Commodity evaluation label generation method and device

Abstract

The invention discloses a commodity evaluation label generation method and device. The method comprises the following steps: acquiring each piece of target comment information of which the emotion type is a target emotion type from each piece of comment information of a target commodity; matching the target comment information with each piece of comment information to obtain a first detection result of the target comment information, and matching the target comment information with each piece of historical comment information of a target user generating the target comment information to obtain a second detection result of the target comment information; determining a truth detection result of the target comment information according to the first detection result and the second detection result; and obtaining an evaluation label of the target commodity according to each piece of credible comment information of which the authenticity detection result is credible in each piece of target comment information. According to the commodity evaluation label generation method provided by the embodiment of the invention, the credibility of the commodity evaluation label can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and device for generating evaluation labels for commodities. Background Art

[0002] With the increasing popularity of online shopping, online shopping has been integrated into consumers' daily lives. Consumers purchase the goods they need through the Internet. When purchasing goods, consumers usually refer to the evaluation labels of the goods to measure whether the goods meet their needs.

[0003] In the related art, the generation of evaluation labels for a product is to mine all its comment information to extract keywords as the evaluation labels for the product or service. However, since there may be malicious comments on a product, such as fake orders to give good reviews or malicious bad reviews, the product may be filled with a large number of comment information that cannot reflect the user's true emotions, resulting in the evaluation labels being unable to reflect the user's true emotions, affecting the credibility of the evaluation labels. Summary of the invention

[0004] In view of the above problems, the present application provides a method and device for generating evaluation labels for commodities, which can improve the credibility of the evaluation labels for commodities.

[0005] In a first aspect, the present application provides a method for generating an evaluation label for a product, comprising: obtaining each target review information whose emotion type is a target emotion type from each review information of the target product; matching the target review information with each review information to obtain a first detection result of the target review information, and matching the target review information with each historical review information of a target user who generates the target review information to obtain a second detection result of the target review information; determining a authenticity detection result of the target review information based on the first detection result and the second detection result; obtaining an evaluation label of the target product based on each credible review information whose authenticity detection result is credible in each target review information; wherein the target emotion type includes a negative emotion type or a positive emotion type.

[0006] In the technical solution of the embodiment of the present application, each target comment information with a positive or negative emotion type is obtained from each comment information of the target product, and for any target comment information, the target comment information is matched with each comment information to obtain a first detection result of the target comment information, and the target comment information is matched with each historical comment information of the target user who generates the target comment information to obtain a second detection result of the target comment information, and after determining the authenticity detection result of the target comment information according to the first detection result and the second detection result, the evaluation label of the target product is obtained according to each credible comment information whose authenticity detection result is credible in each target comment information. Since malicious comments usually have similar comment contents, by using each comment information of the target product and the historical comment information of the target user who generates the target comment information, it is possible to screen out the comment information with a high probability that can reflect the user's true emotion from each target comment information that reflects the user's emotion to form the evaluation label of the product, thereby improving the credibility of the comment information used to generate the evaluation label, and then improving the credibility of the generated evaluation label.

[0007] In some embodiments, from each review information of the target product, each target review information whose sentiment type is the target sentiment type is obtained, including: performing semantic analysis on each review information to obtain the sentiment type corresponding to each review information; and obtaining the target review information whose sentiment type is the target sentiment type from each review information.

[0008] In some embodiments, semantic analysis is performed on each of the review information of the target product to obtain the emotion type corresponding to each of the review information, including: keyword extraction of the review information to obtain at least one keyword of the review information; similarity matching of each of the keywords with a set of emotion terms to determine the emotion type corresponding to each of the keywords; and determining the emotion type corresponding to the review information based on the number of keywords corresponding to the same emotion type.

[0009] In some embodiments, the target comment information is matched with each of the comment information to obtain a first detection result of the target comment information, including: according to the emotion type corresponding to the target comment information, obtaining each first comment information having the same emotion type as the target comment information from each of the comment information of the target product; matching each of the first comment information with the target comment information to determine each first target information in each of the first comment information that matches the target comment information; obtaining the first detection result of the target comment information according to the number of each of the first target information and the proportion of each of the first target information in each of the first comment information.

