

  That is, it is possible to improve the classification accuracy of the document data corresponding to the content, and it is possible to realize the classification from the viewpoint of the document data corresponding to which part of the content.

<Correspondence expression detection>
Next, the content of correspondence detection by the correspondence expression detection means 13 described above will be specifically described. First, as shown in FIG. 2, when the same content expression is detected in the broadcast program metadata divided sentences 1 to N and the viewer questionnaire k divided sentences 1 to M, the detection is based on the phrase and structure of each divided sentence. In order to determine, the same expression can be detected using, for example, linear approximation (linear approximation) or the like. In addition, the same expression can be detected by a tree structure, linear notation or the like for the difference in structure, and the same expression can be detected by a thesaurus, ontology, knowledge processing or the like in which data is hierarchically organized. It is also possible to detect whether or not the expressions are the same by inputting the words and structures to be compared to a binary classifier such as SVM (Support Vector Machine).

  Here, as described above, the correspondence expression detection unit 13 outputs a scale (likelihood) representing the proximity of the correspondence expression for the two divided sentences to be compared. Specifically, the likelihood is determined based on the distance of the tree structure composed of sentence morphological analysis and dependency analysis. In addition, two sentence comparison outputs obtained from a combination of the number of divided sentences of broadcast program metadata and the number of divided sentences of certain viewer echo data k satisfy the predetermined likelihood.

  Here, it demonstrates using a specific example. FIG. 4 is a diagram illustrating an example of broadcast program metadata and viewer echo data. FIG. 4 shows broadcast program metadata 31 for a certain program and echo data 32 after a viewer views a program (content) corresponding to the broadcast program metadata 31.

  At this time, the reverberation data 32 includes a sentence “The Earth is turning around the sun”, and the underlined part of this sentence is “The Earth is turning around the sun”. It is a corresponding expression of the content of the broadcast program “Earth is turning around the sun” shown in the underlined portion of the metadata 31, and it can be determined that “Keshakuren” is the response. Here, the correspondence expression detection in this case will be described.

  FIG. 5 is a diagram showing an example of morphological analysis and dependency analysis in the present invention. Corresponding expression detection means 13 performs morphological analysis and morphological analysis as shown in FIG. 5 for a sentence “... the earth is turning around the sun” included in broadcast program metadata 31. Based on dependency analysis. Similarly, morphological analysis and dependency analysis are performed on the echo data 32.

  In other words, the dependency indicates the modification relationship of the phrases, and since each phrase corrects only one phrase, there is only one dependency destination. Therefore, the data structure of the dependency analysis result is expressed as a tree structure as shown in FIG. 5, and the similarities between the tree structures are compared.

  Next, correspondence expressions are detected by comparing tree structures obtained by the dependency analysis described above. FIG. 6 is a diagram illustrating an example of tree structure comparison. As shown in FIG. 6, the comparison in the tree structure obtained from each dependency analysis is performed by, for example, comparing the linear structures of the trees and detecting the difference as the distance between the tree structures. For example, if two tree structures are ordered trees, it is detected how much insertion, replacement, deletion, etc., should be performed in order to change from one to the other tree structure, and the similarity ( Distance). As a result, the correspondence expression is statistically learned not only for the characteristic words but also for the structure representing the modification relationship, so that the portion of the document included in the echo data is the part corresponding to the program, and the program It is possible to accurately detect which part of the echo is reflected. Therefore, it is possible to realize classification for echo data with high accuracy. The divided sentence comparison method in the present invention is not limited to this.

<Type setting>
Next, the type setting contents in the correspondence expression detection unit 13 will be specifically described. After detecting the corresponding expression portion for the echo data, the correspondence expression detection unit 13 sets the type of whether the echo from the viewer is “good” or “bad”. Here, FIG. 7 is a diagram illustrating an example for setting the type of echo. In FIG. 7, the sentence “XX is bad for my son ’s education because it is bad for my son.” (XX indicates, for example, “the earth is turning around the sun” described above). A tree structure based on the morphological analysis and dependency analysis results described above is shown.

  Here, in order to detect the type of echo of this sentence, a preset thesaurus is used. FIG. 8 is a diagram illustrating an example of a thesaurus. FIG. 8 shows an example of a thesaurus of expressions representing subjective evaluation. That is, the thesaurus according to the present invention is a system in which characteristic expressions (phrases) corresponding to the contents of a program are systematically prepared as reverberation expressions, and these data are systematized hierarchically.

  Specifically, a thesaurus whose program rating is “good” (clean, useful, interesting, beautiful, simple, etc.) or a program whose program rating is “bad” (adverse education, content Is not true, and the like is set in advance, the type of the echo data can be set with high accuracy, and the viewer's echo data can be classified based on this content.

