JP6635966B2 - Visualization device, visualization method, and program - Google Patents

Description

  The present invention relates to a visualization device, a visualization method, and a program.

  Conventionally, when the classification to which a document (text) belongs is multi-valued (three or more), it is estimated which classification the document belongs to (estimation of N-value classification), and one of the most likely classification results is displayed. There are known classification determination methods. For example, when a document describing a failure state of a personal computer is classified into failure types, when the classification is “power supply”, “HDD (Hard Disc Drive)”, “display”, and “fan”, for example, “ When the document "Nothing is displayed on the screen even when the power is turned on" is input, the input document is acquired as shown in FIG. 17 (step S51). When the document is obtained in step S51, it is estimated that the document belongs to the class "display" from the plurality of classes by estimating the N-value class (step S52). When the N-value classification is estimated in step S52, the result information including the classification result of "display" is stored (step S53) and displayed on a display or the like (step S54). As a classifier used for estimating the N-value classification, it is known that an SVM (Support Vector Machine) has very high classification performance (for example, see Non-Patent Document 1). There is also a method of giving the reliability of the classification result and displaying one or more classification results together with the reliability.

Daiya Takamura, "Introduction to Machine Learning for Language Processing," Corona Publishing Co., August 10, 2010, p. 117-127

  However, when the user tries to determine the classification by referring to the classification result displayed by the conventional technology as described above, if only one classification result is displayed, the reliability of the classification result is not known, In some cases, it is difficult for the user to determine whether to adopt the classification result. Even when the classification result is displayed together with the reliability, the reliability of the displayed classification result is known, but a criterion for adopting the classification result (for example, what percentage of reliability can be adopted) It can be difficult to decide. Further, when a document includes a plurality of sentences, it may not be possible to determine which sentence contributes to the classification determination.

  Therefore, an object of the present invention made in view of such a point is to visualize a user's criterion for a classification result of an input document.

In order to solve the above-described problem, a visualization device according to the present invention includes a document acquisition unit that acquires a document based on a user operation, a division unit that divides the document into a plurality of sentences, the document and the plurality of documents. each respective statements, advance the binary classifier estimation unit that estimates whether belonging to each of a plurality of classifications defined, pre-SL document and the plurality of sentences, the document and the plurality of statements, each said classification And a display unit for displaying result information including a classification result indicating whether or not the data belongs to the group.
Further, the visualization device according to the present invention, based on a user operation, a document acquisition unit that acquires a document and a manual assignment classification for the document, a division unit that divides the document into a plurality of sentences, A binary classification estimating unit for estimating whether each of the plurality of sentences belongs to each of a plurality of predetermined classifications, and a classification estimated by the document and the plurality of sentences, the binary classification estimating unit Results, and the result information including the classification result by the manual assignment classification, and the classification result included in the result information is the estimated classification result, or information indicating whether the classification result by the manual assignment classification And a display unit for displaying

Further, a visualization method according to the present invention is a visualization method executed by a visualization device for estimating a classification of a document, wherein the step of acquiring a document based on a user operation and the step of dividing the document into a plurality of sentences dividing the document and each of the plurality of statements, estimating whether belonging to each of a plurality of classifications predetermined said document and said plurality of statements, the document and the plurality of sentences each of which comprising the steps of: displaying the result information including the classification result indicating whether belonging to each of the classification.
Further, the visualization method according to the present invention includes a step of acquiring a document and a classification result by a manual assignment classification for the document based on an operation of a user; a step of dividing the document into a plurality of sentences; And a step of estimating whether or not each of the plurality of sentences belongs to each of a plurality of predetermined classifications, and the document and the plurality of sentences, the estimated classification result, and the classification result by the manual assignment classification. Displaying the result information including the result information including, and the classification result included in the result information, the estimated classification result, or information indicating whether the classification result is based on the manual assignment classification, It is characterized by including.

  Further, a program according to the present invention is a program for causing a computer to function as the visualization device.

  According to the present invention, it is possible to visualize a user's criterion for a classification result of an input document.