[0010] In some embodiments, the target comment information is matched with historical comment information of the target user that generates the target comment information to obtain a second detection result of the target comment information, including: according to the emotion type corresponding to the target comment information, obtaining corresponding second comment information with the same emotion type as the target comment information from the historical comment information of the target user that generates the target comment information; matching each second comment information with the target comment information to determine each second target information in each second comment information that matches the target comment information; obtaining the second detection result of the target comment information according to the number of each second target information and the proportion of each second target information in each second comment information.

[0011] In some embodiments, the authenticity detection result of the target comment information is determined based on the first detection result and the second detection result, including: obtaining an initial detection result of the target comment information based on the first detection result and the second detection result; when the initial detection result is that the comment information is credible, obtaining the authenticity detection result of the target comment information based on the target behavior trajectory of the target user when purchasing the target product.

[0012] In some embodiments, it also includes: when the initial detection result is that the comment information is unreliable, determining the initial detection result as the authenticity detection result of the target comment information.

[0013] In the second aspect, the present application provides a device for generating an evaluation label for a product, comprising: a comment information acquisition module, used to obtain each target comment information with a target emotion type from each comment information of the target product; a comment information detection module, used to match the target comment information with each comment information to obtain a first detection result of the target comment information, and to match the target comment information with each historical comment information of a target user who generates the target comment information to obtain a second detection result of the target comment information; a detection result determination module, used to determine a authenticity detection result of the target comment information based on the first detection result and the second detection result; an evaluation label generation module, used to obtain an evaluation label for the target product based on each credible comment information with the authenticity detection result being credible in each target comment information; wherein the target emotion type includes a negative emotion type or a positive emotion type.

[0014] In a third aspect, the present application provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the method in the implementation of the first aspect when executing the computer program.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, performs the method in the implementation of the first aspect.

[0016] In a fifth aspect, the present application provides a computer program product, which, when running on a computer, enables the computer to execute the method described in the first aspect or any optional implementation manner of the first aspect.

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Moreover, the same reference numerals are used throughout the drawings to represent the same components. In the drawings:

[0019] Figure 1 A first flow chart of a method for generating evaluation labels for commodities in some embodiments of the present application;

[0020] Figure 2 A second flow chart of a method for generating evaluation labels for commodities in some embodiments of the present application;

[0021] Figure 3 A schematic diagram of the structure of a device for generating evaluation labels for commodities in some embodiments of the present application;

[0022] Figure 4 This is a schematic diagram of the structure of an electronic device according to some embodiments of the present application. DETAILED DESCRIPTION

[0023] The following embodiments of the technical solution of the present application are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.

[0025] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is more than two, unless otherwise clearly and specifically defined.

[0026] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0027] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0028] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple paths" refers to more than two paths (including two paths).

[0029] With the increasing popularity of online shopping, online shopping has been integrated into consumers' daily lives. Consumers purchase the goods they need through the Internet. When purchasing goods, consumers usually refer to the review information of the goods to determine whether they need to buy the goods. When there is a large amount of review information for a product, the evaluation tags corresponding to the product are usually mined. Since the evaluation tags can basically represent the characteristics of the product, consumers can measure whether the product meets their needs based on the evaluation tags.

[0030] In the related art, the generation of evaluation labels for a certain product is to conduct information mining on all its comment information to extract keywords as the evaluation labels of the product or service. For example, all the comment information of the product is subjected to keyword extraction, and then the keywords with a frequency greater than the preset frequency are used as the evaluation labels of the product. However, since there may be malicious comments on a certain product, such as positive or negative reviews through fake orders, the product may be filled with a large amount of comment information that cannot reflect the user's true emotions, resulting in the evaluation labels formed also unable to reflect the user's true emotions, affecting the credibility of the evaluation labels.

[0031] In view of the above technical problems, an embodiment of the present application provides a method for generating an evaluation label for a product, which obtains each target comment information with a positive or negative emotion type from each comment information of the target product, and matches the target comment information with each comment information for any target comment information to obtain a first detection result of the target comment information, and matches the target comment information with each historical comment information of the target user who generates the target comment information to obtain a second detection result of the target comment information, and after determining the authenticity detection result of the target comment information according to the first detection result and the second detection result, obtain the evaluation label of the target product according to each credible comment information whose authenticity detection result is credible in each target comment information. Since malicious comments usually have similar comment contents, by using each comment information of the target product and the historical comment information of the target user who generates the target comment information, it is possible to screen out the comment information with a high probability that can reflect the user's true emotion from each target comment information that reflects the user's emotion to form the evaluation label of the product, thereby improving the credibility of the comment information used to generate the evaluation label, and then improving the credibility of the generated evaluation label.

