JP6718535B2 - Evaluation device, evaluation method, and evaluation program - Google Patents

Description

The present invention relates to an evaluation device, an evaluation method, and an evaluation program.

Conventionally, information about brands used by companies and organizations (hereinafter referred to as “company”) has been collected in various media such as TV, newspapers, magazines, Web articles, etc., and the degree of brand exposure in the world and brand The image etc. are evaluated. By evaluating such brands, companies can be useful in making brand strategy decisions.

For example, in Patent Document 1, information about a predetermined brand exposed to the media is collected, the degree of exposure of the brand to the media, the brand interest level indicating the public interest regarding the brand, and the like are calculated, and the information is integrated. There has been proposed a brand evaluation method for calculating a predetermined evaluation index. In addition, brand evaluation methods using various indexes are known (see Patent Documents 2 to 5).

JP, 2005-95249, A JP, 2014-32636, A International Publication No. 2004-066175 International Publication No. 2004-055707 JP, 2003-108738, A

In the brand evaluation method described in Patent Document 1, the information exposed to the media regarding a predetermined brand is acquired and the evaluation index is calculated. In this evaluation method, for example, a company that is the owner of the brand I could not evaluate the gap between the brand image that I expected and the reaction of the public. Moreover, it was not possible to properly evaluate whether the information exposed to the media was positive or negative.

Further, in the conventional brand evaluation method, the media such as TV, newspapers, magazines, etc. have been emphasized, but in recent years, various media including social networking services (SNS), etc., in which the public's reaction is noticeable are comprehensively integrated. And it needs to be evaluated equally.

The present invention has been made in consideration of such circumstances, and evaluates the gap between the brand image that a company expects and the reaction of the public (customers), and determines the public It is an object of the present invention to provide an evaluation device, an evaluation method, and an evaluation program capable of appropriately evaluating whether a reaction has occurred. Further, according to the present invention, when evaluating the gap between the brand image and the reaction of the public, it is possible to appropriately evaluate the accuracy of information exposed to the media and also to evaluate the degree of risk of the company. It is an object of the present invention to provide a simple evaluation device, evaluation method, and evaluation program.

(1): A collection unit that collects data from an information medium (for example, the collection unit 10 in the embodiment), and processing target data related to a predetermined evaluation target is extracted from the data collected by the collection unit. An extracting unit (for example, the screening unit 12 in the embodiment) and a tagging unit (for example, the tagging unit 16 in the embodiment) that performs predetermined tagging on the processing target data extracted by the extracting unit. If, on the basis of the processed data tagged by the tagging unit, calculation unit for calculating an index value for said evaluated (e.g., scoring unit 18 in the embodiment) Bei example and, the tagging section Is a tag indicating the evaluation target, a topic tag indicating a topic, a mirror tag indicating expected contents for the evaluation target, a thermo tag indicating a reaction of a consumer, and a text included in the processing target data. A risk tag indicating a risk related to the evaluation target is tagged, and the calculation unit is tagged in the processing target data with a combination of the tag indicating the evaluation target, the topic tag, and the mirror tag. 1 text, a second text tagged with the combination of the tag indicating the evaluation target, the topic tag and the thermo tag, and a combination of the tag indicating the evaluation target, the topic tag and the risk tag The evaluation device calculates the index value for at least one of the third texts .

(2): In (1), the tagging unit tags the text included in the processing target data using predefined dictionary data.

(3): In (1) or (2), the evaluation target is preset by a company or an organization.

(4) :( 3), wherein the tagging unit, to the text contained in the processing target data, corporate tag is a tag showing the evaluation target, the topic tag, the semantic analysis of the processed data Kansei tags used for the Miratagu showing the contents of the company or organization expects, is to tag the risk tag indicating the risk of the Samotagu, and the company or organization.

(5): In any one of (1) to (4), the analysis unit that divides the text included in the processing target data extracted by the extraction unit into words and associates evaluation targets having the same meaning ( For example, the analysis unit 14) in the embodiment is further included.

(6): In (5), the analysis unit grasps the relationship between the dependency source and the dependency destination of the word included in the text, and the expression strength, multiple negation, and affirmative question of the word included in the text. , Comparison, and dialect interpretation.

(7): In (5) or (6), the analysis unit analyzes the degree to which the evaluation target is positively expressed in the processing target data, and the calculation unit performs the processing by the analysis unit. The index value is calculated based on the analysis result of the target data.

(8): In (3), the calculation unit calculates, for the first text , a first index value indicating an appearance frequency of a first term indicating the content expected by the company or organization, and the first index value is calculated. The second index value indicating the appearance frequency of the second term indicating the reaction of the consumer is calculated for the two texts, and the appearance frequency of the third term indicating the risk of the company or group is calculated for the third text. The third index value based on is calculated.

(9): In (8), is the calculation unit showing a positive content for each of the first index value and the second index value for each of the data collected from the information medium? It is to determine whether it indicates a negative content.

(10): In (8) or (9), the calculation unit weights the data including the first term and the data including the second term based on the type of the information medium, The first index value and the second index value are calculated.

(11): An evaluation system including the evaluation device according to any one of (1) to (10) and an information medium connected to the evaluation device via a network.

(12): A computer collects data from an information medium, extracts processing target data related to a predetermined evaluation target from the collected data, and sets a predetermined value for the extracted processing target data. Is an evaluation method for calculating an index value for the evaluation target based on the tagged processing target data, wherein the evaluation target is set for the text included in the processing target data. A tag indicating a topic, a topic tag indicating a topic, a mirror tag indicating content expected for the evaluation target, a thermo tag indicating a reaction of a consumer, and a risk tag indicating a risk associated with the evaluation target, and the processing. In the target data, the first text tagged with the combination of the tag indicating the evaluation target, the topic tag, and the mirror tag, and the combination of the tag indicating the evaluation target, the topic tag, and the thermo tag A second text, and at least one of the third text tagged with the combination of the tag indicating the evaluation target, the topic tag, and the risk tag is an evaluation method for calculating the index value. ..

