JP6605410B2 - Emotion factor estimation support device, emotion factor estimation support method, and emotion factor estimation support program - Google Patents

Description

  The present invention relates to an emotion factor estimation support apparatus, an emotion factor estimation support method, and an emotion factor estimation support program for estimating a factor that a poster has an emotion appearing in a sentence created by the poster.

  2. Description of the Related Art Conventionally, a technique for identifying a sentence creator's emotion by analyzing a sentence is known. In Non-Patent Document 1, an emotion-causing factor corpus that collects character strings representing factors that caused emotions from a large number of sentences published on the Internet is generated, and the generated emotion-causing factor corpus is referred to as a sentence. There is disclosed a technique for identifying an emotion that the creator had when creating a sentence.

Ryoko Tokuhisa, Kentaro Inui, Yuji Matsumoto, “Emotion estimation based on the emotion-causing factor corpus acquired from the Web”, IPSJ Transactions, Vol. 50, No. 4, 1365-1374, April 2009

  The relationship between a person's feelings and the factors that caused them varies from person to person. For example, young people tend to be happy to have a birthday, while older people tend not to be happy to have a birthday. For example, suppose that an elderly Mr. A posted a message to SNS (Social Networking Service) saying, “Today is a birthday. When Mr. A was not happy that he had a birthday and was happy that his grandson came, Mr. A responded to the message from Mr. A, “It was nice to meet your grandson. If the message “How many?” Is returned, communication with Mr. A will be activated. On the other hand, if the message “Happy Birthday!” Is returned to the message from Mr. A, it is difficult to activate communication with Mr. A.

  As described above, when exchanging messages on the SNS, it may not be possible to know what the person who posted the message is, so there is a possibility of sending an inappropriate reply message. Therefore, it is required to improve the estimation accuracy of the factor that caused the emotion of the message poster so that an appropriate reply message can be created. However, as in the past, when trying to identify emotions using the relationship between the emotions identified using a large number of texts published on the Internet and the character strings that cause emotions, the factors that caused the emotions There was a problem that the probability of not being able to be estimated correctly is high.

  Therefore, the present invention has been made in view of these points, and an object of the present invention is to improve the accuracy of estimating factors that have caused the feelings of contributors appearing in posts.

  The emotion factor estimation support device according to the first aspect of the present invention includes, for each attribute of a message contributor, an emotion word that represents a contributor's emotion and an emotion factor word that represents a factor that caused the emotion represented by the emotion word. A storage unit that stores an emotion factor word dictionary associated with the message, an acquisition unit that acquires a posted message created by the poster, an attribute estimation unit that estimates an attribute of the poster who created the posted message, and the posted message And an emotion factor word associated with the emotion word extracted by the extractor in the emotion factor word dictionary corresponding to the attribute of the poster estimated by the attribute estimation unit And a selection unit that selects a factor character string similar to that from the posted message.

  The selection unit includes at least one of the emotion factor words associated with the emotion word extracted by the extraction unit in the emotion factor word dictionary corresponding to the attribute of the poster estimated by the attribute estimation unit in the posted message. The factor character string including a part of the character string may be selected from the posted message.

  The selection unit may determine a range for selecting the factor character string based on a position of the emotion word extracted by the extraction unit in the posted message.

  The storage unit, in association with the emotion factor word, further stores probability information indicating the probability that the factor that the poster has the emotion represented by the emotion word is the emotion factor word, and the selection unit includes: When the storage unit stores a plurality of the emotion factor words corresponding to the emotion words extracted by the extraction unit, the factor character string related to the emotion factor words having a higher probability indicated by the probability information You may select with priority.

  The acquisition unit acquires a plurality of the posted messages, and the selection unit includes the factor character string included in the posted message created before or after the posted message from which the extraction unit has extracted the emotion word. You may choose.

  If the selection unit determines that the factor character string is not included in the post message from which the extraction unit has extracted the emotion word, the selection unit is prior to the post message from which the extraction unit has extracted the emotion word. Alternatively, the factor character string included in a post message created later may be selected.