It is a block diagram showing the example of composition of the visualization device concerning a 1st embodiment. It is a figure showing the example of a display of the display part in a 2nd embodiment. FIG. 5 is a diagram illustrating an example of result information according to the first embodiment. 5 is a flowchart illustrating an example of a visualization method according to the first embodiment. It is a block diagram showing an example of composition of a visualization device concerning a 2nd embodiment. It is a figure showing the example of a display of the display part in a 2nd embodiment. It is a figure showing an example of the result information of a 2nd embodiment. It is a flow chart which shows an example of the visualization method concerning a 2nd embodiment. It is a figure showing the example of a display of the display part in a 3rd embodiment. It is a figure showing an example of the result information of a 3rd embodiment. It is a flow chart which shows an example of the visualization method concerning a 3rd embodiment. It is a figure showing the example of a display of the display part of a 4th embodiment. It is a figure showing an example of the result information in a 4th embodiment. 13 is a flowchart illustrating an example of a visualization method according to a fourth embodiment. It is a figure showing an example of a 5th embodiment result information. It is a block diagram showing the example of composition of the visualization device concerning the modification of a 1st embodiment. 9 is a flowchart illustrating an example of a classification determination method according to the related art.

[First Embodiment]
First, the functional configuration of the first embodiment of the present invention will be described with reference to FIG. FIG. 1 is a functional configuration diagram of the visualization device 1 according to the first embodiment.

  As shown in FIG. 1, the visualization device 1 includes a document acquisition unit 11, a binary classification estimation unit 12, a storage unit 13, and a display unit 14.

  The document acquisition unit 11 acquires a document based on a user operation. The document acquired by the document acquisition unit 11 is arbitrary, but in the present embodiment, as an example, the user uses a Web application to display “Even if the power is turned on” on a screen shown in FIG. Nothing is displayed on the screen. "And press the send button. In this case, the document acquisition unit 11 acquires the input document.

  The binary classification estimating unit 12 estimates whether or not the document acquired by the document acquiring unit 11 belongs to each of a plurality of classifications based on learning data by a known machine estimation method. Although the classification may be arbitrary, in the present embodiment, examples of the classification will be described as “power supply”, “HDD”, “display”, and “fan”.

  In this case, the binary classification estimation unit 12 estimates whether or not the document acquired by the document acquisition unit 11 belongs to “power”, “HDD”, “display”, and “fan” based on the learning data. . When the document is “Nothing is displayed on the screen even when the power is turned on”, the binary classification estimation unit 12 estimates that the document belongs to the classification of “power” based on the learning data (○), It is assumed that it is estimated that it does not belong to the classification of “HDD” (×), it is estimated that it belongs to the classification of “display” (○), and it is estimated that it does not belong to the classification of “fan” (×).

  The storage unit 13 stores the result information. As shown in FIG. 3, the result information includes an ID (Identification) for uniquely identifying the document, the document, and a classification result. The classification result is information indicating whether the document belongs to each classification (o) or not (x), estimated by the binary classification estimating unit 12 as described above. For example, the storage unit 13 stores the power “O”, HDD “X”, display “O”, and fan “ “X” is stored as the classification result.

  The display unit 14 displays the result information stored in the storage unit 13. Specifically, the display unit 14 displays the document and the classification result on a display or the like. For example, the display unit 14 displays the classification result estimated by the binary classification estimation unit 12 as indicating that the document belongs to the classification of “power” and “display” as shown in FIG. 2B. The display unit 14 may transmit the result information to another information processing device via a communication network.

  Subsequently, a visualization method of the visualization device 1 according to the first embodiment will be described with reference to a sequence diagram shown in FIG.

  First, the document acquisition unit 11 acquires a document based on a user operation (Step S11). In the present embodiment, the document acquisition unit 11 acquires, for example, a document “Nothing is displayed on the screen even when the power is turned on”.

  When the document is obtained in step S11, the binary classification estimating unit 12 estimates the binary classification of the document (step S12). Specifically, the binary classification estimating unit 12 estimates whether the document belongs to each of the predetermined classifications. In the present embodiment, for example, the binary classification estimating unit 12 estimates whether or not the document belongs to the classification “power supply”.