(13): causing a computer to collect data from an information medium, extracting processing target data related to a predetermined evaluation target from the collected data, and performing predetermined processing on the extracted processing target data. Is an evaluation program for causing an index value for the evaluation target to be calculated based on the tagged processing target data , wherein the computer includes a text for the text included in the processing target data. , A tag indicating the evaluation target, a topic tag indicating the topic, a mirror tag indicating the expected contents for the evaluation target, a thermo tag indicating the reaction of the customer, and a risk tag indicating the risk related to the evaluation target. In the processing target data, the first text tagged with the combination of the tag indicating the evaluation target, the topic tag, and the mirror tag, the tag indicating the evaluation target, the topic tag, and the thermo tag The index value is calculated for at least one of the second text tagged with the combination and the third text tagged with the combination of the tag indicating the evaluation target, the topic tag, and the risk tag. , An evaluation program.

According to (1) to (8) and (11) to (13) , the gap between the image of the brand expected by the company or the organization and the reaction of the public (customers) is evaluated, and the activity performed by the company or the organization. It is possible to appropriately evaluate how the world reacted by such things. Moreover, when assessing the gap between the brand image and the reaction of the public, it is possible to assess appropriately including the certainty of the information exposed in the media, and further assess the degree of risk of the company.

According to (9), it is possible to grasp whether each of the data collected from the information medium has a positive content or a negative content.

According to (10), the data collected from various information media can be evaluated comprehensively and equally.

It is a functional block diagram showing an example of an evaluation device in an embodiment. It is a figure which shows an example of the dictionary data memorize|stored in the mirror dictionary in embodiment. It is a figure which shows an example of the dictionary data memorize|stored in the thermo dictionary in embodiment. It is a figure which shows an example of the dictionary data memorize|stored in the risk dictionary in embodiment. It is a figure which shows the other example of the dictionary data memorize|stored in the risk dictionary in embodiment. It is a flow chart which shows an example of a flow of processing of an evaluation device in an embodiment. It is a flow chart which shows an example of a flow of tagging processing of an evaluation device in an embodiment. It is a figure which shows an example of the process target data to which the tagging process was performed in embodiment. It is a flow chart which shows an example of a flow of scoring processing of an evaluation device in an embodiment. It is a figure which shows an example of the scoring process of the evaluation apparatus in embodiment. It is a figure which shows an example of the time series data of the mirror score and thermoscore with respect to the brand A in embodiment. It is a figure which shows an example of the time series data of the mirror score and thermoscore with respect to the brand B in embodiment. It is a figure which shows an example of the classification|category mirror score with respect to the brand A in embodiment. It is a figure which shows an example of the classification|category thermoscore with respect to the brand A in embodiment. It is a figure which shows an example of the time series data of the thermoscore about a brand A, a mirror score, a risk value, and the number of the process target data in embodiment. It is a figure which shows another example of the risk level determination process of the evaluation device in embodiment.

Hereinafter, an evaluation apparatus, an evaluation method, and an evaluation program according to some embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a functional block diagram showing an example of an evaluation device 1 according to an embodiment of the present invention. The evaluation device 1 collects data to be processed (hereinafter, referred to as “processing target data”) from the medium M (information medium) and analyzes the data to evaluate the brand to be evaluated. The media M includes, for example, televisions, newspapers, magazines, web articles, weblogs, SNS such as short text posting services, and reports of shareholders.

The evaluation device 1 calculates three values of “Mirror score (first index value)”, “Thermo score (second index value)”, and “Risk value (third index value)” as the evaluation index related to the brand. .. "Mirror score" is how the public owns the brand being evaluated, as opposed to the content that defines how the brand wants to be seen by the public. It is an index to represent. The “mirror score” indicates the appearance frequency of a term (first term) indicating the content expected by the company. By calculating this mirror score, the gap between the image of the brand expected by the company and the reaction of the public can be grasped.

The “thermo score” is an index showing whether the public feeling can be raised by the activity of the company that owns the brand to be evaluated. "Thermo-score" indicates the frequency of appearance of a term (second term) indicating a public (customer) reaction. By calculating this thermoscore, it is possible to know whether the public feelings about the brand to be evaluated can be increased or whether the public support has been obtained. The “risk value” is an index indicating the degree of risk that has occurred with respect to the brand to be evaluated. The “risk value” is calculated based on the appearance frequency of the term (third term) indicating the risk of the company or the organization and the type of the medium M (information medium).

The evaluation device 1 includes, for example, a collection unit 10, a screening unit 12 (extraction unit), an analysis unit 14, a tagging unit 16, a scoring unit 18 (calculation unit, generation unit), and a display unit 20. The storage unit 22 and the dictionary DB 24 are provided. The dictionary DB 24 is, for example, a company dictionary 30, a topic dictionary 32, a sensitivity dictionary 34, a mirror dictionary 36 (first storage unit), a thermo dictionary 38 (second storage unit), and a risk dictionary 40 (third storage). Section) and.

The collection unit 10 collects data to be processed from the medium M and stores the data in the storage unit 22. The collection unit 10 collects the processing target data via the Internet N, for example. The collection unit 10 collects the processing target data at a predetermined timing such as daily or weekly. The media M as the collection source may be determined in advance by the user of the evaluation device 1. Further, the collection unit 10 may perform a crawling process of periodically collecting documents on the Internet including a predefined character string. The evaluation device 1 may include a reception unit (not shown) that receives an input from the operator P, and the collection unit 10 may collect the processing target data input to the reception unit.