  An output unit that outputs the emotion word extracted by the extraction unit and the factor character string selected by the selection unit in association with the posted message acquired by the acquisition unit may be further included.

  When the selection unit cannot select the emotion factor word associated with the emotion word, the selection unit further includes a reception unit that receives an input of the emotion factor word to be registered in the emotion factor word dictionary in association with the emotion word Also good.

  The acquisition unit further acquires another message created by another person different from the poster related to the posted message created by the poster, and the selection unit includes the factor character included in the other person message. A column may be selected.

  The emotion factor estimation support method according to the second aspect of the present invention includes a step executed by a computer to acquire a posted message created by a message poster, and a step of estimating an attribute of the poster who created the posted message. , An emotion factor word dictionary in which an emotion word included in the posted message is extracted, and an emotion word representing an emotion of the poster and an emotion factor word representing a factor causing the emotion represented by the emotion word are associated with each other A factor character string associated with the emotion factor word associated with the emotion word extracted in the extracting step by referring to an emotion factor word dictionary corresponding to the estimated attribute of the poster, Selecting from posted messages.

  The emotion factor estimation support program according to the third aspect of the present invention includes a step of acquiring a posted message created by a message poster, a step of estimating an attribute of a poster who created the posted message, and the posted message. Is an emotion factor word dictionary in which an emotion word included in a word is extracted, an emotion word representing a poster's emotion, and an emotion factor word representing a factor causing the emotion represented by the emotion word are associated with each other. A factor character string related to the emotion factor word associated with the emotion word extracted in the extracting step is selected from the posted message by referring to the emotion factor word dictionary corresponding to the attribute of the poster who has performed And causing the computer to execute.

  According to the present invention, there is an effect that it is possible to improve the accuracy of estimating the factor causing the poster's emotion appearing in the posting.

It is a figure which shows the structure of the emotion factor estimation assistance system S which concerns on this embodiment. It is a figure which shows the structure of the emotion factor estimation assistance apparatus. It is a table | surface which shows an example of an emotion word dictionary. It is a table | surface which shows an example of an emotion factor word dictionary. It is a figure which shows the example of the screen which the output part 145 output the factor character string. It is a figure which shows the example of the screen which the output part 145 output the factor character string. It is a figure which shows the example of the screen which the output part 145 output the factor character string. It is a figure which shows the modification of an emotion factor word dictionary.

[Overview of this embodiment]
FIG. 1 is a diagram illustrating a configuration of an emotion factor estimation support system S according to the present embodiment. The emotion factor estimation support system S includes an emotion factor estimation support device 1, a poster terminal 2, and an SNS server 3. The emotion factor estimation support device 1 is a computer that acquires a message posted by a poster of the poster terminal 2 (hereinafter referred to as a posted message), and identifies the poster's emotion based on the acquired posted message. The emotion factor estimation support device 1 estimates a factor that the poster has the identified emotion and outputs an estimation result. The posted message is, for example, a message posted on the SNS, and in the present embodiment, a description will be given by taking a tweet as an example.

  The poster terminal 2 is, for example, a smartphone, and is an information processing terminal that allows a poster to create and transmit a tweet. The poster terminal 2 transmits the tweet created by the poster to the SNS server 3 via the network N.

  The SNS server 3 receives tweets transmitted from the contributor terminals 2 of a large number of contributors, and stores them in a storage medium in association with a contributor ID unique to the contributor. The SNS server 3 transmits tweets in response to requests from the emotion factor estimation support apparatus 1 and the poster terminal 2. For example, when the SNS server 3 receives the name of a contributor who receives a tweet input from the emotion factor estimation support apparatus 1, the SNS server 3 stores the past one stored in association with the contributor ID corresponding to the contributor name. The above tweets are searched, and the extracted tweets are transmitted to the emotion factor estimation support apparatus 1.