  When the binary classification is estimated in step S12, the binary classification estimation unit 12 determines whether or not all the classifications have been estimated (step S13). In the example of the present embodiment, the binary classification estimating unit 12 determines whether the binary classification has been estimated for all the classifications of “power supply”, “HDD”, “display”, and “fan”.

  If all the classifications have not been estimated in step S13, the process returns to step S12, and the binary classification estimation unit 12 estimates whether the document belongs to a classification that has not been estimated. If all the classifications have been estimated in step S13, the storage unit 13 stores the result information including the classification results (step S14).

  When the result information is stored in step S14, the display unit 14 displays the result information on a display or the like (step S15).

  As described above, according to the first embodiment, the classification results of all the classifications are stored and displayed for the input document. If a document belongs to a plurality of categories, it means that there is any category, and if it does not belong to any category, it means that the document is lacking. For example, in the case of the classification result shown in FIG. 3, it is indicated that a document "Nothing is displayed on the screen even when the power is turned on" may belong to both the classifications "power" and "display". As a result, it is possible to determine the classification by considering excessive or insufficient information in the input document, deleting unnecessary information, and supplementing missing information. For example, in the case of the classification result shown in FIG. 3, if "power lamp of personal computer is not lit" is supplemented as missing information, the classification can be determined to be "power".

[Second embodiment]
Next, a second embodiment of the present invention will be described with reference to the drawings.

  As shown in FIG. 5, the visualization device 2 according to the second embodiment, like the visualization device 1 according to the first embodiment, includes a document acquisition unit 11, a binary classification estimation unit 12, a storage unit 13, and It has a display unit 14 and further has a dividing unit 15. Note that the same constituent blocks as those of the visualization device 1 according to the first embodiment are denoted by the same reference numerals, and description thereof will be omitted as appropriate.

  The dividing unit 15 divides the document acquired by the document acquiring unit 11. The dividing unit 15 extracts, for example, sentences included in the document one by one. For example, the dividing unit 15 can search for a period "." And extract a character string before the first period as one sentence from the first character of the document. In addition, the dividing unit 15 can extract a character string before the next period from the character after the period as one sentence. Further, when the document has a line feed, the dividing unit 15 may divide the document before and after the line feed.

  As an example, the user uses the Web application to input a document “The operation is very slow and the fan does not make a sound. There is a case where the fan does not start up” on a screen as shown in FIG. Press the send button. In this case, the document acquisition unit 11 acquires the input document. Then, the dividing unit 15 divides the document acquired by the document acquiring unit 11 into a sentence "The operation is extremely slow and the fan does not make a sound." And a sentence "It may not stand up."

  The binary classification estimation unit 12 estimates the binary classification of each sentence divided by the division unit 15 based on the learning data by a known machine estimation method. As described above, when the division unit 15 divides the sentence into “the operation is very slow and the fan does not make a sound” and “there is no possibility of standing up”, the binary classification estimation unit 12 It is estimated whether the sentences divided by the dividing unit 15 respectively belong to “power supply”, “HDD”, “display”, and “fan”.

  Specifically, as shown in FIG. 7, based on the learning data, the binary classification estimating unit 12 determines that the sentence “Power is very slow and the fan does not make a sound”. It is estimated that it does not belong to the classification (x), it is estimated that it does not belong to the classification of "HDD" (x), it is estimated that it does not belong to the classification of "display" (x), and it belongs to the classification of "fan" (o ).

  In addition, the binary classification estimating unit 12 estimates that the sentence “may not start up” belongs to the classification “power” based on the learning data, and classifies the sentence as “HDD”. It is assumed that it is estimated that it does not belong to the category “display”, that it does not belong to the category “display”, and that it does not belong to the category “fan” (×).

  Further, the binary classification estimating unit 12 classifies the entire document “The operation is very slow, the fan does not make a sound, and sometimes does not start up” based on the learning data, and the document belongs to the classification “Power”. (○), presumed that it does not belong to the classification of “HDD” (×), presumed that it does not belong to the classification of “display” (x), and presumed to belong to the classification of “fan” (○) And

  The storage unit 13 stores the result information. As shown in FIG. 7, the result information includes an ID, a document, each of the divided sentences, and a classification result estimated by the binary classification estimation unit 12 and indicating whether each sentence belongs to each classification. , And a classification result indicating whether the entire document belongs to each classification.

