Emotion index value calculation device, emotion index value calculation method, and emotion index value calculation program

Abstract

The present invention provides a device, method, and program for calculating numerical values ​​related to emotions expressed in information that indicates an opinion. [Solution] In a communication system in which an opinion analysis server, one or more opinion collection servers, and multiple analyst terminals are connected to a network, the system control unit of the opinion analysis server 1 acquires multiple opinion information, acquires a polarity value for each of the multiple opinion information that indicates the tendency of the polarity of the emotion expressed in the opinion information and the degree of that tendency, acquires an intensity value for each of the multiple opinion information that indicates the intensity of the emotion in the opinion information, calculates an index value for each of the multiple opinion information by multiplying the polarity value by the intensity value, and transmits screen information to a display device that displays a screen showing at least one of the index value and the statistical value of the index value.

Description

Technical Field

[0001] The present invention relates to a method for calculating a numerical value related to an emotion expressed by information indicating an opinion.

Background Art

[0002] Conventionally, there has been known a technique for analyzing information indicating opinions such as reviews and evaluations and calculating a numerical value indicating the emotion of the information provider. For example, in Patent Document 1, sentiment analysis processing is performed on the text data of a posted article to calculate the positive and negative score of the article, and to obtain label information indicating whether the article on which the sentiment analysis processing has been performed is positive or negative and a score within a predetermined range.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the conventional technique, there are cases where the numerical value calculated as the numerical value indicating the emotion does not change significantly whether the emotion of the information provider is weakly expressed or strongly expressed in the information indicating the opinion. For example, when a numerical value within a predetermined range is calculated, there are cases where there is no significant difference in the calculated numerical value between information including relatively few emotional expressions and information including relatively many emotional expressions.

[0005] The present invention has been made in view of the above points, and an example of the problem is to provide an emotion index value calculation device, an emotion index value calculation method, and an emotion index value calculation program capable of providing numerical value information reflecting the overall strength of the emotion of information indicating an opinion.

Means for Solving the Problems

[0006] One aspect of this application is an opinion information acquisition means for acquiring multiple opinion information, each representing an opinion, and for each of the acquired multiple opinion information, based on the opinion information, the tendency of the polarity of the emotion expressed in the opinion information. In addition to indicating that the polarity of the tendency is either positive or negative, the polarity of either of those The emotion index value calculation device is characterized by comprising: polarity value acquisition means for acquiring polarity values ​​indicating the degree; intensity value acquisition means for acquiring intensity values ​​indicating the intensity of emotion in each of the acquired plurality of opinion information based on the opinion information; index value calculation means for calculating an index value by multiplying the acquired polarity value by the acquired intensity value for each of the acquired plurality of opinion information; and screen information transmission means for transmitting screen information to a display device that displays a screen showing at least one of the calculated index value and an index statistical value which is a statistical value of the index value.

[0007] According to this aspect, a polarity value and an intensity value are calculated for each of the multiple opinion pieces. The polarity value indicates the tendency and degree of the polarity of the emotion expressed in the opinion piece. The intensity value indicates the intensity of the emotion in the opinion piece. An index value is calculated for each of the multiple opinion pieces. The index value is the product of the polarity value and the intensity value. A screen showing at least one of the index value and its statistical value is then displayed on the display device. The index value is influenced by the intensity of the emotion in addition to the tendency and degree of the polarity of the emotion. Therefore, it is possible to provide numerical information that reflects the overall emotional intensity of the opinion piece.

[0008] Another aspect of the present invention is an emotion index value calculation device characterized in that, each time an emotion expression appears in the opinion information, the intensity value is calculated such that the intensity value increases regardless of whether the emotion expression is a positive or negative emotion expression, and the intensity value acquisition means acquires the calculated intensity value.

[0009] According to this aspect, when emotional expressions appear in opinion information, the intensity value increases. The intensity value increases whether the emotional expression is positive or negative. Therefore, the more emotional expressions included in opinion information, the higher the polarity value tends to be, and the larger the absolute value of the index tends to be. Thus, it is possible to calculate an index value related to the abundance of emotional expressions included in opinion information.

[0010] Another aspect of the present invention is an emotion index value calculation device characterized in that, when both positive and negative emotions appear in the opinion information, the intensity value is calculated to include both the increase in the intensity value for the positive emotion and the increase in the intensity value for the negative emotion. Another aspect of this application is that each of the multiple pieces of opinion information obtained presents a statement of opinion, The sentiment index value calculation device is characterized in that, when the opinion statement indicated by any of the aforementioned plurality of opinion information includes multiple sentences, the polarity value acquisition means acquires the polarity value corresponding to the entire opinion statement including the multiple sentences, the intensity value acquisition means acquires the intensity value corresponding to the entire opinion statement including the multiple sentences, and the index value calculation means calculates the product of the polarity value corresponding to the entire opinion statement and the intensity value corresponding to the entire opinion statement as the index value corresponding to the entire opinion statement. Another aspect of the present invention is an emotion index value calculation device characterized in that each of the acquired plurality of opinion information is associated with one of a predetermined plurality of items, and for each of the plurality of items, an index statistical value calculation means is provided for calculating the index statistical value of the index value calculated for the opinion information associated with the item, and the screen information transmission means transmits screen information to display on the screen, relating item information indicating the item and index statistical value information indicating the index statistical value calculated for the item for each of the plurality of items.

[0011] According to this approach, for each of several predetermined items, a statistical value of the index calculated from the opinion information associated with that item is calculated as an index statistic. Then, for each of the multiple items, the information representing that item and the information representing the calculated index statistic for that item are displayed on the screen in association. Therefore, it is possible to compare the trends of the index values ​​across items.

[0012] Another aspect of the present invention is a sentiment index value calculation device characterized in that the screen information transmission means can accept an operation on the screen to select any of the multiple index statistics values ​​shown in the multiple index statistics value information displayed on the screen, and when any of the index statistics values ​​is selected, transmits screen information to display on the screen the opinion information associated with the item for which the selected index statistics value was calculated and the index value calculated for the opinion information in association with each other.

[0013] This aspect allows users to select one of several indicator statistics shown on the screen for each of multiple items. Once an indicator statistic is selected, the opinion information associated with the item for which the selected indicator statistic was calculated, and the indicator value calculated for that opinion information, are displayed in relation to each other. This makes it possible to compare the trends of indicator values ​​across items and to review the opinion information associated with a desired item.

[0014] Another aspect of the present invention is an emotion index value calculation device characterized in that the aforementioned opinion information indicates the opinion of an evaluator of an item to be evaluated, and for each of the plurality of opinion information obtained, an evaluation value acquisition means is provided to acquire an evaluation value indicating the level of the evaluator's evaluation of the item, and for each of the plurality of items, an evaluation statistics value is provided to calculate an evaluation statistics value which is a statistical value of the evaluation value obtained for the opinion information associated with the item, and the screen information transmission means transmits screen information which further displays a screen showing evaluation statistics value information indicating the evaluation statistics value calculated for each of the plurality of items, in association with the item information indicating the item.

[0015] According to this approach, in addition to opinion information, evaluation values ​​indicating the level of evaluation for the subject being evaluated are obtained. For each of the predetermined items, a statistical value of the evaluation value obtained for the opinion information associated with that item is calculated as an evaluation statistic. Then, for each of the items, the information representing that item, the information representing the calculated index statistics for that item, and the information representing the calculated evaluation statistics for that item are linked and displayed on the screen. Therefore, it is possible to compare the trends of index values ​​and evaluation values ​​across items.

[0016] Another aspect of the present invention is a sentiment index value calculation device characterized in that each of the acquired opinion information is associated with one of a predetermined set of items, and for each of the set of items, a statistical value calculation means is provided to calculate statistical values ​​of the polarity value and the intensity value calculated for the opinion information associated with the item, and the screen information transmission means displays on the screen an index information that indicates an index corresponding to each of the set of items, and that indicates the statistical values ​​of the polarity value and the intensity value calculated for the item at the display position of the index information, and transmits the screen information that allows the user to select one of the displayed index information, and when one of the index information is selected, the sentiment index value calculation device displays on the screen an index information that is associated with the item to which the selected index information corresponds from the set of items, and the index value calculated for the opinion information.

[0017] According to this aspect, the screen displays mark information corresponding to each of multiple items. The display position of each mark information corresponds to the statistical values ​​of polarity and intensity calculated for the opinion information associated with the item to which the mark information corresponds. When any mark information is selected from the screen, the opinion information associated with the item to which the mark information corresponds, and the index value calculated for that opinion information are displayed in association. Therefore, by looking at the display position of the mark information, it is possible to grasp the trends of the polarity and intensity values ​​of the opinion information associated with each item, and to check the opinion information and index value associated with the desired item.

[0018] Another aspect of the present invention relates to a computer-based method for calculating emotion index values, comprising: an opinion information acquisition step of acquiring multiple opinion information, each representing an opinion; and, for each of the acquired multiple opinion information, the tendency of the polarity of the emotion expressed in the opinion information, based on the opinion information. In addition to indicating that the polarity of the tendency is either positive or negative, the polarity of either of those The method for calculating an emotion index value is characterized by including: a polarity value acquisition step of acquiring a polarity value indicating the degree; an intensity value acquisition step of acquiring an intensity value indicating the intensity of emotion in the opinion information based on the opinion information for each of the acquired plurality of opinion information; an index value calculation step of calculating an index value by multiplying the acquired polarity value by the acquired intensity value for each of the acquired plurality of opinion information; and a screen information transmission step of transmitting screen information to a display device that displays a screen showing at least one of the calculated index value and an index statistical value which is a statistical value of the index value.