  When the emotion factor estimation support apparatus 1 receives a tweet, it extracts a word or a character string (hereinafter referred to as an emotion word) indicating the emotion included in the tweet. Moreover, the emotion factor estimation support apparatus 1 estimates the attribute of the poster who created the tweet. The emotion factor estimation support apparatus 1 estimates the attribute of a contributor based on, for example, the style of a tweet or a word or character string included in the tweet. The emotion factor estimation support device 1 may specify the attribute of the poster based on the attribute information of the poster sent from the SNS server 3 together with the tweet.

  Subsequently, the emotion factor estimation support apparatus 1 uses a word or a character string (hereinafter referred to as emotion) that causes a poster having an estimated attribute in the tweet created by the poster to have an emotion expressed by the extracted emotion word. Search for the cause string). The emotion factor estimation support apparatus 1 can select an emotion factor character string suitable for the attribute of the poster by referring to the emotion factor word dictionary in which the attribute of the poster and the emotion factor word are associated with each other. The emotion factor estimation support apparatus 1 outputs an emotion factor character string by displaying the emotion factor character string specified by the search or transmitting it to another computer.

Thus, in the emotion factor estimation support system S, it is possible to estimate an emotion factor character string suitable for the attribute of the poster in a tweet created by the poster and output the estimated emotion factor character string. As a result, the user of the emotion factor estimation support system S can appropriately grasp the factors that have the emotion appearing in the tweet by the poster who created the tweet. It is possible to make an optimal response.
Hereinafter, the configuration and operation of the emotion factor estimation support apparatus 1 will be described in detail.

[Configuration of Emotion Factor Estimation Support Device 1]
FIG. 2 is a diagram illustrating a configuration of the emotion factor estimation support apparatus 1. The emotion factor estimation support apparatus 1 includes a communication unit 11, an operation reception unit 12, a storage unit 13, and a control unit 14.
The communication unit 11 is a communication interface for transmitting and receiving information to and from the poster terminal 2 and the SNS server 3 via the network N. The communication unit 11 includes, for example, a LAN (Local Area Network) controller. When the communication unit 11 receives a tweet as a posted message from the poster terminal 2 or the SNS server 3, the communication unit 11 inputs the received tweet to the control unit 14.

  The operation reception unit 12 is a user interface that receives an operation by the user of the emotion factor estimation support apparatus 1. The operation reception unit 12 is configured by, for example, a keyboard, a mouse, and a display. For example, the user of the emotion factor estimation support apparatus 1 can input a contributor name for which an emotion factor word is estimated.

  The storage unit 13 includes a storage medium such as a read only memory (ROM), a random access memory (RAM), and a hard disk. The storage unit 13 stores, for example, an emotion factor estimation support program that is executed by the control unit 14 to execute the emotion factor estimation support method. The storage unit 13 stores an emotion word dictionary 131 and an emotion factor word dictionary 132.

  The emotion word dictionary 131 is a dictionary used by the control unit 14 to extract emotion words included in the tweet. In the emotion word dictionary 131, emotion type information indicating an emotion type is associated with an emotion word. FIG. 3 is a table showing an example of the emotion word dictionary stored in the storage unit 13. As shown in FIG. 3, the emotion word may be a pictograph instead of a character string.

  The emotion factor word dictionary 132 is an emotion factor word dictionary in which for each attribute of the poster, an emotion word representing the poster's emotion is associated with an emotion factor word representing a factor causing the emotion represented by the emotion word. . FIG. 4 is a table showing an example of the emotion factor word dictionary. In the emotion factor word dictionary, the emotion type, the emotion word, the emotion factor word, and the attributes of the poster are associated with each other. The attribute of the contributor is an arbitrary element that characterizes the contributor, such as a residence area, hometown, school, favorite food, in addition to the age, sex, hobby, and temperament shown in FIG.

  In the example shown in FIG. 4, when the emotion type is “恍惚” and the emotion word is “rich”, an example of the emotion factor word when the poster attribute is male is “work is well”. An example of the emotion factor word when the poster attribute is female is “good shopping”. The reason why men feel that they are “full” is that the work is often smooth, but as a factor that women feel “full”, good shopping was possible. It shows that there are many cases.

