Merchant scoring system based on commercial socialized comment feedback

Abstract

The application discloses a merchant scoring system based on commercial socialized comment feedback, and realizes comment real-name system by comparing commenters with ID photos in a database, so that the risk value of brushing good comments can be greatly improved, and user behavior of brushing good comments can be effectively avoided; according to user feature information determined according to product labels, the application classifies comments, and compares comments with the same focus on the same product with each other, so that selectively grouping can improve the accuracy of authenticity judgment; the use and perception of the same product of users with the same user feature information should be similar, and by comparing the comment with other comments with the same user feature information, false good comments can be obtained; the final historical good comment rate obtained by the application is the result of all real comments of the product, and the final historical good comment rate can truly reflect the quality of the corresponding product, and can provide real reference value for subsequent consumers.

Description

Technical Field

[0001] This invention relates to the field of e-commerce reputation scoring technology, and in particular to a merchant scoring system based on socialized product review feedback. Background Technology

[0002] With the rapid development of e-commerce platforms, online shopping has become an important way for people to purchase products in their daily lives. In online shopping, most consumers look at reviews left by previous buyers, which provides valuable reference for later purchases. However, for merchants, taking advantage of this common buyer psychology, they often take measures to increase their positive review rate. A common method is "brushing" (fake positive reviews), where merchants pay people to pose as customers to artificially inflate their online store's ratings and attract more customers. This practice leads to inaccurate information about the merchant provided to ordinary users, negatively impacting their shopping experience. Current technology lacks effective methods to identify brushing behavior and to analyze user comments on merchants based on social media reviews. Summary of the Invention

[0003] The purpose of this invention is to provide a merchant scoring system based on socialized product review feedback that can effectively identify fake positive reviews and obtain a more authentic merchant positive review rate.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a merchant scoring system based on socialized product review feedback, including an identity verification database, a user feature database, and a preference classification database. The identity verification database stores several platform IDs and corresponding ID card photos. The user feature database stores several product tags and corresponding user feature information. The user feature information reflects factors that can influence the content of users' reviews of products with corresponding product tags. The preference classification database stores several feature keywords and corresponding preference groups.

[0005] The merchant scoring system also includes identity verification strategies, user preference classification strategies, review authenticity identification strategies, and positive review rate scoring strategies.

[0006] The identity verification strategy involves obtaining a real-time photo captured by a camera as the image to be identified, identifying the image to be identified to obtain several facial feature information, comparing the facial feature information and the corresponding platform ID to be identified with the data in the identity verification database, and if the comparison is successful, proceeding to the user preference classification strategy; otherwise, the process ends.

[0007] The user preference classification strategy obtains the product tag of the product to be reviewed as the matching tag, searches the user feature database according to the matching tag to obtain the corresponding user feature information as the feature information to be filled, obtains the subjective feature information input by the user according to the feature information to be filled, identifies the subjective feature information to obtain subjective feature keywords, searches the preference classification database according to the subjective feature keywords to obtain the corresponding preference group as the actual preference group, and enters the review authenticity identification strategy.

[0008] The review authenticity identification strategy calculates the deviation value between a review submitted by a platform ID for a product to be reviewed and other reviews of the same product within the same actual preference group to obtain a real-time deviation value. It then obtains reviews of similar products reviewed by the platform ID on the platform as historical reviews, calculates a historical cumulative deviation value based on these historical reviews, updates the historical cumulative deviation value using the historical cumulative deviation value and the real-time deviation value, and calculates the true value based on the new historical cumulative deviation value. If the true value is less than or equal to a preset true threshold, the corresponding review is marked as a genuine review; if the true value is greater than the preset true threshold, the corresponding review is marked as a fake review, and the positive review rate scoring strategy is then applied.

[0009] The positive review rate scoring strategy calculates the following: if a genuine review is received, a real-time positive review rate is obtained based on the content of the genuine review; a new historical positive review rate is then obtained by updating the historical positive review rate based on the real-time positive review rate. If a fake review is received, a reputation depreciation value is generated; a new historical reputation value is then obtained by updating the historical reputation value based on the reputation depreciation value, and the updated historical reputation value is lower than the previous historical reputation value. If the historical reputation value is lower than a preset reputation threshold, the historical positive review rate is reset to the initial positive review rate.