  The display unit 14 displays the result information stored in the storage unit 13. Specifically, the display unit 14 displays the document, each sentence, and the classification result on a display or the like. For example, as illustrated in FIG. 6B, the display unit 14 classifies the classification result estimated by the binary classification estimation unit 12 into the “fan” and “power” classifications of the entire document, and The sentence "Very slow, no fan sound" belongs to the category of "fans", and the sentence "may not start up" indicates that it belongs to the category of "power". The display unit 14 may transmit the result information to another information processing device via a communication network.

  Next, a visualization method of the visualization device 2 according to the second embodiment will be described with reference to a sequence diagram shown in FIG.

  First, the document acquisition unit 11 acquires a document based on a user operation (step S21). For example, the document acquisition unit 11 uses the Web application to display, on the screen shown in FIG. 6A, “the operation is very slow, there is no fan sound, and the user may not start up”. Is input and when the send button is pressed, the input document is acquired.

  When the document is obtained in step S21, the dividing unit 15 divides the document (step S22). For example, the dividing unit 15 describes the text "The operation is very slow and the fan does not sound. There is a case where the fan does not start." Sometimes it is. "

  When the document is divided in step S22, the binary classification estimating unit 12 estimates the binary classification of the divided sentence (step S23). Specifically, the binary classification estimating unit 12 estimates whether or not the divided sentence belongs to a predetermined classification. In the present embodiment, for example, the binary classification estimating unit 12 estimates whether or not the sentence “The operation is very slow and the fan does not sound” included in the document belongs to the classification of “power supply”.

  When the binary classification is estimated in step S23, the binary classification estimation unit 12 determines whether or not all the classifications have been estimated (step S24). In the example of the present embodiment, the binary classification estimating unit 12 determines whether or not the document belongs to each of the classifications “power”, “HDD”, “display”, and “fan”.

  If the sentence has not been estimated for all the classes in step S24, the process returns to step S23, and the binary classification estimating unit 12 estimates whether the sentence belongs to a class that has not been estimated yet. When sentences have been estimated for all the classes in step S24, the binary classification estimating unit 12 determines whether or not all the divided sentences have been estimated (step S25). In the present embodiment, it is determined whether or not any of the divided sentences “the operation is extremely slow and the fan does not make a sound” and the sentence “it may not start up” are both estimated.

  If all sentences have been determined in step S25, the storage unit 13 stores result information including the classification result (step S26). If all sentences have not been determined in step S25, the process returns to step S23 to determine whether a sentence that has not been estimated belongs to each category.

  When the result information is stored in step S26, the display unit 14 displays the result information on a display or the like (step S27). For example, as shown in FIG. 6B, the display unit 14 classifies the classification result estimated by the binary classification estimating unit 12 into the “fan” and “power” classifications of the entire document, and Is very slow, and the fan does not sound. "The sentence" Fan "belongs to the category, and" Sometimes it does not start up. "

  As described above, according to the second embodiment, the dividing unit 15 divides a document into a plurality of sentences. Then, the binary classification estimating unit 12 estimates whether or not each of the plurality of sentences corresponds to each of the classifications. Therefore, if the input document includes a plurality of sentences, the classification result is stored and displayed for each sentence. This makes it possible to reconsider the classification by referring to the classification result estimated for each sentence, even when the classification of the entire document cannot be determined.

[Third Embodiment]
Next, a third embodiment of the present invention will be described with reference to the drawings.

  As shown in FIG. 1, the visualization device 1 according to the third embodiment includes a document acquisition unit 11, a binary classification estimation unit 12, a storage unit 13, and a display unit 14, as in the first embodiment. . Note that the configuration blocks are the same as those in the first embodiment, and a description of the configuration blocks will be omitted.