  Also, in FIG. 4, when the emotion type is “恍惚” and the emotion word is “happy”, an example of the emotion factor word when the poster attribute is a woman in her twenties is “Today's birthday Is. When the emotion type is “grief” and the emotion word is “spicy”, an example of the emotion factor word when the poster attribute is a woman in their 40s is also “birthday”. This indicates that the emotion factor word “birthday” can be a factor that has different emotions depending on the poster attribute. As described above, in the emotion factor word dictionary 132, since the emotion word and the emotion factor word are associated with each poster attribute, the emotion factor estimation support apparatus 1 displays the emotion appearing in the posted message. It is possible to estimate with high accuracy the factors that will be held.

Returning to FIG. 2, the configuration of the control unit 14 will be described.
The control unit 14 is, for example, a CPU (Central Processing Unit), and executes a program stored in the storage unit 13 to obtain an acquisition unit 141, an attribute estimation unit 142, an extraction unit 143, a selection unit 144, and an output unit 145. Function.

  The acquisition unit 141 acquires a posted message created by the poster. In the present embodiment, the acquiring unit 141 transmits the poster name input in the operation accepting unit 12 to the SNS server 3, and posts a tweet created by the poster corresponding to the poster name transmitted to the SNS server 3. Get as. The acquisition unit 141 inputs the acquired tweet to the attribute estimation unit 142.

  The attribute estimation unit 142 estimates the attribute of the poster who created the posted message. In the present embodiment, the attribute estimation unit 142 estimates the attribute of the poster based on the tweet style input from the acquisition unit 141 and the words included in the tweet.

  For example, when the tweet ends with a masculine expression, the attribute estimation unit 142 estimates that the attribute of the poster is male, and when the suite sentence ends with a feminine expression, Presume that the attribute is female. Moreover, when the place name is included in the tweet, the attribute estimation unit 142 can estimate the poster's residence area, hometown, and the like by combining with other words in the tweet. The attribute estimation unit 142 inputs a tweet and the attribute of the poster who has been estimated to the extraction unit 143.

  The extraction unit 143 extracts emotion words included in the posted message. In the present embodiment, the extraction unit 143 extracts emotion words included in the tweet received from the attribute estimation unit 142 by performing morphological analysis with reference to the emotion word dictionary shown in FIG. When the emotion word is recorded in the emotion word dictionary in association with the poster attribute, the extraction unit 143 extracts the emotion word corresponding to the poster attribute input from the attribute estimation unit 142 from the tweet. The extraction unit 143 notifies the selection unit 144 of the tweet, the attribute of the poster, and the extracted emotion word.

  The selection unit 144 extracts a factor character string similar to the emotion factor word associated with the emotion word extracted by the extraction unit 143 in the emotion factor word dictionary corresponding to the attribute of the poster estimated by the attribute estimation unit 142 from the posted message. select. The emotion factor word dictionary corresponding to the attribute of the poster may be an emotion factor word dictionary provided independently for each attribute of the poster, and as shown in FIG. 4, emotions corresponding to a plurality of attributes. Of the factor word dictionary, the part corresponding to the attribute of the poster may be used.

  First, the selection unit 144 extracts an emotion factor word corresponding to the attribute of the poster input from the extraction unit 143 in the emotion factor word dictionary shown in FIG. Subsequently, the selection unit 144 searches for a factor character string that is a character string similar to the extracted emotion factor word in the tweet. The factor character string is, for example, a character string that includes at least a part of the emotion factor word, and is a character string that is the same as the emotion factor word or a character string that is highly correlated with the emotion factor word. When a character string that is more similar to the emotion factor word than a predetermined degree is found, the selection unit 144 selects the found character string as the factor character string.

  Here, the selection unit 144 may determine a target range for selecting the factor character string based on the position of the emotion word extracted by the extraction unit 143 in the tweet. For example, when the emotion word is within a predetermined range from the end of the posted message, the selection unit 144 searches for the factor character string from the character string before the emotion word in the posted message. For example, when the emotion word is within a predetermined range from the head of the tweet, the selection unit 144 searches for the factor character string from the character string after the emotion word in the tweet.