[0010] Preferably, in the authentication strategy, multi-angle photos captured in real time by a camera are used as the images to be identified.

[0011] Preferably, in the user preference classification strategy, if the product label is beauty or clothing, the facial feature information is obtained as objective feature information, the objective feature information is identified to obtain objective feature keywords, and the actual preference group is determined based on the objective feature keywords and the subjective feature keywords.

[0012] Preferably, the objective feature information includes skin color.

[0013] Preferably, the objective feature information includes facial shape.

[0014] Preferably, in the comment authenticity identification strategy, comments submitted by the platform ID for the product to be reviewed are used as comments to be identified. The comments to be identified are marked as short reviews, medium reviews, or long reviews according to the number of words. Historical comments submitted by the platform ID that are similar to the product to be reviewed are obtained. The text repetition rate between the comments to be identified and the historical comments is calculated to obtain a first repetition value. Other comments within the same actual preference group for the same product to be reviewed are obtained as reference comments. The text repetition rate between the comments to be identified and the reference comments is calculated to obtain a second repetition value.

[0015] If the comment to be identified is a short comment, the first difference and the second difference are obtained by obtaining the difference between the preset first repetition threshold and the first repetition value and the second repetition value, respectively. The true value is obtained based on the first difference, the second difference and the historical cumulative deviation value.

[0016] If the review to be identified is a neutral review, then the difference between the preset second repetition threshold and the first repetition value and the second repetition value are respectively used to obtain the third difference and the fourth difference. The true value is obtained based on the third difference, the fourth difference and the historical cumulative deviation value.

[0017] If the comment to be identified is a long comment, the fifth difference and the sixth difference are obtained by obtaining the difference between the preset third repetition threshold and the first repetition value and the second repetition value, respectively. The true value is obtained based on the fifth difference, the sixth difference and the historical cumulative deviation value.

[0018] The first repetition threshold is greater than the second repetition threshold, which is greater than the third repetition threshold.

[0019] As a preferred option, the merchant scoring system also includes a review matching database, which stores several product tags and corresponding review keywords;

[0020] In the comment authenticity identification strategy, comments submitted by the platform ID for the product to be commented are used as comments to be matched. The comments to be matched are identified to obtain comment keywords as real-time comment words. The real-time comment words are matched in the comment matching database to obtain the comment matching degree. The true value is obtained based on the comment matching degree and the historical cumulative deviation value.

[0021] Preferably, the merchant scoring system also includes a contact list database, which stores the platform ID and other platform IDs that have social communication with the platform ID. The merchant scoring system also includes a sharing chain identification strategy.

[0022] The sharing chain identification strategy generates a sharer field. If a platform ID is obtained based on the other platform ID entered in the sharer field, it is used as the sharing source ID. The platform ID and the sharing source ID are entered into the contact list database for matching. The platform ID and the sharing source ID form a sharing chain. The sharing recommendation value is obtained based on the chain length of the sharing chain.

[0023] In the positive review rate scoring strategy, the real-time positive review rate is obtained based on the content of the real reviews and the sharing and recommendation value.

[0024] Preferably, in the sharing chain identification strategy, the sharing duration is obtained by obtaining the chat duration of the platform ID and the sharing source ID, and the sharing recommendation value is obtained based on the sharing duration and the chain length of the sharing chain.

[0025] Preferably, in the sharing chain identification strategy, the chat content obtained from the platform ID and the sharing source ID is used as the sharing content, the keywords related to the product tag are identified as the sharing keywords, the frequency of the sharing keywords is counted to obtain the sharing frequency, and the sharing seed value is obtained based on the sharing frequency, the sharing duration and the chain length of the sharing chain.

[0026] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0027] 1. This invention implements real-name registration for comments by comparing the reviewer's ID photo with the ID photo in the database, which can significantly increase the risk of fake positive reviews and thus effectively prevent users from engaging in fake positive reviews.