[0019] Another aspect of this invention is a computer that provides an opinion information acquisition means for acquiring multiple opinion information, each representing an opinion, and for each of the acquired multiple opinion information, based on the opinion information, the tendency of the polarity of the emotion expressed in the opinion information. In addition to indicating that the polarity of the tendency is either positive or negative, the polarity of either of thoseA polarity value acquisition means for acquiring a polarity value indicating a degree, a strength value acquisition means for acquiring, for each of the plurality of pieces of opinion information acquired, a strength value indicating the intensity of the emotion in the opinion information based on the opinion information, an index value calculation means for calculating an index value by multiplying the acquired polarity value by the acquired strength value for each of the plurality of pieces of opinion information acquired, and a screen information transmission means for transmitting screen information for causing a display device to display a screen indicating at least one of the calculated index value and an index statistical value which is a statistical value of the index value. A sentiment index value calculation program characterized by functioning as.

Effect of the Invention

[0020] According to the present invention, it is possible to provide numerical information reflecting the overall emotional strength of information indicating opinions.

Brief Description of the Drawings

[0021] [Figure 1] It is a diagram showing an example of the schematic configuration of the communication system S according to the present embodiment. [Figure 2] It is a block diagram showing an example of the schematic configuration of the opinion analysis server 1 according to the present embodiment. [Figure 3] It is a diagram showing an example of the functional block of the system control unit 11 of the opinion analysis server 1 according to the present embodiment. [Figure 4] It is a diagram showing an example of the difference in numerical values of emotions. [Figure 5] (a) is a diagram showing an example of the calculation of the sentiment index value. (b) is a diagram showing an example of opinion information and the calculated numerical values. [Figure 6] It is a diagram showing an example of an analysis screen. [Figure 7] It is a diagram showing an example of an analysis screen when selecting the sentiment index statistical value. [Figure 8] It is a diagram showing an example of the analysis screen information transmission process executed by the system control unit 11 of the opinion analysis server 1 according to the present embodiment. [Figure 9]This figure shows an example of the review information transmission process executed by the system control unit 11 of the opinion analysis server 1 according to this embodiment. [Figure 10] This figure shows an example of an analysis screen. [Figure 11] This figure shows an example of the analysis screen when selecting mark information corresponding to an item. [Figure 12] This figure shows an example of the analysis screen information transmission process executed by the system control unit 11 of the opinion analysis server 1 according to this embodiment. [Modes for carrying out the invention]

[0022] [1. First Embodiment] A first embodiment of the present invention will be described in detail below with reference to the drawings.

[0023] [1-1. Communication System Configuration] First, the configuration and functional overview of the communication system S according to this embodiment will be explained with reference to Figure 1. Figure 1 is a diagram showing an example of the overview configuration of the communication system S according to this embodiment.

[0024] As shown in Figure 1, the communication system S may consist of an opinion analysis server 1, one or more opinion collection servers 2, and multiple analyst terminals 3. These devices are connected to a network NW. The network NW is constructed using, for example, the internet, a dedicated communication line (e.g., a CATV (Community Antenna Television) line), a mobile communication network (including base stations, etc.), and a gateway.

[0025] Opinion analysis server 1 may be a server device that analyzes opinions from multiple opinion providers. Opinion analysis server 1 calculates an emotional index value as a result of opinion analysis and provides information on the emotional index value to analyst terminal 3. The emotional index value is a numerical value that serves as an indicator of the opinion provider's emotions, and may be a numerical value of the emotions expressed in that opinion. Each opinion provider provides opinion information indicating their opinion on the subject of the opinion. The subject of the opinion may indicate what the opinion is about. Examples of subjects of opinion include products, services, shops, service providers, people, staff, companies, places, facilities, websites, applications, functions, programs, movies, books, music, content, events, news, etc. The subject of opinion that is being evaluated is also called the subject of evaluation. Examples of opinion providers include users of the subject of opinion, purchasers of the subject of opinion, viewers of the subject of opinion, respondents to surveys about the subject of opinion, etc. Examples of opinions from opinion providers include impressions, evaluations, reviews, word of mouth, complaints, inquiries, tweets, feedback, etc. An opinion provider who provides an evaluation as an opinion on the subject of evaluation is also called an evaluator. Furthermore, the opinion information and other information that opinionators provide as customers about the subject of their opinions are also called VOC (Voice of Customer). Examples of subjects for which VOC opinions are collected include products, services, stores, companies, websites, etc.

[0026] Opinion information, which expresses an opinion, may also be information that can express the feelings of the person expressing the opinion. Opinion information may also be information that represents sentences or texts, for example. One example of opinion information is text data. Examples of text data include text data and document data. Other examples of opinion information include audio data that represents voice and video data that represents video. Audio data or video data may be converted into opinion information as text data by speech recognition.

[0027] The opinion collection server 2 may be a server device that collects opinion information from opinion providers. For example, the opinion collection server 2 may receive opinion information from an opinion provider's terminal device (not shown) and store the received opinion information in a predetermined database. The opinion collection server 2 may also be a server device that performs processing related to the subject of the opinion. For example, suppose the opinion collection server 2 is a device that performs processing related to a website that enables e-commerce. In this case, examples of the subject of the opinion include goods, services, stores, service providers, and website operators. Alternatively, the opinion collection server 2 may be a server device that performs processing related to SNS (Social Network Service). For example, the opinion collection server 2 may collect posts on SNS as opinion information.

[0028] Each analyst terminal 3 may be a terminal device used by an analyst who utilizes the opinion analysis server 1. The analyst terminal 3 may receive information, including sentiment index values, as the result of opinion analysis from the opinion analysis server 1 and display that information on the screen. The analyst may be, for example, an employee of a business that analyzes opinions from opinion givers. Alternatively, the analyst may be an employee of a business that can collect opinion information through a website or the like. Alternatively, the analyst may be a provider of the product, service, program, etc. that is the subject of the opinion. Alternatively, the analyst may be an individual. Examples of analyst terminals 3 include personal computers, tablet computers, smartphones, etc. A web browser may be pre-installed on each analyst terminal 3.

[0029] [1-2. Configuration of the Opinion Analysis Server] Next, the configuration of the opinion analysis server 1 will be described with reference to Figure 2. Figure 2 is a block diagram showing an example of the overview configuration of the opinion analysis server 1 according to this embodiment. As shown in Figure 2, the opinion analysis server 1 comprises a system control unit 11, a system bus 12, an input / output interface 13, a storage unit 14, and a communication unit 15. The system control unit 11 and the input / output interface 13 are connected via the system bus 12.

[0030] The system control unit 11 is composed of a CPU (Central Processing Unit) 11a, a ROM (Read Only Memory) 11b, a RAM (Random Access Memory) 11c, and the like.

[0031] The input / output interface 13 performs interface processing between the storage unit 14 and the communication unit 15 and the system control unit 11.

[0032] The storage unit 14 is composed of, for example, a hard disk drive. This storage unit 14 stores an opinion analysis program. The opinion analysis program may be a program that causes the system control unit 11 to perform a process of analyzing opinion information. The opinion analysis program may be acquired from, for example, another device via a network NW, or it may be recorded on a recording medium such as magnetic tape, optical disc, or memory card and read via a drive device.

[0033] The communication unit 15 is composed of, for example, a network interface card. The communication unit 15 connects to a device different from the opinion analysis server 1 via the network NW and controls the communication status with the connected device.

[0034] [1-3. Overview of System Control Unit Functions] Next, the functional overview of the system control unit 11 of the opinion analysis server 1 will be described with reference to Figures 3 to 7. Figure 3 is a diagram showing an example of the functional blocks of the system control unit 11 of the opinion analysis server 1 according to this embodiment. The system control unit 11 may function as an opinion information acquisition unit 1101, a polarity value acquisition unit 1102, an intensity value acquisition unit 1103, an index value calculation unit 1104, a statistical value calculation unit 1105, and a screen information transmission unit 1106, as shown in Figure 3, by the CPU 11a reading and executing various program codes included in the opinion analysis program.

[0035] The opinion information acquisition unit 1101 may acquire multiple opinion information. The storage unit 14 of the opinion analysis server 1 may store multiple opinion information. For example, the opinion analysis server 1 may acquire multiple opinion information from the opinion collection server 2 or the analyst terminal 3 and store the acquired opinion information in the storage unit 14 in advance. In this case, the opinion information acquisition unit 1101 may acquire opinion information from the storage unit 14. Alternatively, the opinion information acquisition unit 1101 may acquire multiple opinion information from the opinion collection server 2 or the analyst terminal 3. For example, the opinion information acquisition unit 1101 may acquire information such as the opinion collection server 2 that is the source of the opinion information and its URL (Uniform Resource Locator) from the analyst terminal 3. Based on this information, the opinion information acquisition unit 1101 may acquire opinion information from the opinion collection server 2. Also, for example, the analyst terminal 3 may upload opinion information to the opinion analysis server 1.