  The selection unit 144 determines whether to search for a factor character string from the character string before the emotion word or to search for the factor character string from the character string after the emotion word based on the content and style of the emotion word. Also good. For example, if the emotion word includes an expression that is likely to include an emotion factor word before the emotion word, such as “I am happy”, the selection unit 144 uses the character string before the emotion word. Search for the cause string. When the selection unit 144 includes an expression that is likely to include an emotion factor word after the emotion word, such as “I am happy”, the selection unit 144 selects the factor character string from the character string after the emotion word. Explore.

  Moreover, the selection part 144 may select the factor character string contained in the tweet created before or after the tweet from which the extraction part 143 extracted the emotion word. For example, when the extraction unit 143 determines that the factor character string is not included in the tweet from which the emotional word is extracted, the selection unit 144 is created before or after the tweet from which the extraction unit 143 has extracted the emotional word. Select the factor string included in the tweet.

  The selection unit 144 searches for an emotion factor word in a tweet before the tweet from which the emotion word is extracted, or searches for an emotion factor word in a tweet after the tweet from which the emotion word is extracted based on the content of the emotion word You may decide whether to do it. For example, when the tweet before the tweet including the emotion word includes an expression that is likely to include the emotion factor word, such as “So happy,” the selection unit 144 may include the expression before the tweet including the emotion word. Search for the factor string in the tweet. In addition, the selection unit 144 extracts the emotion word when the tweet behind the tweet from which the emotion word is extracted includes an expression that is highly likely to include the emotion factor word, such as “I am happy”. Search for the factor string in the tweet after the tweet.

The output unit 145 outputs the emotion word extracted by the extraction unit 143 and the factor character string selected by the selection unit 144 in association with the tweet acquired by the acquisition unit 141.
5 to 6 are diagrams illustrating examples of screens on which the output unit 145 outputs the factor character string. FIG. 5 shows an example in which a factor character string is included in a tweet including an emotion word. The child A who created the tweet shown in FIG. 5A is a woman in her 20s. In this case, the selection unit 144 selects “birthday” as the emotion factor word corresponding to the emotion word representing the word “happy”. The output unit 145 indicates that “birthday” is a factor character string by underlining the “birthday” selected by the selection unit 144.

  The child B who created the tweet shown in FIG. 5B is a woman in her 40s. In this case, the selection unit 144 selects emotions corresponding to women in their twenties among “birthday” and “grandchild will come” which are emotion factor words corresponding to the emotional word representing the word “I am happy.” Instead of selecting the factor word “birthday”, “grandchild will come” corresponding to a woman in her 40s is selected. The output unit 145 indicates that “grandson will come” is a factor character string by underlining “grandson will come”. The manner in which the output unit 145 outputs the factor character string is arbitrary. In addition to underlining, the user of the emotion factor estimation support apparatus 1 can change the factor character string by changing the color or size. You may make it recognizable.

  FIG. 6 shows an example in which an emotion factor word is not included in a tweet including an emotion word. The tweet shown in FIG. 6 (a2) includes the emotion word “I did it!”, But this tweet does not include the emotion factor word. Therefore, the selection unit 144 selects a factor character string corresponding to the attribute of the A child included in the tweet of FIG. 6 (a1) posted before the tweet of FIG. 6 (a2), and the output unit 145 The character string selected by the selection unit 144 is underlined and output.

  Similarly, the emotion factor word is not included in the tweet shown in FIG. Therefore, the selection unit 144 selects a factor character string corresponding to the attribute of the B child included in the tweet of FIG. 6 (b1) posted before the tweet of FIG. 6 (b2), and the output unit 145 The character string selected by the selection unit 144 is underlined and output.