  Here, if an expression preset in the thesaurus or the like is set separately from the contents of the program, an appropriate likelihood corresponding to the contents of the program cannot be obtained, and therefore an appropriate classification cannot always be performed. Specifically, for example, “Efficiency has advanced and night trains can no longer run” is a “good response” for broadcasts related to the noise of residents near the track. “Bad response”. Therefore, in the present invention, by setting the expression corresponding to the contents of the program, the type can be set with high accuracy even in the above case.

  In this way, by using comparison, tree structure learning, thesaurus etc. by tree structure using morphological analysis and dependency analysis, it is possible to detect what kind of response to which part of the broadcast program with high accuracy. it can. Note that the type of type is not limited to good or bad, and can be arbitrarily set to other types according to the contents of the echo data.

<Other examples>
Here, the detection of the corresponding expression and the setting of the type in the present invention are not limited to this. For example, by detecting the asserted sentence, detecting the same expression, detecting the causal relationship to the dependency in the sentence, etc. It is possible to determine what the viewer has made to the viewer.

  That is, for example, in the echo data 32 shown in FIG. 4 described above, “the earth is turning around the sun”, which is the same expression as the broadcast program metadata, is the object of assertion, and the contents claimed by “Keshikan” Therefore, it can be determined that this sentence has a causal relationship and is a claim sentence.

  In addition to the above, for example, in the echo data, it is determined whether the divided sentence is an asserted sentence or an objective sentence based on a preset condition, and the asserted sentence is determined from the determination result. The correspondence expression can be detected and the type can be set.

  Here, the assertion sentence is a sentence expressing a subjective opinion. For example, the assertion sentence may have expressions that cannot be shared with many people, or may have linguistic features unique to subjective expressions. In addition, as an assertion sentence, "I think", "I like", etc. correspond, for example. In addition, an objective sentence is a part that describes a so-called fact, and is usually an expression that can be shared with many people, and may have a linguistic feature unique to the objective expression. As an objective sentence, for example, “is a fact” is applicable. Note that whether the sentence is an assertion sentence or an objective sentence can be determined by preparing learning data such as a corpus in which correct data is registered in advance and using the corpus.

  That is, by extracting the assertion sentence in the viewer echo data, it is possible to extract the portion (the main point) that the viewer wants to assert most in the viewer echo data. In addition, by performing detection based on the above-described tree structure using an assertion sentence, it is possible to perform correspondence expression detection and type setting with higher accuracy.

<About the claim>
Here, the claim sentence mentioned above is demonstrated concretely. As the assertion sentence, for example, there are “ambiguousness / relativeity”, “embedded sentence”, and the like.

  “Ambiguity / relativity” is almost pure or assertive (preferred claim), for example, “I like apples”, “I don't like bananas”, etc. As another example, there is a claim having an external standard (ethical claim), for example, “clear education is not preferable”, “decrease in academic ability is not preferable”, and the like. Furthermore, there is an assertion (logical assertion) having a causal relationship, for example, “the wind blows and the candy shop makes money”.

  The “embedded sentence” includes, for example, “I say I like apples (opinion + facts)”, “I like ringing birds (facts + opinions)”, and the like.

  In addition, the asserted sentence can be determined by an auxiliary verb such as a guess (“maybe”) or possible (“can do”), a conjunctive such as forward or reverse, and the like.

  When there are a plurality of asserted expressions in a certain viewer echo data (for example, when a good opinion and a bad opinion are included), the distribution of scores belonging to the upper group of each expression and the score of the lower group The boundary is set from the result. Therefore, it is possible to set a plurality of types from the viewer echo data, and to detect the corresponding expression part. In general, as a method of dividing into K sets, for example, a K-means method or the like can be used.

<Example of generated display screen>
Next, an example of a display screen generated by the display screen generation unit 16 will be described with reference to the drawings. FIG. 9 is a diagram showing an example of a display screen for classification results in the present invention.

  As shown in FIG. 9, as a display result, for example, a pie chart 41 indicates which part of the entire program has a large response. Further, for example, when a display area for “resonance to the middle board” in the pie chart 41 is selected (clicked) on the screen, the classification result and the breakdown (ratio) of the types for the “resonance to the middle board” are displayed in the pie chart 42. Is displayed.

  Further, when a display area for “good”, for example, is selected from the pie chart 42, the subcategory “good” of the response to the middle board and its breakdown are displayed by the pie chart 43.

  As another example, for example, when the display area for “resonance to the middle stage” is selected on the screen, among the audience reverberation data classified as the middle stage, among “start”, “middle”, and “end” The pie chart 44 may display the middle classification and its breakdown showing which part is the echo. Furthermore, by selecting the pie charts 43 and 44, the specific contents of the echo (the full text of the echo data, the corresponding expression portion, information on the viewer who created the echo data, etc.) can be displayed.

