Information processing device and program

Abstract

Providing a new method for reliably estimating inter-brand affinity. [Solution] Information processing device 1 includes a memory unit 14 that stores a program, a processor 11 that executes the program, and a communication unit 17 that communicates with SNS server devices 3-1 and 3-2 via an internet line 2. By executing the program, the processor functions as a collection unit 24 that requests the SNS server device to transmit multiple words that have been added along with hashtags to posts by multiple users of the SNS service and receives the requests, an extraction unit 25 that extracts words that represent a brand (brand words) from the multiple words, an extraction unit 26 that extracts words (non-brand words) that are added to the same content as each brand word as co-occurring words, a counting unit 27 that counts the number of times the extracted co-occurring words appear for each brand word, and a calculation unit 28 that calculates a score (affinity score) that represents the affinity between brands based on the number of times the co-occurring words appear.

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device and a program. [Background technology]

[0002] In recent years, brand affinity has become important in marketing. Brand affinity refers to a relationship in which different brands share common values ​​and images, and when these coincide, a synergistic effect is created. When brands with high affinity work together, the brands' credibility and appeal increases, and they are expected to have a stronger influence. Brand affinity can also be utilized when searching for influencers that are suitable for your own brand. It is effective to employ influencers related to other companies' brands that have a high affinity with your own brand as influencers for your own brand.

[0003] Conventional methods for estimating brand affinity use the following factors:

[0004] Commonalities among brand buyers - Degree of overlap between the customer demographics of both brands (age, gender, income, place of residence, values, etc.) Commonalities in the scenes in which the brand is used Similarity of brand image · Social media follower overlap On the other hand, if a collaboration strategy is pushed forward despite the low affinity between the brands, it may lead to consumer confusion and sales may not grow as expected. Not only that, but there is also the risk of damaging the brand image and reducing brand value, so careful judgment is required. Summary of the Invention [Problem to be solved by the invention]

[0005] New methods are expected to reliably estimate inter-brand affinity. [Means for solving the problem]

[0006] The information processing device according to this embodiment includes a storage unit that stores a program, a processor that executes the program, and a communication unit that communicates with an SNS server device that provides an SNS service via an internet connection. By executing the program, the processor functions as a means for requesting the SNS server device to transmit a plurality of words that have been added together with hashtags to each of a plurality of posts by a plurality of users using the SNS service and receiving the same, a means for extracting words that represent a brand (brand words) from the plurality of words, a means for extracting words that have been added together with hashtags to the same content as each brand word as co-occurring words (non-brand words), a means for counting the number of times the extracted co-occurring words appear for each brand word, and a means for calculating a score (affinity score) that represents the affinity between brands based on the number of times the co-occurring words appear. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a configuration diagram of an entire system including an information processing device according to this embodiment. [Figure 2] FIG. 2 is a diagram showing the physical configuration of the information processing apparatus of FIG. [Figure 3] FIG. 3 is a diagram illustrating a functional configuration of the processor in FIG. [Figure 4] FIG. 4 is a diagram showing an example of a user database relating to SNS users stored in the storage unit of FIG. [Figure 5] FIG. 5 is a diagram showing an example of a brand list stored in the storage unit of FIG. [Figure 6] FIG. 6 is a diagram showing an example of a profile screen in an SNS service. [Figure 7] FIG. 7 is a diagram showing an example of a post details screen in the SNS service. [Figure 8] FIG. 8 is a flowchart showing a processing procedure performed by the processor of FIG. [Figure 9]FIG. 9 is a diagram showing an example of the list of hashtag words collected in step S12 of FIG. [Figure 10] FIG. 10 is a diagram showing an example of the number of occurrences of co-occurring words counted in step S15 of FIG. [Figure 11] FIG. 11 is a diagram showing an example of a vector relating to the occurrence counts of co-occurring words generated in step S16 of FIG. [Figure 12] FIG. 12 is a diagram showing an example of a graphic representation of the affinity scores between brands calculated in step S16 of FIG. [Figure 13] FIG. 13 is a supplementary diagram illustrating the process flow of steps S18, S19, and S20 in FIG. [Figure 14] FIG. 14 is a diagram showing an example of the investigation report generated in step S21 of FIG. DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. The configuration of an entire system including an information processing device according to this embodiment is shown in Fig. 1. SNS server devices 3-1 and 3-2 that provide SNS services are connected to an information processing device 1 according to this embodiment via an internet line 2.