[Operation Flowchart of Emotion Factor Estimation Support Device 1]
FIG. 7 is a flowchart of an operation in which the emotion factor estimation support apparatus 1 estimates an emotion factor.
First, the acquisition unit 141 acquires a tweet of a poster who is a target for estimating an emotion factor from the SNS server 3 (S11). Subsequently, the attribute estimation unit 142 estimates the attribute of the poster based on the tweet acquired by the acquisition unit 141 (S12).

  Subsequently, the extraction unit 143 determines whether or not an emotion word is included in the tweet (S13). If the emotion word is not included in the tweet (NO in S143), the extraction unit 143 returns to Step S11 and proceeds to the processing of the next tweet. When the emotion word is included in the tweet (YES in S143), the extraction unit 143 specifies the position of the emotion word in the tweet (S14).

  Subsequently, the selection unit 144 searches for the attribute of the poster and the emotion factor word corresponding to the emotion word based on the position of the emotion word specified by the extraction unit 143 (S15). When the selection unit 144 determines that there is an emotion factor word (YES in S16), the output unit 145 displays the emotion factor word so that the user can recognize it (S17). If it is determined that there is no emotion factor word (NO in S16), the selection unit 144 searches for emotion factor words in a number of tweets within a predetermined range before and after (S18). In this case, the selection unit 144 may search for a tweet within a range determined for each contributor, for example. By doing in this way, since an emotion factor word can be searched in consideration of the habit of the author's tweet creation, the probability that an emotion factor word can be appropriately selected can be improved.

  When the selection unit 144 determines that there are emotion factor words in the preceding and following tweets (YES in S19), the output unit 145 displays the emotion factor words (S17). If the selection unit 144 determines in step S19 that there is no emotion factor word, the selection unit 144 returns to step S11 and proceeds to processing the next tweet.

  The control part 14 performs the process from step S11 to step S19 with respect to all the tweets used as the object which estimates an emotion factor. When the process is completed for all tweets (YES in S20), control unit 14 ends the emotion factor estimation process.

(Modification 1)
FIG. 8 is a diagram showing a modification of the emotion factor word dictionary. The emotion factor word dictionary shown in FIG. 8 represents an emotion word in association with an emotion factor word when there are a plurality of emotion factor words for the same emotion type and emotion word combination in the same poster attribute. Unlike the emotion factor word dictionary shown in FIG. 4, it is the same in other respects in that the factor that the contributor has emotions includes probability information indicating the probability of the emotion factor word.

  In this case, when the storage unit 13 stores a plurality of emotion factor words corresponding to the emotion word extracted by the extraction unit 143, the selection unit 144 relates to an emotion factor word having a higher probability indicated by the probability information. Select the factor string with priority. For example, when a woman in her twenties makes a tweet that says “I was happy to have a date because it is a birthday today,” “birthday” and “date” are emotion factor words for the emotion word “I was happy”. .

  In this case, the selection unit 144 selects “date” having a higher probability as an emotion factor word and “date” as a factor character string. Then, the output unit 145 outputs, for example, a tweet in which “date” is underlined. By doing in this way, even if the user of the emotion factor estimation support device 1 includes a plurality of factor character strings similar to a plurality of emotion factor words in the posted message, the poster becomes the emotion word. It is possible to grasp the factors that have the corresponding emotion with high accuracy.

  When a plurality of factor character strings similar to a plurality of emotion factor words are included in the tweet in this way, the output unit 145 displays so that the user of the emotion factor estimation support apparatus 1 can recognize the plurality of factor character strings. In addition, the probability corresponding to each factor character string may be displayed. By doing in this way, the user of the emotion factor estimation support apparatus 1 can comprehensively determine the factor that the poster has the emotion corresponding to the emotion word based on the displayed plurality of factor character strings and the probability of each. Can be estimated.

(Modification 2)
The emotion factor estimation support apparatus 1 may be configured to update the emotion factor word dictionary shown in FIGS. 4 and 8. For example, when the selection unit 144 cannot select an emotion factor word associated with an emotion word, the operation reception unit 12 receives and receives an input of an emotion factor word that is registered in the emotion factor word dictionary in association with the emotion word. The emotion factor words are registered in the emotion factor word dictionary 132.