The screening unit 12 excludes, from the processing target data collected by the collection unit 10, processing target data that is not related to the brand to be evaluated. The screening unit 12 extracts, from the data collected from the medium M (information medium), data relating to a brand preset by a company or an organization. For example, the screening unit 12 reads the dictionary data stored in the company dictionary 30. This dictionary data becomes the brand to be evaluated. Then, the screening unit 12 excludes the processing target data including the homonyms of this brand. Further, for example, the screening unit 12 excludes the processing target data acquired from the predefined specific URL, excludes the processing target data acquired from the predefined weblog and the SNS, and sets the same text. If there is a large amount of data including "", it may be excluded as abnormal data.

The analysis unit 14 divides the text included in the processing target data into word levels, and performs a morphological analysis process that associates brands with the same meaning. For example, the analysis unit 14 associates alphabetical notation, kanji notation, katakana notation, hiragana notation, and the like indicating the same brand, and processes them as data indicating the same brand. Further, regarding the alphabetical representation, the analysis unit 14 may process the data as data indicating the same brand even when there is a difference in uppercase letters, lowercase letters, and mixed letters of uppercase and lowercase letters. Further, the analysis unit 14 may process text including erroneous writing (such as erroneous kanji conversion) as data indicating a brand to be evaluated. As a result, it is possible to correct the fluctuation of the notation.

In addition, the analysis unit 14 grasps the relationship between the modification source and the modification destination of the words included in the text, and the strength of expression of the words included in the text, multiple denial, affirmative question, modification, comparison, and dialect. Performs parsing processing for interpretation. Interpreting the strength of the expression means, for example, “extreme” in the expression “A product is extremely good” and “slightly” in the expression “A product is slightly good” means that the former “extreme” is more It is to interpret it as a strong expression. Further, to interpret the multiple negation is to correctly interpret an expression including a double negation such as “A product is not bad” as a positive expression.

In addition, interpreting an affirmative question is, for example, interpreting a question sentence expressed with a positive intention of "A product is a good product, isn't it?" In addition, interpreting dependency means correctly interpreting the meaning of a modifier even when the positions of modifiers are reversed, such as the expression "A product is good, right?" is there. This "good product is right?" is interpreted as a positive expression.

In addition, interpreting the comparison means that the product A became worse than the expression "the product A was better than the previous model" comparing the current model of the product A and the previous model. To make a judgment and interpret it as a negative expression. Further, to interpret a dialect is to correctly interpret the meaning of the dialect “Menkoi” in the expression “A product is a product of Menkoi”, judging that it is the intent of the standard word “kawaii”. .. This "A product is a mental product" is interpreted as a positive expression. By performing the syntax analysis processing as described above, it is possible to improve the accuracy of meaning understanding.

The tagging unit 16 refers to the company dictionary 30, the topic dictionary 32, the sensitivity dictionary 34, the mirror dictionary 36, the thermo dictionary 38, and the risk dictionary 40, and refers to the text included in each of the processing target data. Then, "company tag", "topic tag", "sensitivity tag", "mirror tag", "thermo tag", and "risk tag" are tagged. Details of the tagging process will be described later.

The scoring unit 18 performs a scoring process for calculating a mirror score, a thermoscore, and a risk value based on the processing target data that has been tagged by the tagging unit 16. For example, the scoring unit 18 calculates a mirror score for a character string tagged with a combination of “company tag”, “topic tag”, and “mirror tag” in the processing target data. Further, for example, the scoring unit 18 calculates a thermoscore for a character string tagged with a combination of “company tag”, “topic tag”, and “thermotag” in the processing target data. Further, for example, the scoring unit 18 calculates a risk value for a character string tagged with a combination of “company tag”, “topic tag”, and “risk tag” in the processing target data. Further, for example, the scoring unit 18 calculates a risk value for a character string tagged with a combination of “company tag” and “risk tag” in the processing target data.

Some or all of the collection unit 10, the screening unit 12, the analysis unit 14, the tagging unit 16, and the scoring unit 18 are realized by a processor (computer) executing a program (software). Further, some or all of these may be realized by hardware such as LSI (Large Scale Integration) and ASIC (Application Specific Integrated Circuit), or may be realized by a combination of software and hardware.

The display unit 20 displays the mirror score, the thermoscore, and the risk value calculated by the scoring unit 18. The user of the evaluation device 1 can evaluate the target brand by confirming the mirror score, the thermoscore, and the risk value displayed on the display unit 20. The display unit 20 is, for example, a liquid crystal display or an organic EL (Electroluminescence) display device.

The storage unit 22 stores the processing target data collected by the collection unit 10. The storage unit 22 stores, for example, the type of the collection source medium and the processing target data in association with each other.

The dictionary DB 24 stores dictionary data used in various processes performed by the screening unit 12, the analysis unit 14, the tagging unit 16, and the scoring unit 18. The storage unit 22 and the dictionary DB 24 are realized by a ROM (Read Only Memory), a RAM (Random Access Memory), a HDD (Hard Disk Drive), a flash memory, or the like.

The company dictionary 30 stores dictionary data of brands that are owned by a company and are to be evaluated. The company dictionary 30 stores terms that characterize the company, such as a company name, a product name, a service name, and a company president's name.