[0009] As shown in FIG. 2, the information processing device 1 has a processor 11. A RAM 12, a ROM 13, a storage unit 14, an input device 15, a display 16, and a communication unit 17 are connected to the processor 11 via a system bus 10. The processor 11 is configured by, for example, a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor 11 executes a "brand affinity score estimation processing program" loaded from the storage unit 14 to the RAM 12.

[0010] The RAM 12 functions as the main memory, work area, etc. of the processor 11. The ROM 13 or storage unit 14 stores the BIOS (Basic Input Output System) executed by the processor 11, an operating system program (OS), a user database of multiple users who use the SNS service (see Figure 4), a "brand affinity score estimation processing program," and other programs for realizing various functions, as well as various data required for these processes. The input device 15 includes a keyboard (KB), a pointing device such as a mouse or touch panel, etc. The display 16 is typically realized by an LCD (Liquid Crystal Display).

[0011] 3 shows the functional configuration of the processor 11. By executing the "brand affinity score estimation processing program," the processor 11 functions as a control unit 21, a memory management unit 22 that manages writing and reading to and from the memory unit 14, a user extraction unit 23, a hashtag word collection unit 24, a brand word extraction unit 25, a co-occurrence word extraction unit 26, a co-occurrence word counting unit 27, a brand affinity score calculation unit 28, a brand posting rate calculation unit 29, a user / brand affinity score calculation unit 30, and a report editing unit 31.

[0012] The user extraction unit 23 extracts multiple users to be applied to the brand affinity score estimation process from a user database of multiple users using the SNS service stored in the memory unit 14, based on extraction conditions specified by an investigator, such as demographic information (e.g., gender, age (generation)), and items such as hobbies and preferences. As illustrated in FIG. 4, the user database is composed of predetermined items such as accounts, number of posts, demographic information (gender, age, place of residence), hobbies, and preferences for each of multiple users using the SNS service. The processor 11 periodically requests the SNS server devices 3-1 and 3-2 to provide information on the above items regarding multiple users using the SNS service. The memory management unit 22 updates the user database in the memory unit 14 based on the information on the above items received from the SNS server devices 3-1 and 3-2. When starting the brand affinity score estimation process, the processor 11 may send a user extraction request to the SNS server devices 3-1 and 3-2 along with extraction conditions specified by an investigator, such as demographic information (e.g., gender, age (generation)), and items such as hobbies and preferences, and receive the user information.

[0013] FIG. 6 shows a profile screen including a list of posts by a certain user, and FIG. 7 shows a detailed screen of a certain post. For each post, multiple words are set with a hashtag "#" attached. The hashtag word collection unit 24 uses the accounts of the extracted multiple users to request the SNS server devices 3-1, 3-2 to transmit multiple words (hereinafter referred to as hashtag words) attached with a hashtag "#" for each of the multiple posts posted by each of the users, and receives the requests. The memory management unit 22 associates the received hashtag words with the user account, post ID, and post date, and stores them in the memory unit 14 (see FIG. 9).

[0014] The storage unit 14 stores a brand list, as shown in Fig. 5. The brand list is a list of multiple brand names (character strings written in their original, formal spelling). Each brand name is associated with character strings written in various spellings that differ from the original spelling, as well as character strings representing common names and aliases (colloquial names) that are naturally used by the general public.

[0015] The brand word extraction unit 25 extracts hashtag words (referred to as brand words) that represent any brand from the hashtag words by comparing the hashtag words with the brand list. Since the brand list associates original spellings (brand names) with spellings and nicknames that differ from the original spelling, even if users write hashtag words in their own spellings, it is possible to extract almost all brand words from the hashtag words. The brand word extraction unit 25 also counts the number of times a brand word appears for each user.