  In the third embodiment, the document acquisition unit 11 acquires a document and acquires a classification (manually assigned classification) corresponding to the document based on a user operation. As an example, using a Web application, the user inputs a document “The startup is poor” on a screen as shown in FIG. 9A, and inputs “HDD” as a failure classification as a manually assigned classification. . Then, the user presses the transmission button. In this case, the document acquisition unit 11 acquires the document “The startup is bad.” And acquires the failure classification “HDD” as the manually assigned classification.

  The binary classification estimating unit 12 estimates whether or not the acquired document belongs to each classification by using a known machine estimation method based on the learning data. For example, based on the learning data, the binary classification estimating unit 12 estimates, based on the learning data, that the document does not belong to the classification of “power” (×) based on the learning data, It is assumed that it is estimated that it belongs to the category of “HDD” (（), that it does not belong to the category of “display” (×), and that it does not belong to the category of “fan” (×).

  The storage unit 13 stores the result information. As shown in FIG. 10, the result information includes an ID, a document, a classification result of the machine estimation estimated by the binary classification estimation unit 12, and a manual assignment classification. For example, as shown in FIG. 10, the storage unit 13 stores a manually assigned classification “HDD” corresponding to a document “bad start” of ID “C”. Further, the storage unit 13 stores the power supply “×”, the HDD “○”, the display “×”, and the fan “×” as the classification result of the machine estimation.

  The display unit 14 displays the result information stored in the storage unit 13. More specifically, the display unit 14 displays the classification result of the machine estimation estimated by the binary classification estimation unit 12 and the manual assignment classification on a display or the like. For example, as shown in FIG. 9B, the display unit 14 displays the classification result estimated by the binary classification estimation unit 12 as indicating that the document belongs to the classification “HDD” and the input failure classification (manual). This indicates that the “assignment classification” is “HDD”. The display unit 14 may transmit the result information to another information processing device via a communication network.

  Subsequently, a visualization method of the visualization device 3 according to the third embodiment will be described with reference to a sequence diagram shown in FIG.

  First, the document acquisition unit 11 acquires a document and a hand-carrying classification based on a user operation (step S31). For example, as shown in FIG. 9A, when the user inputs a document “Wrong start-up” and a manual assignment classification “HDD” as shown in FIG. 9A, the input document and the manual assignment classification are input. To get.

  The processes in steps S32 and S33 of the visualization method performed by the visualization device 3 in the present embodiment are the same as the processes in steps S12 and S13 in the first embodiment.

  If the documents are estimated for all the categories in step S33, the storage unit 13 stores the result information (step S34). The result information according to the third embodiment includes a manually assigned classification in addition to the classification result according to the first embodiment. For example, as shown in FIG. 10, the storage unit 13 stores result information including a classification result of a power supply “×”, an HDD “「 ”, a display“ × ”, a fan“ × ”, and a manual classification of“ HDD ”. Is stored.

  When the result information is stored in step S34, the display unit 14 displays the result information on a display or the like (step S35). For example, as illustrated in FIG. 9B, the display unit 14 determines that the possibility of belonging to the classification “HDD” is high in the determination by the machine learning, and the input failure classification (manually assigned classification). Is "HDD".

  As described above, according to the third embodiment, the document acquisition unit 11 further acquires the manual assignment classification of the document based on the operation of the user. Then, the display unit 14 displays the classification result estimated by the binary classification estimation unit 12 and the manually assigned classification as result information. Therefore, the user can recognize the classification result estimated based on the learning data, and finally decide the classification of the document in consideration of the manual assignment classification indicating the judgment of the user who acquired the document. it can.

[Fourth embodiment]
Next, a fourth embodiment of the present invention will be described with reference to the drawings.

  As shown in FIG. 5, the visualization device 2 according to the fourth embodiment includes a document acquisition unit 11, a binary classification estimation unit 12, a storage unit 13, a display unit 14, and a division unit, as in the second embodiment. A unit 15 is provided. Note that the configuration blocks are the same as in the second embodiment, and a description of the configuration blocks will be omitted.