The topic dictionary 32 stores dictionary data of topics that are supposed to be topical in the world. For example, when information such as a news release is announced to the public regarding a brand to be evaluated, the topic dictionary 32 stores the topics described in this news release. The topic dictionary 32 stores, for example, terms associated with a brand to be evaluated, such as “new model”, “launch start”, “announcement”, and “world premiere”.

The sensitivity dictionary 34 stores dictionary data used for semantic analysis of the processing target data. The sensitivity dictionary 34 stores dictionary data necessary for positive and negative determination processing of processing target data performed by the scoring unit 18.

The mirror dictionary 36 stores dictionary data that defines how the brand of the company wants to be seen in the world, and in what language the brand should be expressed. FIG. 2 is a diagram showing an example of dictionary data stored in the mirror dictionary 36 in this embodiment. The mirror dictionary 36 stores at least one major category, at least one medium category associated with each major category, and a keyword associated with each major category. This keyword is used as dictionary data.

For example, in the example shown in FIG. 2, “trust” is stored as the major category 1, “safe” is stored as the intermediate category 1-1 associated with this major category 1, and “high quality” is stored as the intermediate category 1-2. Is remembered. Furthermore, "safety", "safety", and "trust" are stored as keywords associated with the middle category 1-1, and "high quality", "safety", and "unbreakable" are stored as the keywords associated with the middle category 1-2. Is remembered. It should be noted that the mirror dictionary 36 does not need to have a data structure of large classification, middle classification, and keyword, and may define only large classification, finer classification, or classification. Instead, only the keyword may be stored.

The thermo dictionary 38 stores dictionary data for evaluating whether or not the activity of the company that owns the brand to be evaluated could raise public sentiment. FIG. 3 is a diagram showing an example of dictionary data stored in the thermo dictionary 38 in the present embodiment. In the example shown in FIG. 3, the thermo dictionary 38 stores “safety”, “satisfaction”, “like”, and “expectation” as terms that represent positive emotions of the world. In addition, the thermo dictionary 38 stores, as terms that represent negative emotions of the world, terms that are opposite to the above-mentioned terms that represent positive emotions, such as “anxiety”, “dissatisfaction”, “dislike”, and “disappointment”. ing. Further, the thermo dictionary 38 may store keywords associated with the above-mentioned terms expressing emotions. The thermo-dictionary 38 may store, for example, "come check", "look forward" and the like as keywords associated with "expectations".

The risk dictionary 40 stores dictionary data indicating risks that a company should grasp regarding a brand to be evaluated. FIG. 4A is a diagram showing an example of dictionary data stored in the risk dictionary 40 in this embodiment. FIG. 4B is a diagram showing another example of dictionary data stored in the risk dictionary 40 in this embodiment. The risk dictionary 40 includes at least one major category, at least one medium category associated with each major category, at least one minor category associated with each major category, and The keywords associated with the subcategories are stored. This keyword is used as dictionary data.

For example, in the example shown in FIGS. 4A and 4B, “company origin” is stored as the major category 1, “management quality” is stored as the major category 1-1 associated with this major category 1, and the major category 1— “Product quality” is stored as 2. Further, “achievement” is stored as the sub-class 1-1-1 associated with the mid-class 1-1, and “personnel” is stored as the sub-class 1-1-2. Further, “performance deterioration”, “stock price drop”, and “fraud” are stored as the keywords associated with the sub-class 1-1-1, and “restructuring”, “keywords associated with the sub-class 1-1-2” are stored. Hard work” and “dismissal” are remembered. Note that the risk dictionary 40 does not need to have a data structure of large classification, middle classification, small classification, and keyword, and may set only large classification or only large classification and middle classification, or define more detailed classification. Alternatively, only the keywords may be stored without defining the classification.

Next, the operation of the evaluation device 1 in this embodiment will be described. FIG. 5 is a flowchart showing an example of the processing flow of the evaluation device 1 in the present embodiment.

First, the collection unit 10 collects the processing target data from the medium M and stores it in the storage unit 22 (step S101). The collection unit 10 collects the processing target data via the Internet N, for example. The collection unit 10 collects the processing target data at a predetermined timing such as daily or weekly. Further, the evaluation device 1 may include a reception unit (not shown) that receives an input from the operator P, and the collection unit 10 may collect the processing target data input to the reception unit.

Next, the screening unit 12 reads the processing target data stored in the storage unit 22 and performs a screening process for excluding data unrelated to the brand to be evaluated (step S103). For example, the screening unit 12 reads dictionary data (brand to be evaluated) from the company dictionary 30 and excludes data including homonyms of this brand.

Next, the analysis unit 14 divides the text included in the processing target data into word levels, and performs a morphological analysis process that associates brands with the same meaning (step S105). For example, the analysis unit 14 correlates alphabetical notation, kanji notation, katakana notation, hiragana notation and the like indicating the same brand, and processes them as data indicating the same brand to correct the fluctuation of the notation.

Next, the analysis unit 14 grasps the relationship between the modification source and the modification destination of the words included in the text, and the strength of expression of the words included in the text, multiple negatives, affirmative doubts, modification, comparison, and dialect. Parsing processing is performed to interpret (step S107).

Next, the tagging unit 16 selects a “company tag”, a “topic tag”, a “sensitivity tag”, a “mirror tag”, a “thermo tag”, and a “risk tag” for the text included in each of the processing target data. Tagging is performed (step S109).

FIG. 6 is a flowchart showing an example of the flow of the tagging process (step S109) of the evaluation device 1 according to this embodiment. First, the tagging unit 16 extracts one data out of the processing target data subjected to the syntax analysis processing by the analysis unit 14 (step S201).