[0036] Each of the multiple opinion pieces acquired by the opinion information acquisition unit 1101 may be information associated with one of a predetermined set of items. These items may be related to the opinion piece or the opinion itself, or they may be related to the subject of the opinion. For example, each opinion piece acquired by the opinion information acquisition unit 1101 may have item identification information attached to it to identify the item. Alternatively, the opinion information acquisition unit 1101 may associate each opinion piece with one of the items based on the opinion piece. Alternatively, the opinion piece may be grouped by item. For example, the opinion information acquisition unit 1101 may acquire an archive file containing the opinion piece for each item. In addition, the storage unit 14 may store the opinion piece in a different directory for each item. Examples of items include the period during which the opinion piece was sent from the opinion user's terminal device to the opinion collection server 2, the subject of the opinion, words included in the opinion piece, the subject of the opinion (or the subject of evaluation), etc. Examples of the period during which the opinion piece was sent include the period during which the opinion was submitted by the opinion user, the period during which a complaint was received from the opinion user, the period during which an inquiry was made from the opinion user, etc. Examples of time periods include time zones, dates, weeks, months, and years. When the opinion collection server 2 stores opinion information in the database, it may store the opinion information in association with transmission time information. The transmission time information may indicate the time when the opinion information was sent to the opinion collection server 2. In particular, the transmission time information may indicate the date and time or date when the opinion information was sent. Each opinion information acquired by the opinion information acquisition unit 1101 may have transmission time information attached as item identification information. By referring to the transmission time information, it is possible to identify which time period the opinion information is associated with. Alternatively, each opinion information acquired by the opinion information acquisition unit 1101 may have time period information attached that indicates the period during which the opinion information was sent. The subject of the opinion may indicate what the opinion is mainly about. For example, suppose a user who has used a store in an online shopping mall posts a review of that store. In this case, the commenter is the user, the store is the subject of the opinion, and the review is the opinion. Examples of the subject of the review in this case include products, delivery, product range, price, etc.For example, an opinion giver may select one of several predetermined themes and send opinion information to the opinion collection server 2. In this case, the opinion collection server 2 may associate the opinion information with theme information indicating the selected theme and store it in a database. The opinion information acquisition unit 1101 may attach theme information as item identification information to each opinion information acquired by the opinion information acquisition unit 1101. Alternatively, the opinion information acquisition unit 1101 may estimate the theme of the opinion information by analyzing the opinion information. If the item associated with the opinion information is a word, one opinion information may be associated with multiple words. If the item associated with the opinion information is an opinion subject, the opinion information acquisition unit 1101 will acquire one or more opinion information indicating opinions on each of several different opinion subjects.

[0037] If the subject of the opinion is an evaluation subject that is evaluated by the opinion giver, the opinion information acquisition unit 1101 may or may not acquire an evaluation value for each of the multiple opinion pieces acquired, indicating the level of the opinion giver's evaluation of that evaluation subject. Examples of evaluation values ​​include a numerical value indicating whether the evaluation subject is good or bad, satisfaction with the evaluation subject, understanding of the evaluation subject, NPS (Net Promoter Score), etc. For example, the opinion giver inputs opinion information and an evaluation value. The evaluation value may be input within a predetermined range of numbers. For example, the evaluation value may be input on a 5-point scale from 1 to 5. The opinion giver's terminal device transmits the opinion information and evaluation value to the opinion collection server 2. The opinion collection server 2 may associate the received opinion information and evaluation value and store them in a database. Each opinion piece acquired by the opinion information acquisition unit 1101 may have an evaluation value attached to it. The opinion information acquisition unit 1101 may acquire an evaluation value by acquiring opinion information to which an evaluation value has been attached.

[0038] The polarity value acquisition unit 1102 may acquire an emotional polarity value for each of the multiple opinion information acquired by the opinion information acquisition unit 1101, based on that opinion information. The emotional polarity value may be a numerical value indicating the tendency of the emotional polarity expressed in the opinion information and the degree of that tendency. The emotional polarity may indicate whether the opinion giver's emotions are positive or negative. The tendency of the emotional polarity may indicate whether the opinion giver's emotions tend towards positive or negative polarity. The degree of the tendency of the emotional polarity may indicate how high the tendency of that polarity is.

[0039] Figure 4 shows an example of differences in emotional values. The emotional polarity value may be a number determined within a predetermined range centered around 0. In other words, the emotional polarity value may be a normalized number. For example, the emotional polarity value may be a number within the range of -1 to +1. An emotional polarity value of 0 may indicate that the opinionator's emotions are in a neutral state. A positive emotional polarity value may indicate that the opinionator's emotions tend to be positive. The closer the positive emotional polarity value is to the maximum value of +1, the stronger the positive tendency may be. A negative emotional polarity value may indicate that the opinionator's emotions tend to be negative. The closer the negative emotional polarity value is to the minimum value of -1, the stronger the negative tendency may be. As shown in Figure 4, the more relatively positive emotional expressions there are in the opinion information, the higher the emotional polarity value tends to be. On the other hand, the more relatively negative emotional expressions there are in the opinion information, the lower the emotional polarity value tends to be.

[0040] The polarity value acquisition unit 1102 may, for example, acquire emotion polarity values ​​from an external system of the opinion analysis server 1 by using an API (Application Programming Interface) provided by that system. In this case, the polarity value acquisition unit 1102 transmits the opinion information to be used for calculating emotion polarity values ​​to that system. The system calculates emotion polarity values ​​based on the received opinion information. Alternatively, the polarity value acquisition unit 1102 may acquire emotion polarity values ​​by calculating them itself. Examples of emotion polarity calculation algorithms include Naive Bayes, SVM, BERT, SentiWordNet, rule-based algorithms, etc. An example of an emotion polarity calculation method is shown below. The entity that calculates emotion polarity values ​​is the emotion polarity value calculation unit. This emotion polarity value calculation unit may be the polarity value acquisition unit 1102 or the system providing the aforementioned API. The emotion polarity value calculation unit subdivides the opinion information into sentences and words by morphological analysis. The emotion polarity value calculation unit determines whether each word is a word that indicates a positive emotion or a word that indicates a negative emotion. For example, a database serving as a word dictionary may have positive and negative words pre-registered. The emotion polarity calculation unit may make a determination by referring to this database. Positive score, negative score, and neutral score may be set as parameters for calculating the emotion polarity value. The initial value of each score may be 0. If the emotion polarity calculation unit finds a sentence containing positive words, it increases the positive score. If the emotion polarity calculation unit finds a sentence containing negative words from the opinion information, it increases the negative score. If the emotion polarity calculation unit finds a sentence from the opinion information that does not contain either positive or negative words, it increases the neutral score. If a sentence containing both positive and negative words is found, the emotion polarity calculation unit may determine which score to increase based on the number of positive words, the number of negative words, the degree of positivity of the words, the degree of negativity of the words, etc. The degree of positivity and the degree of negativity of the words may be stored in the word dictionary database.The amount of score increase may be predetermined. Alternatively, the amount of score increase may be determined by the length of the sentence, the number of positive words, the number of negative words, the degree of positivity of the words, the degree of negativity of the words, etc. The amount of score increase may be determined within a predetermined range, or not. For example, the amount of score increase may be determined within a range from 0 to 1. Once processing is complete for all sentences included in the opinion information, the sentiment polarity calculation unit may normalize each score. For example, the sentiment polarity calculation unit calculates the sum of the positive score, negative score, and neutral score. The sentiment polarity calculation unit may normalize each score by dividing it by its sum. The sentiment polarity calculation unit may calculate the sentiment polarity value by subtracting the normalized negative score from the normalized positive score.

[0041] The intensity value acquisition unit 1103 may acquire an emotion intensity value for each of the multiple opinion information acquired by the opinion information acquisition unit 1101, based on that opinion information. The emotion intensity value may be a numerical value indicating the intensity of emotion in that opinion information. The emotion intensity value may also indicate the overall emotion intensity of the opinion information.

[0042] As shown in Figure 4, the emotion intensity value may be a numerical value determined within the range of 0 to infinity. The higher the emotion intensity value, the stronger the emotion. The more emotional expressions there are in the opinion information, the higher the emotion intensity value. Here, the emotion intensity value increases regardless of whether the emotional expressions found are positive or negative. Therefore, the longer the sentences included in the opinion information, the higher the emotion intensity value tends to be. Here, if the emotion polarity value is positive, the opinion information tends to be positive. If positive expressions appear in this opinion information, the emotion intensity value increases, and even if negative expressions appear, the emotion intensity value increases. If the emotion polarity value is negative, the opinion information tends to be negative. If negative expressions appear in this opinion information, the emotion intensity value increases, and even if positive expressions appear, the emotion intensity value increases.