  As described above, by displaying the classification result as a table, a graph, or the like by the display screen generation unit 16, it is possible for the user or the like to easily grasp the classification result. In the present invention, it is not limited to a pie chart, it can also be displayed graphically by a line graph, a radar graph, etc., and each item obtained as a classification result and its breakdown (ratio) can be displayed in a table format. You can also. Further, for example, in the case of a program that is broadcast daily or weekly, the previous classification result may be stored in the storage means 15 and the statistical result may be generated from the data and displayed.

<Execution program>
Here, the document data classification device 10 described above can perform document data classification in the present invention using the dedicated device configuration described above, but generates an execution program for causing a computer to execute the processing in each configuration. For example, the document data classification process according to the present invention can be realized by installing the program in a general-purpose personal computer, server, or the like.

<Hardware configuration>
Here, a hardware configuration example of a computer capable of executing document data classification processing according to the present invention will be described with reference to the drawings. FIG. 10 is a diagram illustrating an example of a hardware configuration capable of realizing document data classification processing according to the present invention.

  10 includes an input device 51, an output device 52, a drive device 53, an auxiliary storage device 54, a memory device 55, a CPU (Central Processing Unit) 56 that performs various controls, and a network connection device. 57 are connected to each other by a system bus B.

  The input device 51 includes a keyboard and a pointing device such as a mouse operated by the user, a voice input device, and the like, and inputs various operation signals and voice signals such as a program execution instruction from the user. The output device 52 has a display, a speaker, and the like that display various windows and data necessary for operating the computer main body for performing the processing in the present invention. Display or audio output is possible.

  In the present invention, the execution program installed in the computer main body is provided by a portable recording medium 58 such as a CD-ROM. The recording medium 58 on which the program is recorded can be set in the drive device 53, and the execution program included in the recording medium 58 is installed in the auxiliary storage device 54 from the recording medium 58 via the drive device 53.

  Further, the drive device 53 can record the execution program according to the present invention in the recording medium 58. As a result, the recording medium 58 can be used for easy installation on a plurality of other computers, and document data classification processing can be easily realized.

  The auxiliary storage device 54 is a storage means such as a hard disk, and can store an execution program according to the present invention, a control program provided in a computer, etc., and perform input / output as necessary. The auxiliary storage device 54 can also be used as storage means for storing the above-mentioned broadcast program metadata, viewer echo data, classified viewer echo data, and the like.

  Based on a control program such as an OS (Operating System) and an execution program read from the auxiliary storage device 54 and stored in the memory device 55, the CPU 56 inputs various calculations and data with each hardware component. Each processing in the document data classification processing can be realized by controlling processing of the entire computer such as output. Various information necessary during the execution of the program can be acquired from the auxiliary storage device 54 and stored.

  The network connection device 57 obtains an execution program from another terminal connected to the communication network or executes the program by connecting to a communication network such as a telephone line or a LAN (Local Area Network) cable. The execution result obtained in this way or the execution program in the present invention can be provided to other terminals.

  With the hardware configuration as described above, a document data classification process can be realized at a low cost without requiring a special device configuration. In addition, the document data classification process can be easily realized by installing the program.

<Document data classification procedure>
Next, a document data classification process procedure using an execution program (document data classification program) according to the present invention will be described with reference to a flowchart. FIG. 11 is a flowchart illustrating an example of a document data classification processing procedure.

  In FIG. 11, first, broadcast program metadata corresponding to content such as a broadcast program is input (S01), and one or more documents included in the input broadcast program metadata are divided into predetermined units such as sentence units, for example. (S02). Next, one or a plurality of viewer echo data obtained from the viewer is input (S03), and one or a plurality of documents included in the input viewer echo data is divided into predetermined units such as sentence units (for example). S04).

  Next, as described above, the corresponding expression is detected based on the divided sentence of the broadcast program metadata obtained in S02 and the divided sentence of one viewer echo data obtained in S04 (S05). In S05, the echo type is also set. Further, after the process of S05 is completed, it is determined whether or not the corresponding expression has been detected for all viewer echo data (S06).

  Here, when the corresponding expression detection has not been completed for all viewer echo data (NO in S06). Returning to S05, the corresponding expression is detected based on the divided sentence of the broadcast program metadata obtained in S02 and the divided sentence of other viewer echo data that has not yet been processed. Further, in the processing of S06, when the corresponding expression detection is completed for all the viewer echo data (YES in S06), the viewer echo data is classified (S07).