[0016] The co-occurrence word extraction unit 26 extracts, as co-occurrence words associated with the brand, hashtag words (non-brand words) used together with the brand word, rather than brand words, in other words, hashtag words (non-brand words) attached to the same content as the brand word. Non-brand words represent the image and awareness that the user who posted the brand word has of the brand, the products and services that the user considers to be related to the brand, the target demographic and usage scenarios that the user considers, etc.

[0017] The co-occurrence word counting unit 27 classifies all the extracted non-brand words posted by the multiple users by brand word (brand) and counts the number of occurrences of each non-brand word (see FIG. 10). In other words, it calculates the distribution of the number of occurrences of non-brand words as co-occurrence words related to each brand.

[0018] The inter-brand affinity score calculation unit 28 calculates a score (affinity score) representing the affinity between brands based on the frequency of appearance of non-brand words. Specifically, the number of types of non-brand words that appear across brand words used by all surveyed users is defined as the degree, and a vector is generated whose elements are the frequency of appearance of non-brand words counted for each brand word (see Figure 11). The cosine similarity between the vectors of any two brands is calculated as the affinity score. If the vector of a certain brand a is denoted as A and the vector of another brand b is denoted as B, the cosine similarity CS(AB) is calculated using the following formula: A·B represents the dot product of vector A and vector B. |A| represents the length (norm) of vector A, and |B| represents the length (norm) of vector B.

[0019] CS(AB)=A·B / |A||B| Cosine similarity is an index that measures the similarity between two vectors. A vector related to a brand is based on the frequency of co-occurring brand words. Therefore, the cosine similarity CS(AB) between vector A related to brand a and vector B related to brand b represents the degree of similarity and commonality between brands a and b in terms of the general user's image and awareness of brands a and b, the products and services associated with brands a and b, and the target demographics and usage scenarios of brands a and b that users consider. In other words, it represents the degree of affinity, or the degree of mutual association, or familiarity, between the two brands. The value of cosine similarity CS(AB) is calculated within a range from a minimum of "-1" to a maximum of "+1." Affinity between brands is lowest at "-1," and the closer it is to the maximum value of "+1," the higher the affinity.

[0020] Although the cosine similarity is optimal as the affinity score, it is not limited to this, and may be, for example, the Euclidean distance between vectors for each brand word.

[0021] The inter-brand affinity score calculation unit 28 can calculate multiple affinity scores between two different combinations of brands (brand pairs). The inter-brand affinity score calculation unit 28 outputs multiple affinity scores for multiple brand pairs in a unified manner, for example, as a correlation diagram (see FIG. 12).

[0022] Furthermore, by setting multiple conditions for extracting users, user groups with different attributes can be extracted. For example, a user group of men in their 40s and a user group of women in their 20s can be extracted, and the inter-brand affinity score calculation unit 28 can calculate affinity scores for the same two brands from the user group of men in their 40s and the user group of women in their 20s. By outputting the affinity scores calculated from the user group of men in their 40s and the user group of women in their 20s side by side, it is possible to recognize changes in affinity scores depending on the user group.

[0023] The calculation (estimation) of affinity scores between brands can be expanded to the calculation (estimation) of affinity scores between users and brands. The calculation (estimation) of affinity scores between users and brands is realized by the brand post rate calculation unit 29 and the user / brand affinity score calculation unit 30. For example, if a user is an influencer for a certain brand, it is possible to estimate the degree of affinity the user has with other brands (such as the user's own brand) with which the user has not been involved in the past. For example, this can be used to discover new influencers.

[0024] The brand post rate calculation unit 29 calculates, for each user, the proportion of posts in which each brand word is attached along with a hashtag (brand post rate) out of all posts posted by each user. The brand post rate indicates the proportion of posts each user makes to each brand. The user / brand affinity score calculation unit 30 calculates (estimates) the affinity score between the user and non-posted brand words by multiplying the affinity score (already calculated by the inter-brand affinity score calculation unit 28) between the brand words attached to the user's own posts (posted brand words) and brand words attached to posts of other users but not to the user's posts, by the brand post rate for the posted brand words.