[0043] The intensity value acquisition unit 1103 may, for example, acquire emotion intensity values ​​from an external system of the opinion analysis server 1 by using an API provided by that system. In this case, the intensity value acquisition unit 1103 transmits the opinion information to be used for calculating emotion intensity values ​​to that system. The system calculates emotion intensity values ​​based on the received opinion information. Alternatively, the intensity value acquisition unit 1103 may acquire emotion intensity values ​​by calculating them itself. Similar to the case of emotion polarity values, emotion intensity values ​​can be calculated using algorithms such as Naive Bayes, SVM, BERT, SentiWordNet, or rule-based algorithms. An example of a method for calculating emotion intensity values ​​is shown below. The entity that calculates emotion intensity values ​​is the emotion intensity value calculation unit. This emotion intensity value calculation unit may be the intensity value acquisition unit 1103 or the system providing the aforementioned API. The emotion intensity value calculation unit may calculate emotion intensity values ​​such that the emotion intensity value increases (or rises) each time an emotion expression appears in the opinion information, regardless of whether the emotion expression is a positive or negative emotion expression. The intensity value acquisition unit 1103 may acquire the emotion intensity value calculated in this way. The emotion polarity value calculation unit subdivides the opinion information into sentences and words by morphological analysis. The emotion intensity value calculation unit initializes the emotion intensity value to 0. The emotion polarity value calculation unit searches the opinion information for words that indicate positive emotions and words that indicate negative emotions. The emotion polarity value calculation unit then determines whether each sentence included in the opinion information contains at least one of either positive words or negative words. The emotion polarity value calculation unit increases the emotion polarity value each time it finds a sentence that contains at least one of either positive words or negative words. The emotion polarity value increases whether a sentence containing positive words or a sentence containing negative words is found. The emotion polarity value calculation unit does not increase the emotion polarity value if it finds a sentence that does not contain either positive or negative words. The amount of increase in the emotion intensity value may be predetermined. Alternatively, the increase in emotional intensity may be determined by the length of the sentence, the number of positive and negative words, the degree of positivity of the words, the degree of negativity of the words, etc.The amount of increase in the emotion intensity value may be determined within a predetermined range, or it may not. For example, the amount of increase in the emotion intensity value may be determined within the range of 0 to 1. Once processing is complete for all sentences included in the opinion information, the calculation of the emotion intensity value is complete. Note that the emotion polarity value calculation unit does not necessarily need to increase the emotion intensity value at the moment an emotion expression is found. It is sufficient if the emotion intensity value increases in the end. Also, the emotion polarity value calculation unit may increase the emotion intensity value each time an emotion-indicating word is found, rather than increasing the emotion intensity value for each sentence.

[0044] The index value calculation unit 1104 may calculate an emotional index value for each of the multiple opinion pieces obtained by the opinion information acquisition unit 1101 by multiplying the emotional polarity value obtained by the polarity value acquisition unit 1102 by the emotional intensity value obtained by the intensity value acquisition unit 1103. Figure 5(a) shows an example of calculating an emotional index value. Here, the emotional index value calculated for the opinion piece included in the VOC is specifically called the VOC score. In general opinion analysis, emotional polarity values ​​or the normalized score described above are used as indicators of emotion. The closer the emotional polarity value calculated for an opinion piece is to -1, the more important that opinion piece tends to be, and the closer the emotional polarity value is to +1, the more important that opinion piece tends to be. In contrast, the emotional intensity value indicates the overall emotional intensity in the opinion piece. The higher the emotional intensity value calculated for an opinion piece, the more emotions are contained in that opinion piece. Therefore, the higher the emotional intensity value, the more important that opinion piece may be. The index value calculation unit 1104 calculates the emotional index value by multiplying the emotional polarity value by the emotional intensity value. Therefore, it is possible to improve the resolution of the indicator compared to general indicators. The larger the absolute value of the sentiment index calculated for opinion information, the more important that opinion information may be.

[0045] Figure 5(b) shows an example of opinion information and the calculated values. As shown in Figure 5(b), let's assume the opinion information is the sentence "It's okay." This is a neutral sentence that does not contain any positive or negative words. Therefore, the emotion polarity value, emotion intensity value, and emotion index value are all 0.0.

[0046] Let's assume the opinion is, "The delivery was fast, which is good. However, the app is a little difficult to use, which is a shame." The first sentence contains the positive word "good." The second sentence contains the negative words "difficult to use" and "a shame." In this case, the positive and negative cancel each other out, resulting in an emotion polarity value of 0.0. On the other hand, because there are two sentences containing emotional expressions, the emotion intensity value is 1.8. However, the emotion index value is 0.0.

[0047] Assume the opinion information is the following sentence: "It's convenient because it's compact and doesn't take up much space. I also like it because it's easy to use. However, the instruction manual was difficult to understand." The first sentence contains the positive word "convenient." The second sentence contains the positive word "I like it." The third sentence contains the negative word "difficult to understand." In this case, the emotional polarity value is 0.4. Since there are three sentences containing emotional expressions, the emotional intensity value is 2.6. Therefore, the emotional index value is 1.04.

[0048] Let's assume the opinion information is: "The initial setup was difficult and a bit troublesome. However, it's multi-functional so it seems like it will be convenient to use. Once you've set it up, it's hassle-free and easy to use. I'm very satisfied with this price. However, it will probably take some time to remember all the functions." The first sentence contains the negative word "troublesome". The second sentence contains the positive word "convenient". The third sentence contains the positive word "easy". The fourth sentence contains the positive word "very satisfied". The fifth sentence contains the negative word "took time". In this case, the emotion polarity value is 0.4. Since there are five sentences containing emotional expressions, the emotion intensity value is 4.3. Therefore, the emotion index value is 1.72. As shown in this example and the previous example, even with the same emotion polarity value, the emotion index value can be different.

[0049] Let's assume the opinion statement is "difficult to use." This statement contains negative words. In this case, the emotional polarity value is 0.9. Since there is one sentence containing an emotional expression, the emotional intensity value is 0.9. Therefore, the emotional index value is -0.81.

[0050] Let's assume the opinion is, "It malfunctioned less than a month after I received it, which was troublesome. I contacted the support center, but their response was poor." The first sentence contains the negative words "malfunctioned" and "troublesome." The second sentence contains the negative word "bad." In this case, the emotional polarity value was 0.9. Since there are two sentences containing emotional expressions, the emotional intensity value became 1.8. Therefore, the emotional index value is -1.44.

[0051] The statistical value calculation unit 1105 may calculate statistical values ​​of the sentiment index values ​​calculated by the index value calculation unit 1104. These statistical values ​​may be representative values. Examples of statistical values ​​include the mean, median, maximum, minimum, and mode. The statistical values ​​of the sentiment index are called sentiment index statistics. The statistical value calculation unit 1105 may calculate sentiment index statistics for the entire set of opinion information acquired by the opinion information acquisition unit 1101. If each opinion information is associated with one of several items, the statistical value calculation unit 1105 may calculate sentiment index statistics for each item. That is, for each of the multiple items, the statistical value calculation unit 1105 may calculate index statistics for the index values ​​calculated for the opinion information associated with that item.

[0052] Furthermore, the statistical value calculation unit 1105 may also calculate statistical values ​​of evaluation values ​​if evaluation values ​​are obtained by the opinion information acquisition unit 1101. The type of statistical value for evaluation values ​​and the type of statistical value for sentiment indicators may be the same or different. The statistical value calculation unit 1105 may calculate statistical values ​​of evaluation values ​​for the entire set of opinion information obtained by the opinion information acquisition unit 1101. The statistical value calculation unit 1105 may also calculate statistical values ​​of evaluation values ​​for each item. That is, for each of the multiple items, the statistical value calculation unit 1105 may calculate index statistical values ​​of the index values ​​obtained for the opinion information associated with that item.

[0053] The screen information transmission unit 1106 may transmit screen information to a predetermined display device that displays a screen showing at least one of the emotion index value calculated by the index value calculation unit 1104 and the statistical value of that emotion index value. This screen may be an analysis screen showing the results of the opinion information analysis. The screen information may be information that shows a specific screen or information that causes the display device to display a specific screen. The screen information may be, for example, an HTML (HyperText Markup Language) document. In this case, the screen can be replaced with a web page or a page. Alternatively, the screen information may be information about the content or each element displayed on the display device. Alternatively, the screen information may be information that instructs the display device to display a specific screen. The display device is a device capable of displaying any screen. The display device may be, for example, the analyst terminal 3.

[0054] The screen may display at least one of the following: emotional index value information showing emotional index values, or emotional index statistical value information showing emotional index statistics. The emotional index value information or emotional index statistical value information may be the emotional index value or statistics themselves as numerical values. Alternatively, the emotional index value information or emotional index statistical value information may be information that indicates the emotional index value or statistics by its display position on the screen. An example of this information is a graph. The type of graph is not particularly limited. Alternatively, the emotional index value information or emotional index statistical value information may be a symbol, mark, or image that indicates the emotional index value or statistics. Alternatively, the emotional index value information or emotional index statistical value information may indicate the emotional index value or statistics by the color or size of the information.

[0055] The screen information transmission unit 1106 may display at least one of the multiple opinion information obtained by the opinion information acquisition unit 1101 on the screen. Here, the screen information transmission unit 1106 may transmit screen information to display the opinion information, as well as information indicating at least one of the following: the emotion index value, emotion polarity value, and emotion intensity value calculated based on the opinion information, the evaluation value obtained for the opinion information, and information associated with the opinion information. When displaying multiple opinion information, the screen information transmission unit 1106 may display the opinion information in descending order of emotion index value, descending order of emotion index value, or descending order of the absolute value of the emotion index value. This makes it easier for the analyst to review the opinion information, prioritizing opinion information with stronger positive or negative emotions.

[0056] When the statistical value calculation unit 1105 calculates sentiment index statistics for each of several items, the screen information transmission unit 1106 may transmit screen information to display on the screen each of those items, associating the item information representing that item with the sentiment index statistics information representing the sentiment index statistics calculated for that item. The statistical value calculation unit 1105 may transmit screen information to the analyst terminal 3 to display such a screen. The item information may be, for example, the name of the item. For example, the item information and the sentiment index statistics information may be associated by being displayed in the same area. Alternatively, the display position of the item information and the display position of the sentiment index statistics information may indicate that the information is associated.