  In the fourth embodiment, the document acquisition unit 11 acquires a document based on a user operation, and acquires a manual assignment classification corresponding to the document. As an example, the user uses a Web application to input a document such as “The startup is poor. The same is true even when the power is turned off and started.” On the screen shown in FIG. "HDD" is input in the failure classification of "1. Then, the user presses the transmission button. In this case, the document obtaining unit 11 obtains a document “The startup is poor. The same is true even if the power is turned off and restarted.” Further, “HDD” is acquired as a manual assignment classification.

  Then, the dividing unit 15 divides the document acquired by the document acquiring unit 11 into a sentence “Poor start-up.” And a sentence “Even if the power is turned off and started again.”

  Further, the binary classification estimating unit 12 estimates whether or not each sentence divided by the dividing unit 15 belongs to each classification by using a known machine estimation method based on the learning data. For example, as shown in FIG. 13, the binary classification estimating unit 12 estimates that the sentence “bad startup” based on the learning data does not belong to the classification of “power” (×), and It is assumed that it is estimated that it belongs to the category of “HDD” (（), that it does not belong to the category of “display” (×), and that it does not belong to the category of “fan” (×).

  In addition, the binary classification estimating unit 12 estimates that the sentence “The same applies even if the power is turned off and restarted” is based on the learning data, and that the sentence does not belong to the classification “power” (×). It is assumed that it is estimated that it does not belong to “HDD”, that it does not belong to “display” (×), and that it does not belong to “fan” (×). Further, the binary classification estimating unit 12 does not belong to the classification of “power supply” based on the learning data for the entire document “The startup is bad. The same applies even if the power is turned off and restarted.” ), It is estimated that it belongs to “HDD” (○), it is estimated that it does not belong to “display” (×), and it is estimated that it does not belong to “fan” (×).

  The storage unit 13 stores the result information. As shown in FIG. 13, the result information includes an ID, a document, each divided sentence, and a classification result estimated by the binary classification estimation unit 12 indicating whether each sentence belongs to each classification, And a classification result indicating whether or not the entire document belongs to each classification, and a manual assignment classification.

  As described above, the document acquisition unit 11 acquires the document “The startup is poor. The same applies even if the power is turned off and restarted.” And the manually assigned classification “HDD”. The result information stored in the storage unit 13 when it is estimated will be described. As shown in FIG. 13, the storage unit 13 responds to the document “ID has a poor start-up. The power supply “×”, the HDD “○”, the display “×”, and the fan “×” which are the classification results based on the machine estimation are stored. In addition, the storage unit 13 stores the power supply “×”, the HDD “」 ”, the display“ × ”, and the fan“ × ”in response to the divided sentence“ The startup is bad. ” Further, the storage unit 13 stores the power supply “×”, the HDD “×”, the display “×”, and the fan “×” in response to the divided statement “the same applies even if the power is turned off and restarted”. Store.

  The display unit 14 displays the result information stored in the storage unit 13. Specifically, the display unit 14 displays a document, each sentence, a classification result indicating whether each sentence belongs to each classification, a classification result indicating whether a document including each sentence belongs to each classification, and The assignment classification is displayed on a display or the like.

  For example, as illustrated in FIG. 12B, the display unit 14 determines that the entire document is determined to be highly likely to belong to the classification “HDD” by the determination based on the machine learning, and that the input failure classification (manually This indicates that the “assignment classification” is “HDD”. Further, the display unit 14 displays that the sentence “the start is poor” is highly likely to belong to the classification “HDD”, and the sentence “the same applies even when the power is turned off” is displayed. Displays that it is determined that there is a high possibility that it does not belong to the classification. The display unit 14 may transmit the result information to another information processing device via a communication network.

  Subsequently, a visualization method of the visualization device 4 according to the fourth embodiment will be described with reference to a sequence diagram shown in FIG.

  First, the document obtaining unit 11 obtains a document and a manual assignment classification based on a user operation (step S41). For example, as shown in FIG. 12A, a document “The startup is poor. The same is true even after the power is turned off.” And a manually assigned classification “HDD” are input by a user operation. In this case, the input document and the manually assigned classification are acquired.

  The processing of steps S42 to S45 of the visualization method performed by the visualization device 3 in the present embodiment is the same as the processing of steps S22 to S25 of the second embodiment.