Next, the tagging unit 16 selects a “company tag”, a “topic tag”, a “sensitivity tag”, a “mirror tag”, a “thermo tag”, and a “risk tag” for the text included in each of the processing target data. Tagging is performed (step S203). When the character string of the dictionary data stored in the company dictionary 30 is included in the processing target data, the tag attaching unit 16 adds the “company tag” to the character string in the processing target data. When the character string of the dictionary data stored in the topic dictionary 32 is included in the processing target data, the tag attaching unit 16 adds a “topic tag” to the character string in the processing target data. When the character string of the dictionary data stored in the sensitivity dictionary 34 is included in the processing target data, the tag attaching unit 16 adds the “kansei tag” to the character string in the processing target data. When the character string of the dictionary data stored in the mirror dictionary 36 is included in the processing target data, the tag attaching unit 16 adds a “mirror tag” to the character string in the processing target data. When the character string of the dictionary data stored in the thermo dictionary 38 is included in the processing target data, the tag attaching unit 16 adds a “thermo tag” to the character string in the processing target data. If the character string of the dictionary data stored in the risk dictionary 40 is included in the processing target data, the tag attaching unit 16 adds a “risk tag” to the character string in the processing target data.

FIG. 7 is a diagram showing an example of the processing target data on which the tagging process is performed in the present embodiment. In the example shown in FIG. 7, “company tag” is added to the character strings “company A” and “product A”, “topic tag” is added to the character string “start of sale”, and "Mirror tag" is added to the character string "Target", "Thermo tag" and "Kansei tag" are added to the character string "Expectation", and "Risk tag" is added to the character string "Fault". Has been granted.

Next, the tagging unit 16 determines whether the tagging process has been completed for all the process target data (step S205). When the tagging unit 16 determines that the tagging process has not been completed for all the process target data, the tagging unit 16 extracts the process target data for which the tagging process has not been performed and performs the tagging process. On the other hand, when the tagging unit 16 determines that the tagging process has been completed for all the process target data, the tagging process is completed.

Next, the scoring unit 18 performs a scoring process for calculating a mirror score, a thermoscore, and a risk value based on the processing target data tagged by the tagging unit 16 (step S111).

FIG. 8 is a flowchart showing an example of the flow of the scoring process (step S111) of the evaluation device 1 according to this embodiment. First, the scoring unit 18 extracts one piece of data to be processed that has been tagged by the tagging unit 16 (step S301).

Next, the scoring unit 18 targets the processing target data tagged with the combination of the “company tag”, the “topic tag”, and the “mirror tag”, and sets the mirror score (hereinafter, “individual mirror”) for each processing target data. "Score") is calculated (step S303). For example, the scoring unit 18 counts the number of “mirror tags” and calculates an individual mirror score with one “mirror tag” as one point.

Next, the scoring unit 18 targets the processing target data tagged with the combination of the “company tag”, the “topic tag”, and the “thermo tag” for each processing target data (hereinafter, “individual thermostat”). "Score") is calculated (step S305). For example, the scoring unit 18 counts the number of “thermo tags” and calculates an individual thermo score with one “thermo tag” as one point.

FIG. 9: is a figure which shows an example of the scoring process of the evaluation apparatus 1 in this embodiment. In the example shown in FIG. 9, “process target data 1” includes a character string with a “topic tag” of “new model”, a character string with a “company tag” of “product A”, It includes a character string with "mirror tag (major classification "Yorokobi")" that was "cool" and a character string with "thermo tag (keyword associated with "expectation")" that is "check". In this case, the scoring unit 18 calculates the individual mirror score as “1.0” for the combination of “product A”, “new model”, and “horizontal scale”. Further, the scoring unit 18 calculates the individual thermoscore as "1.0" by combining "product A", "new model" and "expectation".

"Processing target data 2" is a character string with a "topic tag" of "new model", a character string with a "mirror tag (major classification "Yorokobi")" of "cool", and "Check it out" And a character string to which a "thermo tag (keyword associated with "expectation")" is added. In this case, since the character string to which the “company tag” is added does not exist, the scoring unit 18 sets the individual mirror score and the individual thermoscore to “0.0”.

“Processing target data 3” is a character string with a “topic tag” of “new model”, a character string with “company tag” of “product A” and “company A”, and “cool” Character string to which the "mirror tag (major classification "Yorokobi") is added and a character string to which the "thermo tag (keyword associated with "expectation")" called "exact check" is added. That is, the "processing target data 3" includes two character strings "Product A" and "A company" to which the "company tag" is added. In this case, the scoring unit 18 calculates an individual mirror score and an individual thermoscore for each of these two character strings.

Here, in the "processing target data 3", the positive expression "The display was very cool" is included, but this may not express the product A or the company A positively. In other words, the expression "the exhibition was very cool" can be interpreted as a positive expression about the "exhibition" method. In this case, the scoring unit 18 calculates an individual mirror score and an individual thermoscore that are lower than the processing target data 1 that directly and positively expresses the product A. For example, the scoring unit 18 calculates the individual mirror score as “0.5” for the combination of “product A”, “new model” and “horizontal scale”, and calculates “Company A”, “new model” and “horizontal scale” The individual mirror score is calculated as "0.5" by the combination of. In addition, the scoring unit 18 calculates the individual thermoscore as “0.5” by combining “product A”, “new model” and “expectation”, and calculates “Company A”, “new model” and “expectation”. The individual thermoscore is calculated as "0.5" by the combination of.

Next, the scoring unit 18 targets the processing target data tagged with the combination of the “company tag”, the “topic tag”, and the “risk tag”, and sets the risk value for each processing target data (hereinafter, “individual A risk value” is calculated (step S307). For example, the scoring unit 18 counts the number of “risk tags” and scores each “risk tag” as one point.