[0057] When the statistical value calculation unit 1105 calculates the statistical value of the evaluation value, the screen information transmission unit 1106 may transmit screen information to the analyst terminal 3 that displays a screen showing the sentiment index statistical value and the evaluation value statistical value. In this case, in addition to the sentiment index statistical value information, the screen may also display evaluation statistical value information showing the evaluation value. The evaluation statistical value information may be the evaluation value itself. Alternatively, the evaluation statistical value information may be information that indicates the evaluation value by the display position of the evaluation statistical value information on the screen. Alternatively, the evaluation statistical value information may be a symbol, mark, or image that indicates the evaluation value. Alternatively, the evaluation statistical value information may indicate the evaluation value by the color or size of the information.

[0058] If statistical values ​​for evaluation have been calculated for each of the multiple items, the screen information transmission unit 1106 may transmit screen information to display on the screen, relating the item information representing that item, the emotional index statistical value information representing the emotional index statistical value calculated for that item, and the evaluation statistical value information representing the evaluation statistical value calculated for that item. The statistical value calculation unit 1105 may transmit screen information to the analyst terminal 3 to display such a screen. In this case, the emotional index statistical value information and the evaluation statistical value information may be displayed relating to each other for each item.

[0059] If sentiment index statistical information is displayed on the screen for each of multiple items, the screen information transmission unit 1106 may transmit screen information to the analyst terminal 3 that allows the user to select any of the sentiment index statistical values ​​shown on the screen. Examples of operations for selecting a sentiment index statistical value include selecting the sentiment index statistical information that represents that sentiment index statistical value, selecting the item information displayed in association with that sentiment index statistical information, and selecting the evaluation statistical information displayed in association with that sentiment index statistical information. If any of the sentiment index statistical values ​​are selected by such operations, the screen information transmission unit 1106 may transmit screen information to display the opinion information associated with the item associated with the selected sentiment index statistical value on that screen.

[0060] For example, the screen information may include an event handler corresponding to an event in which the analyst selects a sentiment index statistic. This event handler may be a program that causes the analyst terminal 3 to send item identification information to the opinion analysis server 1 to identify the item associated with the sentiment index statistic selected from the screen. The screen information transmission unit 1106 may send one or more opinion information items associated with the item identified by the item identification information received from the analyst terminal 3 to the analyst terminal 3. This allows the screen information transmission unit 1106 to display the opinion information on the screen. Alternatively, the screen information transmission unit 1106 may send screen information containing all opinion information acquired by the opinion information acquisition unit to the analyst terminal 3. Each opinion information included in the screen information may have item identification information of the item associated with that opinion information attached to it. In this case, the event handler corresponding to the event in which the sentiment index statistic is selected may be a program that causes the analyst terminal 3 to display on the screen the opinion information included in the screen information, with the item identification information of the item associated with the sentiment index statistic selected from the screen attached to it.

[0061] When displaying opinion information associated with an item associated with a selected sentiment index statistic, the screen information transmission unit 1106 may transmit screen information to display the opinion information and at least one of the sentiment index value, sentiment polarity value, sentiment intensity value calculated for that opinion information, and evaluation value obtained for that opinion information, relating them to each other on the screen. When displaying multiple opinion information items, the screen information transmission unit 1106 may display the opinion information in descending order of sentiment index values, descending order of sentiment index values, or descending order of the absolute values ​​of the sentiment index values.

[0062] Figure 6 shows an example of an analysis screen. The example screen shown in Figure 6 is an example where the evaluation target is a store that operates in an online shopping mall, the opinion information is the text of reviews about the store from users who have purchased products from that store, and the item is month. As shown in Figure 6, the analysis screen 100 includes the number of reviews 110, the overall average sentiment index value 120, the overall average rating value 130, the graph area 140, the review list display area 170, etc. The number of reviews 110 indicates the total number of reviews posted by users. The overall average sentiment index value 120 indicates the average sentiment index value as a sentiment index statistic for all posted reviews. The overall average sentiment index value 120 is an example of sentiment index statistic information. The overall average rating value 130 indicates the average rating value as a rating statistic for all posted reviews. The overall average rating value 130 is an example of rating statistic information.

[0063] Graph area 140 is the area where graphs showing the relationship between months and the average sentiment index values ​​and average rating values ​​are displayed. Graph area 140 includes month labels 141, 142, 143, and 144. Each month label contains information indicating the month. For example, month labels 141, 142, 143, and 144 each indicate September, October, November, and December 2023. Each month label is an example of item information. Furthermore, graph 140 includes a sentiment index value graph 150 and a rating value graph 160. The sentiment index value graph 150 is a graph showing the relationship between months and the average sentiment index values. The sentiment index value graph 150 may be a line graph, for example. Marks 151, 152, 153, and 154 are displayed on the sentiment index value graph 150. Each mark contains information indicating the correspondence with an item and the sentiment index value at the display position of the mark. For example, marks 151, 152, 153, and 154 are associated with September, October, November, and December 2023, respectively. Furthermore, the higher a mark is displayed, the higher the average sentiment index value of reviews posted in the month associated with that mark. Each mark is an example of sentiment index statistics. Rating graph 160 is a graph showing the relationship between months and the average rating value. Rating graph 160 may also be a line graph, for example. Similar to sentiment index graph 150, marks are also displayed on rating graph 160. Again, the higher a mark is displayed, the higher the average value included in reviews posted in the month associated with that mark.

[0064] The review list display area 170 may be an area where a list of submitted reviews is displayed. For example, multiple review information 180 may be displayed in the review list display area 170. Review information 180 is information about the submitted review. The analyst can display other review information 180 by operating the scroll bar provided in the review list display area 170. Each review information 180 includes the submission date 181, sentiment index value 182, rating value 183, and review text 184. The submission date 181 indicates the year and month in which the review was submitted. If multiple review information 180 have the same submission year and month, the submission date 181 may be displayed together for all of them. The sentiment index value 182 indicates the sentiment index value calculated based on the review text 184 of that review. The rating value 183 indicates the rating value entered by the user when submitting the review. The review text 184 indicates the opinion information submitted as the body of the review. If the entire review text 184 cannot be displayed in review information 180, the analyst may perform a prescribed operation to display the entire review text 184.

[0065] Furthermore, the analysis screen 100 may display information showing the overall ratio of positive opinions to negative opinions, and the ratio of positive opinions to negative opinions for each item. Additionally, if opinions are categorized by subject, the analysis screen 100 may display information showing the overall category ratio and the category ratio for each item. Finally, the analysis screen 100 may display a word cloud showing each word in a size corresponding to its frequency of occurrence.

[0066] Figure 7 shows an example of the analysis screen when selecting sentiment index statistics. In Figure 7, the same elements as in Figure 6 are given the same symbols. On the analysis screen 100, the analyst selects one of the marks from the graph area 140. For example, as shown in Figure 7, the analyst places the mouse pointer P over mark 153 and clicks the mouse. Mark 153 corresponds to November 2023. In response to this operation, as shown in Figure 7, multiple review information 180 for reviews posted in November 2023 is displayed in the review list display area 170.

[0067] There are various examples of how sentiment index values ​​and sentiment index statistics can be used. For example, if the opinion information consists of inquiries or complaints about the subject of the opinion, opinion information with negative and lower sentiment index values ​​can be processed with higher priority. Also, for example, by referring to opinion information with lower sentiment index values ​​with higher priority, it is possible to determine which aspects of the subject of the opinion should be prioritized for improvement. Also, for example, by referring to opinion information with positive and higher sentiment index values ​​with higher priority, it is possible to improve the subject of the opinion so that those with positive opinions are more satisfied. Furthermore, when sentiment index statistics are calculated for multiple items, it is possible to understand the strengths and weaknesses of the subject of the opinion.

[0068] [1-4. Operation of the communication system] Next, the operation of the communication system S will be described with reference to Figures 8 and 9. The processes shown in these figures are illustrative, and any processes may be executed as long as the objective is achieved. The order of the processes is not limited to the order shown in these figures. Also, at least one of the steps shown in these figures may be omitted. The system control unit 11 of the opinion analysis server 1 may execute the processes shown in Figures 8 and 9 according to the program code included in the opinion analysis program.

[0069] Figure 8 shows an example of the analysis screen information transmission process executed by the system control unit 11 of the opinion analysis server 1 according to this embodiment. For example, each opinion information sent from the opinion collection server 2 or the analyst terminal 3 to the opinion analysis server 1 is stored in a predetermined directory in the storage unit 14. Each opinion information is accompanied by item identification information and an evaluation value. The analyst operates the analyst terminal 3 to specify, for example, the path to the directory from which the opinion information was obtained. The analyst also specifies a list of items for which they want the analysis results to be displayed. Then, the analyst selects to display the analysis screen. The analyst terminal 3 sends an analysis screen information request including the specified path and list to the opinion analysis server 1. The system control unit 11 may execute the analysis screen information transmission process in response to receiving the analysis screen information request.