  If all sentences have been estimated in step S45, the storage unit 13 stores the result information (step S46). The result information according to the fourth embodiment includes a manual assignment classification in addition to the classification result according to the second embodiment. For example, as shown in FIG. 13, the storage unit 13 is a manual assignment classification corresponding to the document of ID “D” “The start is poor. The same applies even if the power is turned off and restarted.” HDD, power “x”, HDD “あ る”, display “x”, and fan “x”, which are classification results based on machine estimation, are stored. In addition, the storage unit 13 stores the power supply “×”, the HDD “O”, the display “×”, and the fan “×”, which are the classification results based on the machine estimation, in response to the divided sentence “The rising is bad.” Store. Further, the storage unit 13 responds to the divided sentence “Even if the power is turned off and restarted.”, The power supply “x”, the HDD “x”, and the display “x” which are the classification results based on machine estimation. ”And fan“ x ”.

  When the result information is stored in step S34, the display unit 14 displays the result information on a display or the like (step S47). For example, as illustrated in FIG. 12B, the display unit 14 determines that the entire document is determined to be highly likely to belong to the classification “HDD” by the determination based on the machine learning, and that the input failure classification ( This indicates that “manual assignment classification” is “HDD”. Further, the display unit 14 displays that the sentence “the start is poor” is highly likely to belong to the classification “HDD”, and the sentence “the same applies even when the power is turned off” is displayed. Displays that it is determined that there is a high possibility that it does not belong to the classification.

  As described above, according to the fourth embodiment, similarly to the second embodiment, the dividing unit 15 divides a document into a plurality of sentences. Then, the binary classification estimating unit 12 estimates whether or not each of the plurality of sentences corresponds to each of the classifications. Therefore, the user can recognize whether each sentence included in the document corresponds to each classification. Further, according to the fourth embodiment, similarly to the third embodiment, the document acquisition unit 11 further acquires the manual assignment classification of the document based on the operation of the user. Then, the display unit 14 displays the classification estimated by the binary classification estimation unit 12 and the manually assigned classification as result information. Therefore, the user can recognize the classification estimated based on the learning data and the manual assignment classification indicating the judgment of the user who acquired the document. Therefore, when the user finally decides the classification to which the document belongs, it is possible to make a comprehensive decision in consideration of both the classification result for each sentence and the manually assigned classification.

[Fifth Embodiment]
Next, a fifth embodiment of the present invention will be described with reference to the drawings.

  As shown in FIG. 5, the visualization device 2 according to the fifth embodiment includes a document acquisition unit 11, a binary classification estimation unit 12, a storage unit 13, a display unit 14, and a division unit, as in the fourth embodiment. A unit 15 is provided. Note that the configuration blocks are the same as those in the fourth embodiment, and a description of the configuration blocks will be omitted.

  In the fifth embodiment, the document acquisition unit 11 acquires a plurality of documents and acquires a manual assignment classification corresponding to each document based on a user operation. For example, the document acquisition unit 11 acquires a document “Nothing is displayed on the screen even when the power is turned on” and a manual assignment classification “Display”. In addition, the document acquisition unit 11 acquires a document “The operation is extremely slow and the fan does not make a sound. There is a case where the fan does not start.” And a manual assignment classification “Fan”. In addition, the document acquisition unit 11 acquires a document “The startup is poor. The same applies even if the power is turned off and restarted.” And a manual assignment classification “HDD”.

  In this case, the binary classification estimating unit 12 and the storage unit 13 perform the same processing as that of the third embodiment on the document “Nothing is displayed on the screen even when the power is turned on”. The dividing unit 15, the binary classification estimating unit 12, and the storage unit 13 have a statement that "the operation is very slow and the fan does not make a sound. There is a case where the fan does not start." The same processing as in the fourth embodiment is performed for the document "Same as above when starting up again."

  The display unit 14 displays the result information about each document stored in the storage unit 13 on a display or the like. The display unit 14 may transmit the result information of each document to another information processing device via a communication network together.