[0028] 2. Different types of users have different priorities when it comes to the same product. If all reviews of the same product are judged for authenticity in the same way, the accuracy of the judgment will be low. This invention classifies reviews based on user characteristic information determined by product tags, and compares reviews with the same priorities for the same product. This selective grouping can improve the accuracy of authenticity judgment.

[0029] 3. Users with similar user characteristics should have similar experiences using the same product. If a user submits a positive review after engaging in fraudulent order placement, they will not know the true user experience because they have not actually used the product. Therefore, their review will differ significantly from the genuine review. Comparing these differences can help identify fraudulent positive reviews.

[0030] 4. The final historical positive review rate obtained by this invention is the result of the sum of all real reviews of this product. The final historical positive review rate can truly reflect the quality of the corresponding product and can provide real reference value for subsequent consumers. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of a merchant scoring system.

[0032] The following are explanations of the reference numerals in the attached diagram: 011, Authentication strategy; 012, User preference classification strategy; 013, Review authenticity identification strategy; 014, Sharing chain identification strategy; 015, Positive review rate scoring strategy; 021, Authentication database; 022, User feature database; 023, Preference classification database; 024, Contact list database. Detailed Implementation

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0034] Example 1:

[0035] like Figure 1 As shown, the merchant scoring system based on socialized product review feedback includes an identity verification database 021, a user feature database 022, and a preference classification database 023. The identity verification database 021 stores several platform IDs and corresponding ID card photos. The user feature database 022 stores several product tags and corresponding user feature information. The user feature information reflects factors that can influence users' reviews of products with corresponding product tags (such as occupation, age, height, weight, taste, skin color, etc.). The preference classification database 023 stores several feature keywords and corresponding preference groups.

[0036] The merchant scoring system also includes identity verification strategy 11, user preference classification strategy 012, review authenticity identification strategy 013, and positive review rate scoring strategy 015.

[0037] The identity verification strategy 11 obtains a real-time photo taken by the camera (only photos taken in real time can be captured, not photos directly uploaded to the album) as the image to be identified. It then identifies the image to be identified to obtain several facial feature information. The facial feature information and the corresponding platform ID to be identified are compared with the data in the identity verification database 021. If the comparison is successful, the user preference classification strategy 012 is entered; otherwise, the process ends. This strategy implements real-name registration for comments, significantly increasing the risk of fake positive reviews and effectively preventing users from engaging in such behavior.

[0038] The user preference classification strategy 012 obtains product tags (such as mobile phones, clothing, eye makeup, foundation, cameras, snacks, etc.) of the product to be reviewed as matching tags. Based on these matching tags, it searches the user feature database 022 to obtain corresponding user feature information as fill-in feature information. It then obtains subjective feature information input by the user based on this fill-in feature information, identifies subjective feature keywords, and searches the preference classification database 023 based on these keywords to obtain corresponding preference groups as actual preference groups, which are then used in the review authenticity identification strategy 013. Different types of users have different priorities for the same product; if all reviews of the same product are judged for authenticity in the same way, the accuracy of the judgment is low. This invention classifies reviews based on user feature information determined by product tags and compares reviews with the same priorities for the same product. This selective grouping improves the accuracy of authenticity judgment.

[0039] The aforementioned review authenticity identification strategy 013 calculates a real-time deviation value by comparing the deviation between a review submitted by a platform ID for a product to be reviewed and other reviews of the same product within the same actual preference group. Users with the same user characteristics should have similar perceptions of the same product. If a user submits a positive review after engaging in fraudulent order placement, they will not know the true user experience because they have not actually used the product. Therefore, their review will differ significantly from genuine reviews. Fraudulent positive reviews can be identified through this difference comparison.

[0040] Reviews of similar products commented on by the platform ID are obtained as historical reviews. A historical cumulative deviation value is calculated based on these historical reviews. The deviation value between the historical reviews of this platform ID and other reviews of the same product is obtained using the same method as calculating the real-time deviation value (again, preference groups are categorized), thus determining the authenticity of the historical reviews of this platform ID. A new historical cumulative deviation value is obtained by updating the historical cumulative deviation value and the real-time deviation value. A true value is then obtained based on the new historical cumulative deviation value. If the true value is less than or equal to a preset true threshold, the corresponding review is marked as a genuine review. If the true value is greater than the preset true threshold, the corresponding review is marked as a fake review, and the positive review rate scoring strategy 015 is entered.