[0070] As shown in Figure 8, the opinion information acquisition unit 1101 selects one of several items based on the list of items included in the analysis screen information request (step S101). The opinion information acquisition unit 1101 then searches for opinion information with item identification information of the selected item attached from the directory indicated by the path included in the analysis screen information request. Next, the opinion information acquisition unit 1101 acquires one of the retrieved opinion information along with its evaluation value (step S102). Next, the polarity value acquisition unit 1102 acquires an emotional polarity value based on the acquired opinion information (step S103). As mentioned above, the polarity value acquisition unit 1102 may acquire the emotional polarity value from a predetermined system by transmitting the opinion information to that system, or the polarity value acquisition unit 1102 may calculate the emotional polarity value itself. Next, the intensity value acquisition unit 1103 acquires an emotional intensity value based on the acquired opinion information (step S104). As described above, the intensity value acquisition unit 1103 may acquire an emotion intensity value from a predetermined system by transmitting opinion information to that system, or the intensity value acquisition unit 1103 may calculate the emotion intensity value itself. Next, the index value calculation unit 1104 calculates an emotion index value by multiplying the emotion polarity value by the emotion intensity value (step S105). The index value calculation unit 1104 stores the calculated emotion polarity value, emotion intensity value, and emotion index value in the storage unit 14 in association with the acquired opinion information. Next, the opinion information acquisition unit 1101 determines whether or not all opinion information retrieved for the selected item has been acquired (step S106). If there is opinion information that has not yet been acquired (step S106: NO), the process proceeds to step S102. In step S102, the opinion information acquisition unit 1101 acquires one of the opinion information that has not yet been acquired, along with its evaluation value.

[0071] If all opinion information has been acquired (step S106: YES), the process proceeds to step S107. In step S107, the statistical value calculation unit 1105 calculates the average of the sentiment index values ​​calculated for the selected items as the item average sentiment index value. The statistical value calculation unit 1105 performs the calculation by dividing the sum of the sentiment index values ​​calculated for the selected items by the number of opinion information acquired for the selected items. Next, the statistical value calculation unit 1105 calculates the average of the evaluation values ​​acquired for the selected items as the item average evaluation value (step S108). The statistical value calculation unit 1105 performs the calculation by dividing the sum of the evaluation values ​​acquired for the selected items by the number of opinion information acquired for the selected items. The statistical value calculation unit 1105 stores the calculated item average sentiment index value and item average evaluation value in the storage unit 14, associating them with the item identification information of the selected items. Next, the opinion information acquisition unit 1101 determines whether or not all items listed in the list have been selected (step S109). If there are still items that have not been selected (step S109: NO), the process proceeds to step S101. In step S101, the opinion information acquisition unit 1101 selects one of the items that has not yet been selected.

[0072] If all items are selected (step S109: YES), the process proceeds to step S110. In step S110, the statistical value calculation unit 1105 calculates the average of the sentiment index values ​​calculated for all acquired opinion information as the overall average sentiment index value. Next, the statistical value calculation unit 1105 calculates the average of all acquired evaluation values ​​as the overall average evaluation value (step S111). The statistical value calculation unit 1105 stores the calculated overall average sentiment index value and overall average evaluation value in the storage unit 14.

[0073] Next, the screen information transmission unit 1106 generates data for displaying a graph area 140 that includes a graph showing the relationship between each item, the average sentiment index value for each item, and the average evaluation value for each item (step S112). This data may be, for example, data written in HTML. At this time, the screen information transmission unit 1106 may add label information indicating each item to the data for display. The screen information transmission unit 1106 also sets the coordinates for displaying the marks for each item's average sentiment index value and each item's average evaluation value in the data, based on the numerical value and the item to which the mark is associated. Next, the screen information transmission unit 1106 generates data for a list of review information 180 to be displayed in the review list display area 170, which includes review information 180 for all opinion information (step S113). This data may also be data written in an HTML document. The screen information transmission unit 1106 may generate data that includes opinion information, sentiment index value, evaluation value, and item name, so that each review information 180 includes such information. Next, the screen information transmission unit 1106 generates an HTML document as an example of screen information for the analysis screen (step S114). At this time, the screen information transmission unit 1106 may embed the overall average sentiment index value and the overall average evaluation value into this HTML document. The screen information transmission unit 1106 may also embed the data of the graph area 140 into the HTML document. The screen information transmission unit 1106 may also embed the data of the list of review information 180 into the HTML document so that at least a portion of the list of review information 180 is displayed in the review list display area 170. The screen information transmission unit 1106 may also embed an event handler in the HTML document to send a review information request, including item identification information corresponding to the selected mark, to the opinion analysis server 1 when a mark is selected on the sentiment index value graph 150. The screen information transmission unit 1106 sends the generated HTML document to the analyst terminal 3 (step S115). The analysis screen information transmission process ends. Based on this HTML document, the analyst terminal 3 displays an analysis screen 100, such as the one shown in Figure 6.

[0074] Figure 9 shows an example of the review information transmission process executed by the system control unit 11 of the opinion analysis server 1 according to this embodiment. For example, when the analyst terminal 3 is displaying the analysis screen 100, the analyst selects one of the marks displayed on the sentiment index value graph 150. The analyst terminal 3 sends a review information request to the opinion analysis server 1, which includes item identification information corresponding to the selected mark. The system control unit 11 may execute the review information transmission process in response to receiving the review information request.

[0075] As shown in Figure 9, the screen information transmission unit 1106 obtains item identification information from the review information request (step S201). Next, the screen information transmission unit 1106 obtains all opinion information to which the obtained item identification information has been added, and for each of those opinion pieces of information, obtains the sentiment index value and evaluation value associated with that opinion piece of information (step S202). Next, the screen information transmission unit 1106 generates data for displaying a list of review information 180 (step S203). For each of the obtained opinion pieces of information, the screen information transmission unit 1106 generates data containing that information such as the opinion piece of information, the sentiment index value, the evaluation value, and the name of the item identified by the obtained item identification information (step S203). Next, the screen information transmission unit 1106 transmits the generated data to the analyst terminal 3 (step S204), and the review information transmission process ends. Based on this HTML document, the analyst terminal 3 displays a list of review information 180 in the review list display area 170.

[0076] As described above, according to this embodiment, it is possible to provide numerical information that reflects the overall emotional intensity of the information expressing an opinion.

[0077] Here, when an emotional expression appears in opinion information, the emotional intensity value increases whether the emotional expression is positive or negative. In this case, the more emotional expressions included in the opinion information, the higher the emotional intensity value tends to be, and the larger the absolute value of the emotional index also tends to be. Therefore, it is possible to calculate an emotional index value related to the abundance of emotional expressions included in the opinion information.

[0078] Furthermore, the opinion analysis server 1 may calculate sentiment index statistics for the opinion information associated with each item. Alternatively, the opinion analysis server 1 may display on the screen the item information for each item and the sentiment index statistics information that shows the sentiment index statistics calculated for that item, in association with each item. In this case, the trends in sentiment index values ​​can be compared between items.

[0079] Here, the opinion analysis server 1 may accept an operation via the screen to select any of the multiple emotional index statistics shown in the multiple emotional index statistics information displayed on the screen. Furthermore, if any of the emotional index statistics are selected, the opinion analysis server 1 may display opinion information associated with the item for which the selected emotional index statistics were calculated. In this case, it becomes possible to compare the trends of the emotional index values ​​across items and to confirm the opinion information associated with the desired item.

[0080] Furthermore, the opinion analysis server 1 may acquire evaluation values. The opinion analysis server 1 may also calculate evaluation statistics for each of the multiple items. Additionally, the opinion analysis server 1 may display a screen that further shows evaluation statistics information associated with each of the multiple items. In this case, trends in sentiment index values ​​and evaluation values ​​can be compared across items.

[0081] [2. Second Embodiment] Next, a second embodiment will be described. Except for the points described below, this embodiment may be identical to the first embodiment.

[0082] [2-1. Overview of System Control Unit Functions] First, the functional overview of the system control unit 11 of the opinion analysis server 1 will be explained with reference to Figures 10 and 11. In this embodiment, each of the multiple opinion information acquired by the opinion information acquisition unit 1101 is associated with one of a predetermined set of items. The association between opinion information and items is explained in the first embodiment.

[0083] The statistical value calculation unit 1105 may calculate statistical values ​​for emotion polarity and emotion intensity for each of the multiple items, based on the opinion information associated with that item. Examples of the types of statistical values ​​are described in the first embodiment. The statistical value of emotion polarity is called the emotion polarity statistic. The statistical value of emotion intensity is called the emotion intensity statistic.

[0084] The screen information transmission unit 1106 may transmit screen information to the analyst terminal 3, such as an analysis screen, which displays mark information indicating a mark corresponding to each of multiple items. Specifically, the screen information transmission unit 1106 may display mark information on the screen that indicates the emotional polarity statistics and emotional intensity statistics calculated for each item at the display position of the mark information. For example, the horizontal display position of the mark information may indicate the emotional polarity statistics, and the vertical display position of the mark information may indicate the emotional intensity statistics. Alternatively, the horizontal display position of the mark information may indicate the emotional intensity statistics, and the vertical display position of the mark information may indicate the emotional polarity statistics. The mark information itself may be a point, a figure, a symbol, an image, or the name of the corresponding item. The mark information may include, or may not include, the emotional index statistics calculated for the item to which the mark information corresponds. The mark information may, for example, be information that constitutes a scatter plot showing the relationship between the emotional polarity statistics and the emotional intensity statistics for each item.