[0025] The report editing unit 31 creates a brand affinity survey report listing the inter-brand affinity scores calculated by the inter-brand affinity score calculation unit 28 and the affinity scores between users and brands calculated by the user / brand affinity score calculation unit 30 (see FIG. 14 ). A template is prepared for the report. The inter-brand affinity scores calculated by the inter-brand affinity score calculation unit 28 are laid out in accordance with the template, along with interpretations that have been created in advance and stored in the memory unit 14 in association with the score value ranges. A correlation diagram is created and laid out to visually understand brand names by displaying a distance according to the reciprocal of multiple affinity scores for different combinations of two brands, i.e., by positioning brands with high affinity closer together and brands with low affinity farther apart. For a specific user, the affinity scores between the user and brands to which the user has never posted and with which the user has had little or no previous connection are also laid out in accordance with the template.

[0026] FIG. 8 shows the processing procedure performed by the processor 11. The investigator specifies user extraction conditions related to items such as the number of posts, gender, age (generation), hobbies, and preferences. The user extraction unit 23 searches the user database illustrated in FIG. 4 according to the user extraction conditions, and multiple users who match the user extraction conditions are extracted as users to be applied to the brand affinity score estimation process (S11). The user database is constructed based on information related to items such as the number of posts, gender, age (generation), hobbies, and preferences of multiple users who use the SNS service, obtained from the SNS server devices 3-1 and 3-2, and is updated periodically.

[0027] Using the extracted accounts of the multiple users, the hashtag word collection unit 24 collects from the SNS server devices 3-1 and 3-2 multiple words (hashtag words) that are added to each of the multiple posts posted by each of the users along with the hashtag "#" (S12). As illustrated in Fig. 9, the hashtag words are stored in the storage unit 14 in association with the user account, post ID, and post date.

[0028] The hashtag words are compared with the brand list in FIG. 5. This extracts hashtag words (brand words) that represent any of the brands (S13). The brand list includes not only the original spelling but also various spellings and nicknames, so it is possible to extract almost all brand words. The number of times a brand word appears is counted for each user by the brand word extraction unit 25.

[0029] Next, hashtag words (non-brand words) are extracted together with the extracted brand words, i.e., hashtag words (non-brand words) that are attached with the hashtag # to the same content (post) as the extracted brand words, and are associated with the brand as co-occurring words (S14). For example, in the example of Figure 7, the non-brand words "Sports," "Running," and "socker" are extracted together with the brand word "Niko." The words "Sports," "Running," and "socker" are co-occurring words with the brand "Niko," and are perceived by the user as representing the image of the brand "Niko."

[0030] As illustrated in FIG. 10, the co-occurrence word counting unit 27 classifies all non-brand words (co-occurrence words) related to the extracted posts of all users into brand words (brands), and counts the number of occurrences of each non-brand word (co-occurrence word) (S15).

[0031] The inter-brand affinity score calculation unit 28 generates vectors, as shown in FIG. 11, for each of the extracted brand words (brands), with the frequency of occurrence of non-brand words as element components (S16). Two pairs of the extracted brand words (brands) are combined, and a score (affinity score) representing the affinity between the two brands is calculated based on the two vectors for the two brand words (S17). The affinity score is calculated as the cosine similarity between the two vectors.

[0032] As is well known, cosine similarity is an index that measures how similar two vectors are. Since a vector related to a certain brand is based on the frequency of occurrence of co-occurring brand words, the cosine similarity between a vector related to one brand and a vector related to another brand can be an index that measures whether the two brands are closely or distantly related in terms of the image and awareness that general users have of the brand, the products and services that are related to brands a and b, the target demographic and usage scenarios that users have in mind for the brand, etc.