[0041] The positive review rate scoring strategy 015 calculates the following: If a genuine review is received, a real-time positive review rate is obtained based on the content of the genuine review. A new historical positive review rate is then obtained by updating the historical positive review rate based on the real-time positive review rate. If a fake review is received, a reputation depreciation value is generated. A new historical reputation value is obtained by updating the historical reputation value based on the reputation depreciation value, with the updated historical reputation value being lower than the previous historical reputation value. If the historical reputation value is lower than a preset reputation threshold, the historical positive review rate is reset to the initial positive review rate. The final historical positive review rate obtained by this invention is the sum of all genuine reviews for this product. The final historical positive review rate accurately reflects the quality of the corresponding product and can provide valuable reference for subsequent consumers.

[0042] As another example, in the authentication strategy 11, multi-angle photos captured in real time by a camera are obtained as the image to be identified.

[0043] As another example, in the user preference classification strategy 012, if the product label is beauty or clothing, the facial feature information (such as skin tone, face shape, etc.) is obtained as objective feature information. Objective feature keywords are obtained by identifying the objective feature information, and the actual preference group is determined based on the objective feature keywords and the subjective feature keywords. In the beauty or clothing field, the user's own user feature information has a significant impact on the user's evaluation of the product (compatibility sometimes has a greater impact than quality), especially the user's facial feature information. Furthermore, people often have biases in their perception of their own facial features, and cannot obtain accurate information subjectively. This invention obtains the facial feature information obtained during identity verification strategy 11, thereby objectively obtaining the user's facial feature information and ultimately improving the accuracy of authenticity judgment.

[0044] As another example, in the comment authenticity identification strategy 013, comments submitted by the platform ID for the product to be reviewed are obtained as comments to be identified. The comments to be identified are marked as short reviews, medium reviews, or long reviews according to the number of words. Historical comments submitted by the platform ID that are similar to the product to be reviewed are obtained. The text repetition rate between the comments to be identified and the historical comments is calculated to obtain a first repetition value. Other comments within the same actual preference group for the same product to be reviewed are obtained as reference comments. The text repetition rate between the comments to be identified and the reference comments is calculated to obtain a second repetition value.

[0045] If the comment to be identified is a short comment, the first difference and the second difference are obtained by obtaining the difference between the preset first repetition threshold and the first repetition value and the second repetition value, respectively. The true value is obtained based on the first difference, the second difference and the historical cumulative deviation value.

[0046] If the review to be identified is a neutral review, then the difference between the preset second repetition threshold and the first repetition value and the second repetition value are respectively used to obtain the third difference and the fourth difference. The true value is obtained based on the third difference, the fourth difference and the historical cumulative deviation value.

[0047] If the comment to be identified is a long comment, the fifth difference and the sixth difference are obtained by obtaining the difference between the preset third repetition threshold and the first repetition value and the second repetition value, respectively. The true value is obtained based on the fifth difference, the sixth difference and the historical cumulative deviation value.

[0048] The first repetition threshold is greater than the second repetition threshold, which in turn is greater than the third repetition threshold. As the number of words in a comment increases, the higher the text repetition rate, the greater the likelihood of a fake comment.

[0049] As another example, the merchant scoring system also includes a review matching database, which stores several product tags and corresponding review keywords;

[0050] In the comment authenticity identification strategy 013, comments submitted by the platform ID for the product to be commented are taken as comments to be matched. The comments to be matched are identified to obtain comment keywords as real-time comment words. The real-time comment words are matched in the comment matching database to obtain the comment matching degree. The true value is obtained based on the comment matching degree and the historical cumulative deviation value.

[0051] As another example, the merchant scoring system also includes a contact list database 024, which stores platform IDs and other platform IDs that have social communication with the platform IDs. The merchant scoring system also includes a sharing chain identification strategy 014.