Next, the scoring unit 18 determines whether or not the individual scoring processing has been completed for all the processing target data (step S309). When the scoring unit 18 determines that the individual scoring processing has not been completed for all the processing target data, it extracts the processing target data for which the individual scoring processing has not been performed and performs the individual scoring processing. ..

On the other hand, when the scoring unit 18 determines that the individual scoring processing has been completed for all the processing target data, it performs the weighting processing based on the type of the collection source of each processing target data (step S311). The scoring unit 18 weights the data including the term stored in the mirror dictionary 36 and the data including the term stored in the thermo dictionary 38 based on the type of the medium M (information medium), and will be described later. Calculate the mirror score and thermoscore. For example, based on past statistical data, the average amount of speech of the brand to be evaluated per day for each medium is calculated, and weighted so that the influence of all the media becomes uniform. For example, when the weight of the newspaper is set to the reference value “1”, a weight “2” larger than that of the newspaper is set for a television having a smaller amount of speech than the newspaper. Further, for an SNS that has a large amount of speech compared to newspapers, a weight "0.1" that is smaller than that of newspapers is set.

Next, the scoring unit 18 determines whether the content of the processing target data is a positive expression or a negative expression based on the “sensitivity tag” given to the processing target data. Is performed (step S313). The scoring unit 18 determines, for each of the data to be processed, whether it shows a positive (positive) content or a negative (negative) content for each mirror score and thermoscore. To do.

For example, in the scoring unit 18, the number of places showing a positive expression (the number of affective tags of a positive expression) in one processing target data is the number of places showing a negative expression (the sensitivity of the negative expression). If the number of tags is larger than the number of tags), it is determined that the processing target data is a positive expression as a whole. In addition, when the number of places showing a negative expression is larger than the number of places showing a positive expression in one processing target data, the scoring unit 18 determines that the processing target data as a whole is negative. Judge as an expression. If the number of places showing a positive expression and the number of places showing a negative expression are the same in one processing target data, the scoring unit 18 determines that the processing target data as a whole. Determined to be a neutral expression. In addition, the scoring unit 18 determines that the processing target data to which the “sensitivity tag” is not attached is emotionless.

Next, the scoring unit 18 calculates the final thermoscore and mirror score based on the processing target data that has undergone the emotion determination processing (step S315). For example, the scoring unit 18 sums the individual mirror scores of the processing target data determined to be positive expressions and the individual mirror scores of the processing target data determined to be negative expressions for each brand to be evaluated. Calculate the sum of The scoring unit 18 subtracts a value obtained by subtracting the sum of the individual mirror scores of the processing target data determined to be the negative expression from the total of the individual mirror scores of the processing target data determined to be the positive expression, Calculate as the final Miller score.

Further, for example, the scoring unit 18 may sum the individual thermoscores of the processing target data determined to be positive expressions and the individual processing target data determined to be negative expressions for each brand to be evaluated. Calculate the sum of the thermoscores. Further, the scoring unit 18 subtracts the sum of the individual thermoscores of the processing target data determined to be the negative expression from the sum of the individual thermoscores of the processing target data determined to be the positive expression, Calculate as the final thermoscore. The scoring unit 18 also stores the calculated mirror score and thermoscore in the storage unit 22.

Next, the scoring unit 18 calculates a final risk value based on the processing target data that has been subjected to the emotion determination process, and determines the risk level (step S317). In the determination of the risk level, for example, “high”, “medium”, and “low” of the risk are determined based on a predetermined threshold value. In this case, a threshold may be set for each classification in the risk dictionary shown in FIGS. 4A and 4B. For example, even if the risk value is the same, if the “defect” of the subclass 1-2-1 is considered to be more serious in risk occurrence than the “performance” of the subclass 1-1-1, The threshold value of "defect" of sub-class 1-2-1 is set lower than that of "achievement" of sub-class 1-1-1. With the above, the scoring process is completed.

Next, the scoring unit 18 performs an evaluation result output process of displaying the calculated mirror score, thermoscore, risk value, etc. on the display unit 20 (step S113). The scoring unit 18 generates information capable of displaying an image in which a mirror score, a thermoscore, and a risk value are compared. FIG. 10: is a figure which shows an example of the time series data of the mirror score and thermoscore with respect to the brand A in this embodiment. FIG. 11: is a figure which shows an example of the time series data of the mirror score and thermoscore with respect to the brand B in this embodiment. In the example shown in FIG. 10, the values of the Miller score and the thermoscore significantly decrease on July 20, 2016, and a peak occurs, and the values of the Miller score and the Thermoscore significantly increase on November 1, 2016. Then, a peak occurs, and it can be seen that the values of the mirror score and the thermoscore are around 0 in other periods. Further, in the example shown in FIG. 11, peaks of the mirror score and the thermoscore frequently occur, but it can be seen that the value of each peak is smaller than the peak value of the brand A shown in FIG.

FIG. 12 is a diagram showing an example of classification-based mirror scores for the brand A in the present embodiment. In the example shown in FIG. 12, for each of the major category 1 (trust), major category 2 (new), and major category 3 (challenge) stored in the mirror dictionary shown in FIG. 2, May 1, 2016 to November 1, 2016 Until the day, the average value per day of the mirror score of the processing target data determined to be a positive expression, and the average value per day of the mirror score of the processing target data determined to be a negative expression. It is shown as a bar graph. The mirror score of the processing target data determined to be a negative expression is shown as a negative value. Further, a line graph shows a value obtained by subtracting the average value of the mirror scores of the processing target data determined to be the negative expression from the average value of the mirror scores of the processing target data determined to be the positive expression. Thus, the user of the evaluation device 1 can confirm the mirror score for each classification.