[0033] The affinity scores between brands can then be used to calculate (estimate) the affinity scores between a particular user and brands that the user is not currently interested in.

[0034] The brand post rate calculation unit 29 calculates the ratio of the number of posts that each extracted user has posted about each brand to the total number of posts posted by each extracted user as the brand post rate for that brand (S18). FIG. 13 shows a specific example. Assume that user A is an influencer for a certain brand and has posted about brands 1, 2, and 3. Assume also that the brands extracted from the posts of all extracted users are 1, 2, 3, ..., 6. In other words, user A has never posted about brands 4, 5, or 6, or in other words, user A has not been involved with brands 4, 5, or 6 until now.

[0035] What we want to estimate here is the affinity score between User A and each of Brands 4, 5, and 6. The user / brand affinity score calculation unit 30 extracts the inter-brand affinity scores for two combinations of two brands for brands 1, 2, 3, ..., 6: brands 1, 2, 3 posted by user A and brands 4, 5, 6 not posted by user A. The highest inter-brand affinity score is extracted for each of brands 4, 5, 6 (S19). In this example, the combination with brand 4 shows the highest inter-brand affinity score with brand 3. The combination with brand 5 shows the highest inter-brand affinity score with brand 2. The combination with brand 6 shows the highest inter-brand affinity score with brand 3.

[0036] The user / brand affinity score calculation unit 30 calculates the affinity score between the user and the non-posting brand by multiplying the affinity score between the brand related to the user A's post and the brand to which the user A has not posted by the brand posting rate for the brand related to the post (S20).

[0037] In the example of FIG. 13 , the affinity score between User A and Brand 4, to which User A has no posts, is estimated by multiplying User A's posting rate for Brand 3 by the affinity score between Brand 3, to which User A has posted, and Brand 4, to which User A has no posts. In other words, User A's posting rate for Brand 3 can be said to represent User A's level of interest in Brand 3. Therefore, the affinity score between User A and Brand 4, to which User A has no posts, can be indirectly estimated by discounting the affinity score between Brand 3, to which User A has posted, and Brand 4, to which User A has no posts, by User A's posting rate for Brand 3. Similarly, the affinity score between User A and Brand 5, to which User A has no posts, is estimated by multiplying User A's posting rate for Brand 2, to which User A has posted, by the affinity score between Brand 2, to which User A has posted, and Brand 5, to which User A has no posts. The affinity score between User A and Brand 6, to which User A has no posts, is estimated by multiplying User A's posting rate for Brand 3 by the affinity score between Brand 3 and Brand 6.

[0038] Finally, the report editing unit 31 creates a brand affinity research report (S21). As shown in Fig. 14, the calculated affinity scores between the brands are laid out according to a brand affinity research report template. An interpretation is set in advance for each of a plurality of score value ranges, and the interpretation corresponding to the calculated affinity score between the brands is written alongside the affinity score.

[0039] It is expected that the number of brands appearing in the posts of the extracted users will be very large. For these many brands, the pairings will also vary widely. It is practical to extract a certain number of affinity scores that indicate the highest affinity scores and include only the brand combinations associated with the extracted affinity scores in the report. Furthermore, when there is a brand of interest, such as a company's own brand, it may be appropriate to calculate the affinity scores between the brand of interest and each of the other brands and include only the brand combinations associated with the highest affinity scores in the report.

[0040] Below the affinity score and its interpretation is a visually easy-to-understand correlation diagram in which brand names are spaced apart at distances according to the affinity score value. The closer the affinity score value is to +1.0, the higher the affinity between brands, and the closer it is to -1.0, the lower the affinity between brands, so the distance is set according to the inverse of the affinity score value.

[0041] Furthermore, below the correlation diagram, affinity scores are laid out between specific users designated by the investigator from among the extracted users, such as users who are expected to be influencers for the investigator's own brand or users who act as influencers for a brand with a large number of posts, and brands to which the user has never posted and with which the user has had little or no previous connection.