  Further, a display screen is generated using the viewer echo data classified in S07 (S08), and the generated display screen is output to a display or the like (S09). The document data classification process described above can improve the classification accuracy of document data. In addition, the document data classification process can be easily realized by installing the program.

  As described above, according to the present invention, the corresponding expression portion is detected by dividing and comparing the document included in the metadata corresponding to the content and the document data, and classification is performed based on the detection result. As a result, the classification accuracy of the document data can be improved.

  Thus, for example, when a broadcast station or the like starts broadcasting with metadata added to a broadcast program, or when metadata for a broadcast program is provided from a content provider other than the broadcast station, these metadata are set. By using it as an index for classifying the response to a broadcast program, it is possible to obtain with high accuracy what kind of response has occurred. Therefore, the broadcasting station or the like can perform an appropriate analysis on the program response and reflect it in the program production.

  The content in the present invention is not limited to general programs such as news and variety, and can be applied to any content such as moving images, music, movies, text information, and the like.

  Although the preferred embodiment of the present invention has been described in detail above, the present invention is not limited to the specific embodiment, and various modifications are possible within the scope of the gist of the present invention described in the claims. Can be changed.

It is a figure which shows an example of schematic structure of the document data classification device in this invention. It is a figure which shows an example for demonstrating the comparison method of the division sentence in this invention. It is a figure which shows an example which outputs the list of the viewer echo data for every classification. It is a figure which shows an example of broadcast program metadata and viewer echo data. It is a figure which shows an example of the morphological analysis and dependency analysis in this invention. It is a figure which shows an example of a tree structure comparison. It is a figure which shows an example for setting the classification of reverberation. It is a figure which shows an example of a thesaurus. It is a figure which shows an example of the display screen of the classification result in this invention. It is a figure which shows an example of the hardware constitutions which can implement | achieve the document data classification | category process in this invention. It is a flowchart which shows an example of a document data classification | category process procedure.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Document data classification | category apparatus 11 Broadcast program metadata division | segmentation means 12 Viewer reverberation data division | segmentation means 13 Correspondence expression detection means 14 Correspondence expression classification means 15 Storage means 16 Display screen generation means 21, 31 Broadcast program metadata 22 Viewer reverberation data 23 Classification result 32 Echo data 41 to 44 Pie chart 51 Input device 52 Output device 53 Drive device 54 Auxiliary storage device 55 Memory device 56 CPU
57 Network connection device 58 Recording medium

Claims (7)

A document data classification device for classifying one or a plurality of document data related to the content by metadata corresponding to the content,
Metadata dividing means for dividing a document included in the metadata by a predetermined unit;
Document data dividing means for dividing a document included in the document data by a predetermined unit;
Based on one or more divided sentences obtained by the metadata dividing means and one or more divided sentences obtained by the document data dividing means, at least one of a plurality of contents constituting the content Correspondence expression detection means for detecting a corresponding expression part;
A document data classification apparatus comprising: a correspondence expression classification unit that classifies the document data based on a detection result obtained by the correspondence expression detection unit. The correspondence expression detecting means includes
A morphological analysis is performed on each of one or more divided sentences obtained by the metadata dividing means and one or more divided sentences obtained by the document data dividing means, and a tree structure is obtained from dependency analysis using the analysis results. The document data classification apparatus according to claim 1, wherein the expression part is detected by generating and comparing the generated tree structures. The correspondence expression detecting means includes
3. The sentence data classification apparatus according to claim 1, wherein a predetermined type is set for the document data based on a preset thesaurus. The correspondence expression detecting means includes
The claim sentence and the objective sentence are discriminated for each of one or a plurality of divided sentences obtained by the document data dividing means, and the expression part is detected based on the discriminated sentence. 4. The document data classification device according to any one of items 1 to 3.   5. The document data according to claim 1, further comprising a display screen generation unit configured to generate a screen for displaying at least a graph or a table of the classification result obtained by the correspondence expression classification unit. Classification device. The document data is
The document data classification device according to claim 1, wherein the data provided by each viewer is echo data created by each viewer as a response to the content provided to each viewer. A document data classification program for causing a computer to execute document data classification processing for classifying one or a plurality of document data related to the content by metadata corresponding to the content,
Metadata division processing for dividing a document included in the metadata in a predetermined unit;
Document data division processing for dividing a document included in the document data in a predetermined unit;
Based on one or more divided sentences obtained by the metadata dividing process and one or more divided sentences obtained by the document data dividing process, at least one of a plurality of contents constituting the content A corresponding expression detection process for detecting a corresponding expression part;
A document data classification program for causing a computer to execute a corresponding expression classification process for classifying the document data based on a detection result obtained by the corresponding expression detection process.

Images