  Specifically, when the selection unit 144 determines that there is no emotion factor word in the tweet that is the target of estimating the emotion factor, the output unit 145 is posted with the tweet including the emotion word and before and after the tweet. A predetermined number of tweets are displayed. When the user of the emotion factor estimation support apparatus 1 performs an operation of selecting a character string included in a predetermined number of displayed tweets, the operation reception unit 12 receives the selected character string as an emotion factor word. . By doing in this way, the emotion factor estimation assistance apparatus 1 can increase the number of emotion factor words suitable for the attribute of the poster included in the emotion factor word dictionary.

(Modification 4)
In the above description, an example has been shown in which the acquisition unit 141 acquires a tweet created by a poster as a posted message. However, the acquisition unit 141 is different from the poster related to the posted message created by the poster. Another person's post message (hereinafter referred to as “other message”) may be acquired. For example, the acquisition unit 141 acquires another message such as a reply tweet related to a posted message created by the poster.

  Then, the selection unit 144 may select an emotion factor word corresponding to the attribute of the poster included in the other person message acquired by the acquisition unit 141. By doing in this way, the emotion factor estimation support apparatus 1 can estimate the factor that caused the emotion even if the factor that caused the poster's emotion is in the other person's message. .

(Modification 4)
In the emotion factor word dictionary shown in FIGS. 4 and 8, the emotion factor words are associated with the attributes of the poster such as age, gender, hobby, and temperament. The emotion factor word dictionary is not limited to such a form, and the attribute of the poster is a poster ID, and the storage unit 13 may store a different emotion factor word dictionary for each poster. By doing in this way, the emotion factor estimation assistance apparatus 1 can further improve the precision which extracts the character string which shows the factor which the contributor held the emotion corresponding to an emotion word.

[Effects of Emotion Factor Estimation Support Device 1 of this Embodiment]
As described above, the emotion factor estimation support apparatus 1 according to the present embodiment refers to the emotion factor word dictionary corresponding to the attribute of the poster so that the poster can express the emotion corresponding to the emotion word included in the posted message. It is possible to output a character string with a high probability indicating the factor held. By doing in this way, the user of the emotion factor estimation assistance apparatus 1 can improve the precision which estimates the factor which caused the contributor's emotion appearing in the text.

  In addition, the selection unit 144 can increase the processing speed by determining the target range for selecting the factor character string based on the position of the sentiment word extracted by the extraction unit 143 in the posted message. This increases the possibility of selecting a character string.

  In addition, when there are a plurality of emotion factor words corresponding to the emotion word, the selection unit 144 is a factor character related to the emotion factor word that has a higher probability of being assumed as the emotion factor of the poster based on the attribute of the poster. Select columns preferentially. By doing in this way, even if there are a plurality of character strings similar to the emotion factor word in the posted message, the user of the emotion factor estimation support apparatus 1 can estimate the emotion factor with high accuracy.

  As mentioned above, although this invention was demonstrated using embodiment, the technical scope of this invention is not limited to the range as described in the said embodiment. It will be apparent to those skilled in the art that various modifications or improvements can be added to the above embodiment. In particular, the specific embodiments of the distribution / integration of the devices are not limited to those illustrated above, and all or a part thereof may be added in arbitrary units according to various additions or according to functional loads. It can be configured functionally or physically distributed and integrated.

  In the above description, tweets are exemplified as posted messages. However, posted messages are not limited to tweets, and the present invention can be applied to arbitrary messages. For example, by using the emotion factor estimation support device 1, based on the content of the message sent to the customer consultation center, the reason why the person who sent the message to the customer consultation center is angry can be estimated with high accuracy. Become.