[0052] The sharing chain identification strategy 014 generates a sharer field. If a platform ID is obtained based on the other platform ID entered in the sharer field, it is used as the sharing source ID. The platform ID and the sharing source ID are entered into the contact list database 024 for matching. Then, the platform ID and the sharing source ID form a sharing chain. The sharing seed value is obtained based on the chain length of the sharing chain.

[0053] In the positive review rate scoring strategy 015, the real-time positive review rate is obtained based on the content of the real reviews and the sharing value. If a product is of good quality, according to people's sharing logic, they will be willing to share it with their friends. Furthermore, only if that friend also agrees with the product's quality will they continue to share it with another friend. Therefore, the longer the sharing chain of a product, the more guaranteed its quality. Positive reviews from the same number of people are significantly more reliable than those from people with no social interaction, and the resulting historical positive review rate is closer to the product's true quality.

[0054] In the sharing chain identification strategy 014, the sharing duration is obtained by obtaining the chat duration of the platform ID and the sharing source ID, and the sharing seed value is obtained based on the sharing duration and the chain length of the sharing chain.

[0055] As another example, in the sharing chain identification strategy 014, chat content containing the platform ID and the sharing source ID is used as the sharing content. Keywords related to product tags are identified from the sharing content and used as sharing keywords. The frequency of these sharing keywords is calculated to obtain the sharing frequency. The sharing recommendation value is obtained based on the sharing frequency, the sharing duration, and the length of the sharing chain. The sharing frequency clearly reflects user approval of the product and more accurately reflects product quality.

[0056] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A merchant scoring system based on socialized product review feedback, characterized in that, It includes an identity verification database (021), a user feature database (022), and a preference classification database (023). The identity verification database (021) stores several platform IDs and corresponding ID card photos. The user feature database (022) stores several product tags and corresponding user feature information. The user feature information reflects factors that can influence the user's comments on products with corresponding product tags. The preference classification database (023) stores several feature keywords and corresponding preference groups. The merchant scoring system also includes an identity verification strategy (011), a user preference classification strategy (012), a review authenticity identification strategy (013), and a positive review rate scoring strategy (015); The identity verification strategy (011) obtains multi-angle photos captured in real time by the camera as images to be identified, identifies the images to be identified to obtain several facial feature information, compares the facial feature information and the corresponding platform ID to be identified with the data in the identity verification database (021), if the comparison is successful, it enters the user preference classification strategy (012), if the comparison is unsuccessful, it ends. The user preference classification strategy (012) obtains the product tag of the product to be reviewed as the matching tag, searches the user feature database (022) according to the matching tag to obtain the corresponding user feature information as the feature information to be filled, obtains the subjective feature information input by the user according to the feature information to be filled, identifies the subjective feature information to obtain subjective feature keywords, searches the preference classification database (023) according to the subjective feature keywords to obtain the corresponding preference group as the actual preference group, and enters the review authenticity identification strategy (013); The comment authenticity identification strategy (013) calculates the deviation value between the comment submitted by the platform ID for the product to be commented and other comments for the same product in the same actual preference group to obtain the real-time deviation value, obtains the comments of the same type of products commented by the platform ID in the platform as historical comments, obtains the historical cumulative deviation value based on the historical comments, updates the historical cumulative deviation value based on the historical cumulative deviation value and the real-time deviation value to obtain the new historical cumulative deviation value, obtains the true value based on the new historical cumulative deviation value, if the true value is less than or equal to the preset true threshold, marks the corresponding comment as a real comment, if the true value is greater than the preset true threshold, marks the corresponding comment as a fake comment, and enters the positive review rate scoring strategy (015); The positive review rate scoring strategy (015) is as follows: if a real review is obtained, a real-time positive review rate is obtained based on the content of the real review; and a new historical positive review rate is obtained by updating the historical positive review rate and the real-time positive review rate. If a false review is received, a reputation depreciation value is generated. A new historical reputation value is obtained by updating the historical reputation value based on the historical reputation value and the reputation depreciation value. The updated historical reputation value is less than the historical reputation value before the update. If the historical reputation value is lower than a preset reputation threshold, the historical positive review rate is reset to the initial positive review value. In the comment authenticity identification strategy (013), comments submitted by the platform ID for the product to be reviewed are obtained as comments to be identified. The comments to be identified are marked as short reviews, medium reviews, or long reviews according to the number of words. Historical comments of the same type as the product to be reviewed submitted by the platform ID are obtained. The text repetition rate between the comments to be identified and the historical comments is calculated to obtain a first repetition value. Other comments for the same product to be reviewed within the same actual preference group are obtained as reference comments. The text repetition rate between the comments to be identified and the reference comments is calculated to obtain a second repetition value. If the comment to be identified is a short comment, the first difference and the second difference are obtained by obtaining the difference between the preset first repetition threshold and the first repetition value and the second repetition value, respectively. The true value is obtained based on the first difference, the second difference and the historical cumulative deviation value. If the review to be identified is a neutral review, then the difference between the preset second repetition threshold and the first repetition value and the second repetition value are respectively used to obtain the third difference and the fourth difference. The true value is obtained based on the third difference, the fourth difference and the historical cumulative deviation value. If the comment to be identified is a long comment, the fifth difference and the sixth difference are obtained by obtaining the difference between the preset third repetition threshold and the first repetition value and the second repetition value, respectively. The true value is obtained based on the fifth difference, the sixth difference and the historical cumulative deviation value. The first repetition threshold is greater than the second repetition threshold, which is greater than the third repetition threshold.