[0085] Furthermore, the screen information transmission unit 1106 may transmit screen information to the analyst terminal 3 that displays a screen capable of accepting an operation to select any of the mark information displayed on the screen. If any of the mark information is selected by this operation, the screen information transmission unit 1106 may display on the screen the opinion information associated with the item corresponding to the selected mark information from among a plurality of items, in association with the sentiment index value calculated for that opinion information. For example, as in the first embodiment, the screen information may include an event handler corresponding to the event in which the analyst selects mark information. This event handler may be a program for displaying opinion information associated with the item corresponding to the selected mark information. As in the first embodiment, the screen information transmission unit 1106 may transmit opinion information associated with the item corresponding to the selected mark information to the analyst terminal 3, or the analyst terminal 3 may display opinion information associated with that item from the opinion information already included in the screen information.

[0086] When displaying opinion information associated with an item corresponding to selected mark information, the screen information transmission unit 1106 may transmit screen information to display on the screen, relating the opinion information with at least one of the emotion index value, emotion polarity value, and emotion intensity value calculated for that opinion information, and the evaluation value obtained for that opinion information. When displaying multiple opinion information items, the screen information transmission unit 1106 may display the opinion information in descending order of emotion index value, descending order of emotion index value, or descending order of the absolute value of the emotion index value.

[0087] Figure 10 shows an example of an analysis screen. The example screen shown in Figure 10 is an example where the subject of the opinion is a store that operates in an online shopping mall, the opinion information is a review text from a user who has purchased a product from that store, and the item is the subject of the opinion. Each piece of opinion information may be classified into one of several categories that indicate the subject of the opinion. For example, categories may include product, delivery, product range, price, and others. As shown in Figure 10, the analysis screen 200 includes the number of reviews 210, the overall average sentiment index value 220, a category map 230, and a review list display area 240, etc. The number of reviews 210 indicates the total number of reviews posted by users. The overall average sentiment index value 220 indicates the average value of the VOC score as a sentiment index statistic for all posted reviews. The category map 230 shows the relationship between the average sentiment polarity value and the average sentiment intensity value for each category of the subject of the opinion. The category map 230 includes marks 231, 232, 233, 234, and 235. Marks 231, 232, 233, 234, and 235 are indicator information corresponding to product, delivery, product range, price, and other, respectively. The name of the category to which the mark corresponds may be displayed near each mark. The display position of each mark indicates the average emotional polarity value and the average emotional intensity value of the item to which the mark corresponds. The horizontal axis of category map 230 corresponds to the average emotional polarity value, and the vertical axis corresponds to the average emotional intensity value.

[0088] The review list display area 240 may be an area where a list of submitted reviews is displayed. For example, multiple review information items 250 may be displayed in the review list display area 240. The review information item 250 is information about the submitted review. The analyst can display other review information items 250 by operating the scroll bar provided in the review list display area 240. Each review information item 250 includes the review text 251, the category name 252, and the sentiment index value 253. The review text 251 shows the opinion information submitted as the body of the review. If the entire review text 251 cannot be displayed in the review information item 250, the analyst may perform a predetermined operation to display the entire review text 251. The category name indicates the name of the category to which the review text 251 is classified. The sentiment index value 253 indicates the sentiment index value calculated based on the review text 251.

[0089] Furthermore, the analysis screen 200 may display evaluation statistics and evaluation values ​​for each opinion piece of information. The analysis screen 200 may also display information showing the overall ratio of positive to negative opinions, and the ratio of positive to negative opinions for each item. The analysis screen 200 may also display information showing the number of occurrences of each positive word and each negative word. Furthermore, the analysis screen 200 may display information showing the number of reviews for each category. Finally, the analysis screen 100 may display a word cloud showing each word in a size corresponding to its frequency of occurrence.

[0090] Figure 11 shows an example of the analysis screen when selecting mark information corresponding to an item. In Figure 11, the same elements as in Figure 10 are given the same symbols. On the analysis screen 200, the analyst selects one of the marks from the category map 230. For example, as shown in Figure 11, the analyst places the mouse pointer P over mark 232 and clicks the mouse. Mark 232 corresponds to delivery. In response to this operation, as shown in Figure 11, the review list display area 240 displays multiple review information 250, including the review text 251 related to delivery.

[0091] [2-2. Operation of the communication system] Next, the operation of the communication system S will be explained with reference to Figure 12. Figure 12 is a diagram showing an example of the analysis screen information transmission process executed by the system control unit 11 of the opinion analysis server 1 according to this embodiment. In Figure 12, the same reference numerals are used for the same processes as in Figure 8.

[0092] As shown in Figure 12, the opinion information acquisition unit 1101 selects one of several items (step S101). The opinion information acquisition unit 1101 searches for opinion information to which item identification information of the selected item has been added. Next, the opinion information acquisition unit 1101 acquires one of the retrieved opinion information (step S121). Next, the acquisition of the emotion polarity value, the acquisition of the emotion intensity value, and the calculation of the emotion index value are performed (steps S103 to S105). Next, if there is opinion information among the retrieved opinion information for the selected item that has not yet been acquired (step S106: NO), the process proceeds to step S121.

[0093] If all opinion information has been obtained (step S106: YES), the process proceeds to step S122. In step S122, the statistical value calculation unit 1105 calculates the average of the emotion polarity values ​​obtained for the selected items as the average emotion intensity value for the items. Next, the statistical value calculation unit 1105 calculates the average of the emotion intensity values ​​obtained for the selected items as the average emotion intensity value for the items (step S123). The statistical value calculation unit 1105 stores the calculated average emotion polarity value and average emotion intensity value for the items in the storage unit 14, associating them with the item identification information of the selected items. Next, if there are items in the list that have not yet been selected (step S109: NO), the process proceeds to step S101.

[0094] If all items are selected (step S109: YES), the statistical value calculation unit 1105 calculates the overall average sentiment index value for all acquired opinion information (step S110). Next, the screen information transmission unit 1106 generates data for displaying a category map 230 that shows the relationship between the average sentiment polarity value and the average sentiment intensity value for each item (step S124). This data may be, for example, data written in HTML. At this time, the screen information transmission unit 1106 sets the coordinates for displaying the mark corresponding to each item in the data based on the average sentiment polarity value and average sentiment intensity of that item. The screen information transmission unit 1106 may also embed the item name in the data so that the name of the item is displayed near each mark. Next, the screen information transmission unit 1106 generates data for a list of review information 250 to be displayed in the review list display area 240, which includes review information 250 for all opinion information (step S125). Step S125 may be the same as step S113 shown in Figure 8, except that the list does not need to include evaluation values. Next, the screen information transmission unit 1106 generates an HTML document as an example of screen information for the analysis screen 200 (step S126). At this time, the screen information transmission unit 1106 may embed the overall average sentiment index value into this HTML document. The screen information transmission unit 1106 may also embed the data of the category map 230 into the HTML document. The screen information transmission unit 1106 may also embed the list data of the review information 250 into the HTML document so that at least a portion of the list of review information 250 is displayed in the review list display area 240. The screen information transmission unit 1106 may also embed an event handler in the HTML document to send a review information request containing item identification information corresponding to the selected mark to the opinion analysis server 1 when a mark on the category map 230 is selected. The screen information transmission unit 1106 sends the generated HTML document to the analyst terminal 3 (step S115). The analysis screen information transmission process is completed. Based on this HTML document, the analyst terminal 3 displays an analysis screen 200, such as the one shown in Figure 10.

[0095] The analyst selects one of the marks displayed on the category map 230. The analyst terminal 3 sends a review information request to the opinion analysis server 1, which includes item identification information corresponding to the selected mark. The system control unit 11 may perform the same processing as the review information transmission process shown in Figure 9 in response to receiving the review information request. Here, the list of review information 250 does not have to include evaluation values.

[0096] As described above, according to this embodiment, it is possible to grasp the trends in the emotional polarity value and emotional intensity value of the opinion information associated with each item based on the display position of the mark information, and to confirm the opinion information and emotional index value associated with a desired item.

[0097] (Note 1) An emotion index value calculation device comprising: an opinion information acquisition means for acquiring multiple opinion information, each representing an opinion; a polarity value acquisition means for acquiring, based on the opinion information, a polarity value indicating the tendency of the polarity of the emotion expressed in the opinion information and the degree of said tendency for each of the multiple opinion information acquired; an intensity value acquisition means for acquiring, based on the opinion information, an intensity value indicating the intensity of the emotion in the opinion information for each of the multiple opinion information acquired; an index value calculation means for calculating an index value by multiplying the acquired polarity value by the acquired intensity value for each of the multiple opinion information acquired; and a screen information transmission means for transmitting screen information to a display device that displays a screen showing at least one of the calculated index value and an index statistical value which is a statistical value of said index value.

[0098] (Note 2) The emotion index value calculation device according to Note 1, characterized in that each time an emotion expression appears in the opinion information, the intensity value is calculated such that the intensity value increases regardless of whether the emotion expression is a positive or negative emotion expression, and the intensity value acquisition means acquires the calculated intensity value.

[0099] (Note 3) The sentiment index value calculation device according to Note 1 or 2, wherein each of the acquired opinion information is associated with one of a predetermined set of items, and the device further comprises an index statistical value calculation means for each of the set of items, which calculates the index statistical value of the index value calculated for the opinion information associated with the item, and the screen information transmission means transmits screen information to be displayed on the screen, associating item information indicating the item with index statistical value information indicating the index statistical value calculated for the item, for each of the set of items.