FIG. 13: is a figure which shows an example of the classification|category thermoscore with respect to the brand A in this embodiment. In the example shown in FIG. 13, for each “safety”, “satisfaction”, “like”, and “expectation” stored in the thermo dictionary shown in FIG. , A bar graph showing the average value per day of the thermoscore of the processing target data determined to be a positive expression and the average value per day of the thermoscore of the processing target data determined to be a negative expression .. Also, a line graph shows a value obtained by subtracting the average value of the thermoscores of the processing target data determined to be the negative expression from the average value of the thermoscores of the processing target data determined to be the positive expression. Thereby, the user of the evaluation device 1 can confirm the thermoscore for each classification.

FIG. 14: is a figure which shows an example of the time series data of the thermoscore about a brand A, a mirror score, a risk value, and the number of the process target data in this embodiment. In FIG. 14, it is shown that the average value of the thermoscores by classification on a certain specific day (November 1, 2016) is “71”, and the average value of the mirror scores by classification is “65”. Has been done. Further, in FIG. 14, regarding the risk level of the risk value, the number of cases determined to be “high” is “2”, the number of cases determined to be “medium” is “11”, and it is determined to be “low”. It is shown that the number of cases is “34”. Further, in FIG. 14, the number of pieces of processing target data regarding the brand A in the past 30 days is shown by a line graph. Thus, the user of the evaluation device 1 can confirm the thermoscore, the mirror score, and the risk level regarding the brand A. Although FIG. 14 shows an example in which the thermoscore, the mirror score, and the risk value of a specific day are displayed, the thermoscore, the mirror score, and the risk value of an arbitrary period such as weekly or monthly. May be displayed.

FIG. 15 is a diagram illustrating another example of the risk level determination processing of the evaluation device 1 according to this embodiment. First, the scoring unit 18 determines the validity of the combination of tags assigned to each piece of data to be processed (step S401). The scoring unit 18 determines whether or not each of the processing target data is tagged with a combination of “company tag” and “risk tag”, for example. When the processing target data is tagged with the combination of the “company tag” and the “risk tag”, the scoring unit 18 sets the processing target data as the target of risk level determination. On the other hand, when the processing target data is not tagged with the combination of the “company tag” and the “risk tag”, the scoring unit 18 excludes the processing target data from the target of risk level determination.

Even if the processing target data is tagged with the combination of the “company tag” and the “risk tag”, the scoring unit 18 determines that the processing target data is a positive expression by the emotion determination processing. When the determination is made, this processing target data is excluded from the risk level determination target.

Next, the scoring unit 18 aggregates risk values for each medium M for each of the character strings associated with the “company tag” with respect to the processing target data targeted for risk level determination (step S403). ). Next, the scoring unit 18 determines, for example, a risk level of “large”, “medium”, and “small” for each of the character strings associated with the “company tag”, using a predetermined threshold as a reference ( Step S405). This threshold value is set for each medium M, and for each classification in the risk dictionary shown in FIGS. 4A and 4B.

For example, for each of the media M (for example, “newspaper”, “television”, “web article”, “weblog”, “SNS” regarding “defective” of the sub-class 1-2-1 shown in FIGS. 4A and 4B. In (1), a first threshold value corresponding to the risk level "small", a second threshold value corresponding to the risk level "medium" (first threshold value <second threshold value), and a third threshold value corresponding to the risk level "large". A threshold value (second threshold value<third threshold value) is set. Here, as compared to “newspaper” and “television”, “web articles”, “weblogs”, and “SNS” have a large amount of speech, so “web articles”, “weblogs”, and “SNS” For, a threshold value higher than that of “newspaper” and “TV” may be set. For example, the scoring unit 18 determines that the risk level is “large” when the value of the aggregated risk value is equal to or higher than the third threshold value, and determines the risk level when the value is equal to or higher than the second threshold value and lower than the third threshold value. If the value is equal to or more than the first threshold and less than the second threshold, the risk level is determined to be “small”. Note that the scoring unit 18 may determine “no risk” when the aggregated risk value is less than the first threshold.

Next, the scoring unit 18 totalizes the risk levels determined for each medium M for each of the character strings associated with the "company tag" (step S407). For example, the scoring unit 18 uses a numerical value associated with each of the risk levels “large”, “medium”, and “small”. As the numerical values associated with each of the risk levels “large”, “medium”, and “small”, for example, a value that decreases in the order of the risk levels “large”, “medium”, and “small” is set. The scoring unit 18 calculates the total (numerical value) of the numerical values associated with the risk level determined for each medium for each of the character strings associated with the “company tag”.

The scoring unit 18 totalizes the risk level determination results by cross media as described above, and determines the final risk level based on this total value. For example, the scoring unit 18 sets the final risk levels of “high”, “medium”, “low”, and “safe” for each of the character strings associated with the “company tag” with a predetermined threshold as a reference. Is determined.

The scoring unit 18 may generate a heat map in which the final risk level determined by the scoring unit 18 is expressed in color. In this heat map, for example, character strings associated with “company tags” (character strings of dictionary data stored in the company dictionary 30) are arranged in the column direction, and classifications in the risk dictionary are arranged in the row direction. The final risk level determined by the scoring unit 18 is one square in this heat map.

The calculated mirror score, thermoscore, and risk value may be mapped in a three-dimensional space and the support vector machine may calculate the evaluation value.