  The visualization method of the visualization device 2 according to the fifth embodiment is a combination of the visualization method of the third embodiment and the visualization method of the fourth embodiment. Specifically, the document acquisition unit 11 acquires a plurality of documents and a classification corresponding to each document. When the document includes only one sentence, the visualization device 5 performs the processing of steps S32 to S34 of the third embodiment. When the document includes a plurality of sentences, the visualization device 2 performs the processing of steps S42 to S46 of the fourth embodiment. Then, as shown in FIG. 15, the storage unit 13 stores the result information stored in step S34 and the result information stored in step S46 together.

  As described above, according to the fifth embodiment, the document acquisition unit 11 acquires a plurality of documents and a manual assignment classification corresponding to each document. Therefore, the user can recognize the classification estimated based on the learning data and the manually assigned classification for each of the plurality of documents. Therefore, when the user finally decides the classification to which a plurality of documents belong, it is possible to make a comprehensive decision in view of the classification results of the documents and the sentences included in each document and the manual assignment classification.

  Further, in the above-described embodiment, each function of the visualization devices 1, 2, 3, 4, and 5 stores a program describing processing contents for realizing these functions in a storage unit of the computer, and stores the program in the storage unit of the computer. It may be realized by reading and executing this program by the CPU. Further, the program may be recorded on a recording medium or provided through a network.

  Further, in the first embodiment, as shown in FIG. 16, the visualization device 1 may further include a voice recognition unit 16. The voice recognition unit 16 performs voice recognition processing on the voice uttered by the user and converts the voice into a document. Then, the document acquisition unit 11 acquires the document converted by the speech recognition unit 16, and performs the same processing as described above. This makes it possible to estimate the classification of information represented by speech. The same applies to the second to fifth embodiments.

  Although the above embodiments have been described as representative examples, it will be apparent to those skilled in the art that many changes and substitutions can be made within the spirit and scope of the present invention. Therefore, the present invention should not be construed as being limited by the above-described embodiment, and various modifications and changes can be made without departing from the scope of the claims.

1,2,3,4,5 Visualization device 11 Document acquisition unit 12 Binary classification estimation unit 13 Storage unit 14 Display unit 15 Division unit 16 Voice recognition unit

Claims (5)

A document acquisition unit for acquiring a document based on a user operation;
A dividing unit that divides the document into a plurality of sentences;
A binary classification estimating unit that estimates whether the document and the plurality of sentences each belong to each of a plurality of predetermined classifications,
Before SL document and the plurality of sentences, and a display unit in which the document and each of the plurality of statements to display the result information and a classification result indicating whether belonging to each of the classification,
A visualization device comprising: A document acquisition unit configured to acquire a document and a manual assignment classification for the document based on a user operation;
A dividing unit that divides the document into a plurality of sentences;
A binary classification estimating unit that estimates whether the document and the plurality of sentences each belong to each of a plurality of predetermined classifications,
The document and the plurality of sentences, the classification result estimated by the binary classification estimation unit, and the result information including the classification result by the manual assignment classification, and the classification result included in the result information is the estimated classification A display unit that displays the result or information indicating whether the result is a classification result according to the manual assignment classification,
A visualization device comprising: A visualization method performed by a visualization device that estimates a classification of a document,
Obtaining a document based on a user operation;
Dividing the document into a plurality of sentences;
Estimating whether or not each of the document and the plurality of sentences belongs to each of a plurality of predetermined classifications,
Displaying result information including the document and the plurality of sentences, and a classification result indicating whether each of the document and the plurality of sentences belongs to each of the classifications,
A visualization method comprising: Based on the operation of the user, a document, and a step of acquiring a classification result by manual assignment classification for the document,
Dividing the document into a plurality of sentences;
Estimating whether or not each of the document and the plurality of sentences belongs to each of a plurality of predetermined classifications,
The document and the plurality of sentences, the estimated classification result, and the result information including the classification result by the manual assignment classification, and whether the classification result included in the result information is the estimated classification result, Displaying result information including information indicating whether the result is a classification result by the assigned classification;
A visualization method comprising:   A program for causing a computer to function as the visualization device according to claim 1.

Images