[0042] As described above, according to this embodiment, brand words included in hashtag words in a social networking service are extracted, hashtag words (non-brand words) that are used in the same content as the brand words are applied as co-occurring words of the brand words, and the degree of similarity (cosine similarity, etc.) between the distributions (vectors) of the frequency of appearance of the co-occurring words between the brands is calculated as the affinity between the brands, thereby making it possible to reliably estimate the degree of affinity between brands. Furthermore, this can be extended to estimate the degree of affinity between a given user and a brand to which the user has not posted, i.e., a brand to which the user is presumed to have little interest. For example, this can be effectively used to search for brands to collaborate with and confirm synergistic effects. It can also be used, for example, to search for and confirm influencers suitable for one's own brand.

[0043] Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, as well as within the scope of the invention described in the claims and their equivalents. [Explanation of symbols]

[0044] 1...information processing device, 2...internet line, 3-1, 3-2...SNS server device, 21...control unit, 22...memory management unit, 23...user extraction unit, 24...hashtag word collection unit, 25...brand word extraction unit, 26...co-occurrence word extraction unit, 27...co-occurrence word aggregation unit, 28...brand affinity score calculation unit, 29...brand post rate calculation unit, 30...user / brand affinity score calculation unit, 31...report editing unit

Claims

1. The device includes a storage unit that stores a program, a processor that executes the program, and a communication unit that communicates with an SNS server device that provides an SNS service via an internet line, The processor executes the program, a means for requesting the SNS server device to transmit a plurality of words attached together with a hashtag for each of a plurality of posts by a plurality of users using the SNS service, and receiving the same; a brand word extraction means for extracting words representing brands (brand words) from the plurality of words; A means for extracting words (non-brand words) that are tagged with hashtags to the same content as each of the brand words as co-occurring words; means for counting the number of times the extracted co-occurring words appear for each of the brand words; An information processing device that functions as affinity calculation means that calculates a score (affinity score) representing the affinity between the brands based on the number of times the co-occurring words appear.

2. 2. The information processing apparatus according to claim 1, wherein said affinity calculation means calculates a cosine similarity between vectors for each of said brand words, with the frequency of occurrence of said co-occurring words as an element component, as said affinity score between said brands.

3. 2. The information processing apparatus according to claim 1, wherein said affinity calculation means calculates, as the affinity score between said brands, a Euclidean distance between vectors for each of said brand words, with the frequency of occurrence of said co-occurring words as an element component.

4. The information processing device according to claim 1 , wherein the processor further functions as a means for extracting the plurality of users from all users of the SNS service based on extraction conditions relating to at least one of age, gender, residential area, hobbies, and preferences.

5. The information processing device according to claim 4 , wherein the processor further functions as a means for outputting a change in the affinity score between the brands due to a difference in the extraction conditions.

6. The processor: A means for calculating, for each user, the ratio of the number of posts in which the brand word is attached together with a hashtag to the total number of posts for each user (brand post rate); 2. The information processing device of claim 1, further functioning as a means for calculating an affinity score between each of the users and the non-posted brand words by multiplying, for each of the users, an affinity score between the brand words (posted brand words) attached to the user's own posts and the brand words (non-posted brand words) not attached to the user's own posts but attached to posts of other users by the brand posting rate.

7. the storage unit stores a brand list in which different notations representing the brand are associated with the name of the brand; The information processing apparatus according to claim 1 , wherein the brand word extraction means extracts the brand words by comparing the words with the brand list.

8. A computer that communicates with an SNS server device that provides an SNS service via an Internet line, a means for requesting the SNS server device to transmit a plurality of words attached together with a hashtag for each of a plurality of posts by a plurality of users using the SNS service, and receiving the same; a brand word extraction means for extracting words representing brands (brand words) from the plurality of words; A means for extracting words (non-brand words) that are tagged with hashtags to the same content as each of the brand words as co-occurring words; means for counting the number of times the extracted co-occurring words appear for each of the brand words; a program that functions as affinity calculation means that calculates a score (affinity score) representing the affinity between the brands based on the number of times the co-occurring words appear;

Images