According to the evaluation device of the present embodiment described above, it is possible to evaluate the gap between the brand image that a company expects and the reaction of the public, and appropriately evaluate how the public responds by the activities performed by the company. You can Further, the processing target data collected from various media can be evaluated comprehensively and equally. Moreover, when assessing the gap between the brand image and the reaction of the public, it is possible to assess appropriately including the certainty of the information exposed in the media, and further assess the degree of risk of the company.

Although the embodiments for carrying out the present invention have been described above using the embodiments, the present invention is not limited to these embodiments, and various modifications and substitutions are made within the scope not departing from the gist of the present invention. Can be added.

DESCRIPTION OF SYMBOLS 1... Evaluation device, 10... Collection part, 12... Screening part (extraction part), 14... Analysis part, 16... Tagging part, 18... Scoring part (calculation part, generation part), 20... Display part, 22... Storage unit, 24... Dictionary DB, 30... Corporate dictionary, 32... Topic dictionary, 34... Sensitive dictionary, 36... Mirror dictionary (first storage unit), 38... Thermo dictionary (second storage unit), 40... Risk dictionary ( Third storage unit)

Claims (13)

A collection unit that collects data from the information medium,
From the data collected by the collection unit, an extraction unit that extracts processing target data related to a predetermined evaluation target,
A tagging unit for performing a predetermined tagging on the processing target data extracted by the extraction unit,
Based on the processed data tagged by the tagging unit, Bei example and a calculation unit for calculating an index value for the evaluation,
The tagging unit, with respect to the text included in the processing target data, a tag indicating the evaluation target, a topic tag indicating a topic, a mirror tag indicating expected contents for the evaluation target, a reaction of a consumer. Tag the thermo tag shown, and the risk tag showing the risk related to the evaluation target,
In the processing target data, the calculation unit includes a first text tagged with a combination of a tag indicating the evaluation target, the topic tag, and the mirror tag, a tag indicating the evaluation target, the topic tag, and the thermo tag. The index value is targeted to at least one of the second text tagged with the combination of the second text, and the third text tagged with the combination of the tag indicating the evaluation target, the topic tag, and the risk tag. calculate,
Evaluation device. The tagging unit tags the text included in the processing target data using predefined dictionary data,
The evaluation device according to claim 1. The evaluation target is preset by a company or an organization,
The evaluation device according to claim 1 or 2. The tagging unit, to the text contained in the processing target data, corporate tag is a tag showing the evaluation target, the topic tag, sensitive tag used for semantic analysis of the processed data, the the indicating what company or organization expects Miratagu, tagging the risk tag indicating the risk of the Samotagu, and the company or organization,
The evaluation device according to claim 3. The text included in the processing target data extracted by the extraction unit is divided into words, and the analysis unit further includes an evaluation target having the same meaning.
The evaluation device according to any one of claims 1 to 4. The analysis unit grasps a relationship between a dependency source and a dependency destination of words included in the text, and performs expression strength, multiple negation, affirmative question, comparison, and interpretation of dialects of the words included in the text. ,
The evaluation device according to claim 5. The analysis unit analyzes the degree to which the evaluation target is positively expressed in the processing target data,
The calculation unit calculates the index value based on an analysis result of the processing target data by the analysis unit,
The evaluation device according to claim 5 or 6. The calculation unit,
A first index value indicating the frequency of appearance of the first term indicating the content expected by the company or organization is calculated for the first text ,
A second index value indicating the appearance frequency of the second term indicating the reaction of the customer is calculated for the second text ,
Calculating a third index value based on the frequency of appearance of a third term indicating the risk of the company or organization , targeting the third text ,
The evaluation device according to claim 3. For each of the data collected from the information medium, the calculation unit shows a positive content or a negative content for each of the first index value and the second index value. To determine if
The evaluation device according to claim 8. The calculation unit weights the data including the first term and the data including the second term based on the type of the information medium to determine the first index value and the second index value. calculate,
The evaluation device according to claim 8. An evaluation system comprising: the evaluation device according to any one of claims 1 to 10; and an information medium connected to the evaluation device via a network. Computer
Collect data from information media,
From the collected data, processing target data related to a predetermined evaluation target is extracted, and predetermined processing is performed on the extracted processing target data,
Calculating an index value for the evaluation target based on the tagged processing target data,
Evaluation method ,
With respect to the text included in the processing target data, a tag indicating the evaluation target, a topic tag indicating a topic, a mirror tag indicating expected contents for the evaluation target, a thermo tag indicating a reaction of a consumer, and the evaluation Tag with a risk tag that indicates the risk associated with the subject,
In the processing target data, the first text tagged with the combination of the tag indicating the evaluation target, the topic tag, and the mirror tag, and the tag combining the tag indicating the evaluation target, the topic tag, and the thermo tag Calculating the index value for at least one of the attached second text and the third text tagged with the combination of the tag indicating the evaluation target, the topic tag, and the risk tag,
Evaluation method. On the computer,
Collect data from information media,
From the collected data, the processing target data related to a predetermined evaluation target is extracted,
Let the specified tag be applied to the extracted processing target data,
Calculating an index value for the evaluation target based on the tagged processing target data,
An evaluation program ,
On the computer,
With respect to the text included in the processing target data, a tag indicating the evaluation target, a topic tag indicating a topic, a mirror tag indicating expected contents for the evaluation target, a thermo tag indicating a reaction of a consumer, and the evaluation Tag risk tags that indicate the risks associated with the target,
In the processing target data, the first text tagged with the combination of the tag indicating the evaluation target, the topic tag, and the mirror tag, and the tag combining the tag indicating the evaluation target, the topic tag, and the thermo tag Causing the index value to be calculated for at least one of the attached second text and the third text tagged with the combination of the tag indicating the evaluation target, the topic tag, and the risk tag,
Evaluation program.

Images