DESCRIPTION OF SYMBOLS 1 Emotion factor estimation support apparatus 2 Contributor terminal 3 Server 11 Communication part 12 Operation reception part 13 Memory | storage part 14 Control part 131 Emotion word dictionary 132 Emotion factor word dictionary 141 Acquisition part 142 Attribute estimation part 143 Extraction part 144 Selection part 145 Output part N network S emotion factor estimation support system

Claims (11)

A storage unit that stores an emotion factor word dictionary in which an emotion word that represents an emotion of the poster and an emotion factor word that represents the factor that caused the emotion represented by the emotion word are associated with each attribute of the poster of the message;
An acquisition unit for acquiring a post message created by the contributor;
An attribute estimation unit for estimating the attribute of the poster who created the posted message;
An extraction unit for extracting emotion words included in the posted message;
A factor character string similar to the emotion factor word associated with the emotion word extracted by the extraction unit in the emotion factor word dictionary corresponding to the attribute of the poster estimated by the attribute estimation unit is extracted from the posted message. A selection section to select;
Emotion factor estimation support device having The selection unit includes at least one of the emotion factor words associated with the emotion word extracted by the extraction unit in the emotion factor word dictionary corresponding to the attribute of the poster estimated by the attribute estimation unit in the posted message. Selecting the factor string including a part of the string from the posted message;
The emotion factor estimation support apparatus according to claim 1. The selection unit determines a target range for selecting the factor character string based on the position of the emotion word extracted by the extraction unit in the posted message.
The emotion factor estimation support apparatus according to claim 1 or 2. The storage unit further stores probability information indicating a degree of probability that a factor that a poster has an emotion represented by the emotion word is the emotion factor word in association with the emotion factor word;
When the storage unit stores a plurality of the emotion factor words corresponding to the emotion word extracted by the extraction unit, the selection unit relates to an emotion factor word having a higher probability indicated by the probability information. Preferentially select the factor string;
The emotion factor estimation support apparatus according to any one of claims 1 to 3. The acquisition unit acquires a plurality of the posted messages,
The selection unit selects the factor character string included in a post message created before or after the post message from which the extraction unit has extracted the emotion word.
The emotion factor estimation support apparatus according to any one of claims 1 to 4. If the selection unit determines that the factor character string is not included in the post message from which the extraction unit has extracted the emotion word, the selection unit is prior to the post message from which the extraction unit has extracted the emotion word. Or select the factor string included in the post message created later,
The emotion factor estimation support device according to claim 5. An output unit that outputs the emotion word extracted by the extraction unit and the factor character string selected by the selection unit in association with the posted message acquired by the acquisition unit;
The emotion factor estimation support device according to any one of claims 1 to 6. When the selection unit cannot select the emotion factor word associated with the emotion word, the selection unit further includes a reception unit that receives an input of the emotion factor word to be registered in the emotion factor word dictionary in association with the emotion word.
The emotion factor estimation support device according to any one of claims 1 to 7. The acquisition unit further acquires another person message related to the post message created by the contributor, created by another person different from the contributor,
The selection unit selects the factor character string included in the other person message;
The emotion factor estimation support device according to any one of claims 1 to 8. The computer runs,
Retrieving a posted message created by the message poster;
Estimating the attributes of the poster who created the posted message;
Extracting an emotion word included in the posted message;
An emotion factor word dictionary in which an emotion word representing a contributor's emotion and an emotion factor word representing a factor causing the emotion represented by the emotion word are associated, and the emotion factor corresponding to the estimated attribute of the poster Selecting a factor character string associated with the emotion factor word associated with the emotion word extracted in the extracting step from the posted message by referring to a word dictionary;
Emotion factor estimation support method having Retrieving a posted message created by the message poster;
Estimating the attributes of the poster who created the posted message;
Extracting an emotion word included in the posted message;
An emotion factor word dictionary in which an emotion word representing a contributor's emotion and an emotion factor word representing a factor causing the emotion represented by the emotion word are associated, and the emotion factor corresponding to the estimated attribute of the poster Selecting a factor character string associated with the emotion factor word associated with the emotion word extracted in the extracting step from the posted message by referring to a word dictionary;
Emotion factor estimation support program to make computer execute.

Images