[0100] (Note 4) The emotion index value calculation device according to Note 3, characterized in that the screen information transmission means can accept an operation on the screen to select any of the multiple index statistics values ​​shown in the multiple index statistics value information displayed on the screen, and when any of the index statistics values ​​is selected, the device transmits screen information to display on the screen the opinion information associated with the item for which the selected index statistics value was calculated and the index value calculated for the opinion information in association with each other.

[0101] (Note 5) The sentiment index value calculation device according to Note 3 or 4, further comprising: an evaluation value acquisition means for acquiring an evaluation value indicating the level of evaluation from the evaluator for each of the multiple opinion pieces acquired; and an evaluation statistics calculation means for calculating an evaluation statistics value which is a statistical value of the evaluation value acquired for each of the multiple items with respect to the opinion information associated with the item, wherein the screen information transmission means transmits screen information which further displays an evaluation statistics value indicating the evaluation statistics value calculated for each of the multiple items in association with the item information indicating the item.

[0102] (Note 6) The sentiment index value calculation device according to any one of Notes 1 to 5, wherein each of the acquired opinion information is associated with one of a predetermined set of items, and the device further comprises statistical value calculation means for each of the set of items, which calculates statistical values ​​of the polarity value and the intensity value calculated for the opinion information associated with the item, and the screen information transmission means displays on the screen, for each of the set of items, mark information indicating a mark corresponding to the item, which indicates the statistical values ​​of the polarity value and the intensity value calculated for the item at the display position of the mark information, and transmits screen information that allows the user to select one of the displayed mark information, and when one of the mark information is selected, the device displays on the screen the opinion information associated with the item to which the selected mark information corresponds and the index value calculated for the opinion information in association with the selected item.

[0103] (Note 7) A method for calculating an emotion index value performed by a computer, comprising: an opinion information acquisition step of acquiring a plurality of opinion information, each representing an opinion; a polarity value acquisition step of acquiring a polarity value indicating the tendency of the emotion polarity expressed in the opinion information and the degree of said tendency, based on the opinion information, for each of the plurality of opinion information acquired; an intensity value acquisition step of acquiring an intensity value indicating the intensity of the emotion in the opinion information, based on the opinion information, for each of the plurality of opinion information acquired; an index value calculation step of calculating an index value by multiplying the acquired polarity value by the acquired intensity value, for each of the plurality of opinion information acquired; and a screen information transmission step of transmitting screen information to a display device that displays a screen showing at least one of the calculated index value and an index statistical value which is a statistical value of said index value.

[0104] (Note 8) An emotion index value calculation program characterized in that the computer functions as an opinion information acquisition means for acquiring multiple opinion information, each representing an opinion; a polarity value acquisition means for acquiring a polarity value indicating the tendency of the polarity of the emotion expressed in the opinion information and the degree of said tendency, based on the opinion information for each of the multiple opinion information acquired; an intensity value acquisition means for acquiring an intensity value indicating the intensity of the emotion in the opinion information, based on the opinion information for each of the multiple opinion information acquired; an index value calculation means for calculating an index value by multiplying the acquired polarity value by the acquired intensity value for each of the multiple opinion information acquired; and a screen information transmission means for transmitting screen information to a display device that displays a screen showing at least one of the calculated index value and an index statistical value which is a statistical value of said index value. [Explanation of Symbols]

[0105] 1 Opinion Analysis Server 2. Opinion collection server 3. Analyst terminal 11 System Control Unit 12 System Bus 13 Input / Output Interfaces 14 Storage section 15 Communications Department 1101 Opinion information acquisition department 1102 Polarity Value Acquisition Unit 1103 Intensity value acquisition unit 1104 Indicator Value Calculation Unit 1105 Statistical calculation 1106 Screen Information Transmission Unit NW Network S Communication System

Claims

1. A means for acquiring opinion information that acquires multiple opinion pieces of information each representing an opinion, Polarity value acquisition means for each of the multiple opinion pieces obtained, which, based on the opinion piece, indicates the tendency of the polarity of the emotion expressed in the opinion piece, and if the polarity of the tendency is either positive or negative, acquires a polarity value indicating the degree of that polarity. For each of the multiple opinion pieces obtained, an intensity value acquisition means obtains an intensity value indicating the intensity of emotion in the opinion piece based on the opinion piece, An index value calculation means for each of the multiple opinion pieces obtained, which calculates an index value by multiplying the obtained polarity value by the obtained intensity value, A screen information transmission means transmits screen information to a display device that displays a screen showing at least one of the calculated index value and the index statistical value which is a statistical value of the index value. A device for calculating emotional index values, characterized by comprising the following features.

2. Each time an emotional expression appears in the opinion information, the intensity value is calculated such that it increases, regardless of whether the emotional expression is a positive or negative emotion. The emotion index value calculation device according to claim 1, characterized in that the intensity value acquisition means acquires the calculated intensity value.

3. The emotion index value calculation device according to claim 2, characterized in that when both positive and negative emotions appear in the opinion information, the intensity value is calculated to include both the increase in the intensity value for the positive emotion and the increase in the intensity value for the negative emotion.

4. Each of the obtained opinion information represents the text of the opinion, The emotion index value calculation device according to claim 1, characterized in that, if the opinion statement indicated by any of the opinion information mentioned above includes multiple sentences, the polarity value acquisition means acquires the polarity value corresponding to the entire opinion statement including the multiple sentences, the intensity value acquisition means acquires the intensity value corresponding to the entire opinion statement including the multiple sentences, and the index value calculation means calculates the product of the polarity value corresponding to the entire opinion statement and the intensity value corresponding to the entire opinion statement as the index value corresponding to the entire opinion statement.

5. Each of the obtained opinion information is associated with one of a predetermined set of items, The system further comprises an index statistical value calculation means for each of the aforementioned items, which calculates the index statistical value of the index value calculated for the opinion information associated with the aforementioned item, The emotion index value calculation device according to claim 1, characterized in that the screen information transmission means transmits screen information that displays on the screen, relating item information indicating the item and index statistical value information indicating the index statistical value calculated for the item, for each of the plurality of items.

6. The emotion index value calculation device according to claim 5, wherein the screen information transmission means is capable of receiving an operation on the screen to select any of the multiple index statistics values ​​shown in the multiple index statistics value information displayed on the screen, and when any of the index statistics values ​​is selected, the device transmits screen information to display on the screen the opinion information associated with the item for which the selected index statistics value was calculated and the index value calculated for the opinion information in association with each other.

7. The opinion information indicates the evaluator's opinion on the subject of evaluation, For each of the multiple opinion pieces obtained, an evaluation value acquisition means is provided to acquire an evaluation value indicating the level of evaluation from the evaluator for the subject of evaluation. For each of the aforementioned items, an evaluation statistics calculation means calculates an evaluation statistics value which is a statistical value of the evaluation value obtained for the opinion information associated with the said item, Furthermore, The emotion index value calculation device according to claim 5 or 6, characterized in that the screen information transmission means transmits screen information that further displays a screen showing evaluation statistical value information indicating the evaluation statistical value calculated for each of the plurality of items, in association with the item information indicating the item.

8. Each of the obtained opinion information is associated with one of a predetermined set of items, The system further comprises statistical value calculation means for each of the aforementioned items, which calculates statistical values ​​for the polarity value and the intensity value calculated for the opinion information associated with the aforementioned item, The screen information transmission means transmits screen information that, for each of the plurality of items, displays mark information indicating a mark corresponding to the item, and that indicates statistical values ​​of the polarity value and statistical values ​​of the intensity value calculated for the item at the display position of the mark information, and transmits screen information that allows the user to select any of the displayed mark information, and when any of the mark information is selected, displays on the screen the opinion information associated with the item to which the selected mark information corresponds and the index value calculated for the opinion information in association with it, as described in any one of claims 2 to 6.

9. A method for calculating an emotion index value performed by a computer, An opinion information acquisition step to obtain multiple opinion pieces of information each representing an opinion, A polarity value acquisition step for each of the multiple opinion pieces obtained, which, based on the opinion piece, indicates the tendency of the polarity of the emotion expressed in the opinion piece, and if the polarity of the tendency is either positive or negative, obtains a polarity value that indicates the degree of that polarity. For each of the multiple opinion pieces obtained, an intensity value acquisition step is taken to obtain an intensity value indicating the intensity of emotion in the opinion piece based on the opinion piece. For each of the multiple opinion pieces obtained, an index value calculation step is performed, in which an index value is calculated by multiplying the obtained polarity value by the obtained intensity value. A screen information transmission step involves transmitting screen information to a display device that displays a screen showing at least one of the calculated index value and the index statistical value which is a statistical value of the index value. A method for calculating an emotional index value, characterized by including the following:

10. A computer, A means for acquiring opinion information that acquires multiple opinion pieces of information each representing an opinion, Polarity value acquisition means for each of the multiple opinion pieces obtained, which, based on the opinion piece, indicates the tendency of the polarity of the emotion expressed in the opinion piece, and if the polarity of the tendency is either positive or negative, acquires a polarity value indicating the degree of that polarity. For each of the multiple opinion pieces obtained, an intensity value acquisition means obtains an intensity value indicating the intensity of emotion in the opinion piece based on the opinion piece, An index value calculation means for each of the multiple opinion pieces obtained, which calculates an index value by multiplying the obtained polarity value by the obtained intensity value, Screen information transmission means for transmitting screen information to a display device that displays a screen showing at least one of the calculated index value and the index statistical value which is a statistical value of the index value. A program for calculating emotional index values, characterized by its function as such.

Images