2. The merchant scoring system based on socialized product review feedback as described in claim 1, characterized in that, In the user preference classification strategy (012), if the product label is beauty or clothing, the facial feature information is obtained as objective feature information, the objective feature information is identified to obtain objective feature keywords, and the actual preference group is determined based on the objective feature keywords and the subjective feature keywords.

3. The merchant scoring system based on socialized product review feedback as described in claim 2, characterized in that, The objective feature information includes skin color.

4. The merchant scoring system based on socialized product review feedback as described in claim 2, characterized in that, The objective feature information includes facial features.

5. The merchant scoring system based on socialized product review feedback as described in claim 1, characterized in that, The merchant scoring system also includes a review matching database, which stores several product tags and corresponding review keywords; In the comment authenticity identification strategy (013), the comments submitted by the platform ID for the product to be commented are taken as the comments to be matched. The comments to be matched are identified to obtain the comment keywords as real-time comment words. The comments are matched in the comment matching database according to the real-time comment words to obtain the comment matching degree. The true value is obtained according to the comment matching degree and the historical cumulative deviation value.

6. The merchant scoring system based on socialized product review feedback as described in claim 1, characterized in that, The merchant scoring system also includes a contact list database (024), which stores platform IDs and other platform IDs that have social communication with the platform IDs. The merchant scoring system also includes a sharing chain identification strategy (014). The sharing chain identification strategy (014) generates a sharer field. If a platform ID is obtained, other platform IDs entered in the sharer field are used as the sharing source ID. The platform ID and the sharing source ID are entered into the contact list database (024) for matching. The platform ID and the sharing source ID form a sharing chain. The sharing seed value is obtained according to the chain length of the sharing chain. In the positive review rate scoring strategy (015), the real-time positive review rate is obtained based on the content of the real reviews and the sharing and recommendation value.

7. The merchant scoring system based on socialized product review feedback as described in claim 6, characterized in that, In the sharing chain identification strategy (014), the sharing duration is obtained by obtaining the chat duration of the platform ID and the sharing source ID, and the sharing seed value is obtained based on the sharing duration and the chain length of the sharing chain.

8. The merchant scoring system based on socialized product review feedback as described in claim 7, characterized in that, In the sharing chain identification strategy (014), the chat content obtained by the platform ID and the sharing source ID is used as the sharing content. The keywords related to the product tag are obtained by identifying the sharing content as the sharing keywords. The frequency of the sharing keywords is counted to obtain the sharing frequency. The sharing seed value is obtained based on the sharing frequency, the sharing duration and the chain length of the sharing chain.

Images