[0032] The method and device for generating evaluation labels for commodities disclosed in the embodiments of the present application can be applied to terminal devices to generate evaluation labels for commodities. The terminal device can be any terminal device such as a mobile terminal, a desktop terminal, a vehicle-mounted terminal or a server. The server can be an independent server or a server cluster composed of multiple servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence sampling point devices.

[0033] According to some embodiments of the present application, a method for generating a product evaluation label is provided, and the method can be applied to the aforementioned terminal device. Figure 1 As shown, the method for generating the evaluation label of the product includes:

[0034] S101, obtaining target review information with a target sentiment type from each review information of the target product;

[0035] S102, matching the target comment information with each of the comment information to obtain a first detection result of the target comment information, and matching the target comment information with each of the historical comment information of the target user who generated the target comment information to obtain a second detection result of the target comment information;

[0036] S103, determining a truthfulness test result of the target comment information according to the first test result and the second test result;

[0037] S104, obtaining an evaluation label of the target product according to each credible review information whose authenticity detection result is credible in each target review information;

[0038] The target emotion type includes a negative emotion type or a positive emotion type.

[0039] In some embodiments, the target product may be an item or service for which an evaluation label needs to be generated. When the evaluation label of the target product needs to be generated, all the review information for the target product can be obtained from the database storing the review information of any product through the API interface. For any review information, the corresponding emotion type can be determined. Exemplarily, when commenting on the target product, the rating evaluation information of the service level of the target product is usually attached, such as 1-5 star rating evaluation information, and the rating evaluation information can be used to reflect the user's emotion towards the target product. If the rating evaluation information is 4-5 stars, it reflects that the user's emotion towards the target product is positive; if the rating evaluation information is 1-2 stars, it reflects that the user's emotion towards the target product is negative; if the rating evaluation information is 3 stars, it reflects that the user's emotion towards the target product is neutral. Therefore, for any review information, the emotion type of the review information can be determined according to the rating evaluation information corresponding to the review information. If the rating evaluation information corresponding to the review information is 4-5 stars, it can be determined that the emotion type of the review information is a positive emotion type.

[0040] Considering that malicious comments are usually good or bad reviews, there is a greater possibility of malicious comments appearing in positive and negative emotional type comments. At the same time, positive and negative emotional type comments can usually better reflect the advantages or disadvantages of the target product, so that the evaluation labels generated by using positive and negative emotional type comments can better reflect the characteristics of the product. Therefore, after obtaining the emotional type of each comment information, the comment information with positive or negative emotional type can be obtained from each comment information as the target comment information, so as to more specifically screen the comment information, improve the screening efficiency of the comment information, and at the same time improve the accuracy of the generated evaluation labels.

[0041] After obtaining each target review information, for any target review information, the target review information can be matched with each review information of the target product, such as matching the target review information with each review information of the target product by text similarity. If the similarity between a certain review information and the target review information reaches a preset similarity, such as 85%, it means that the review information matches the target review information.

[0042] After matching the target comment information with the comment information of the target product for text similarity, the first detection result of the target comment information can be obtained based on the number of comment information that matches the target comment information. If the number of comment information that matches the target comment information in each comment information reaches the first preset number, it means that the target comment information may be a repeated malicious comment, and at this time, it can be determined that the first detection result of the target comment information is unreliable; otherwise, it can be determined that the first detection result of the target comment information is credible. Among them, the first preset number can be set according to actual conditions. For example, the first preset number can be determined based on the total number of each comment information. The total number of each comment information is positively correlated with the first preset number, that is, the more the total number of comment information of the target product, the larger the first preset number. Exemplarily, the first preset number can be 1 / 4 of the total number of each comment information.

[0043] Alternatively, the first detection result of the target comment information can be obtained based on the proportion of the number of comment information that matches the target comment information. For example, the number of comment information that matches the target comment information can be divided by the total number of comment information to obtain the proportion of the number of comment information that matches the target comment information, and the proportion can be compared with the first preset proportion. If the proportion reaches the first preset proportion, it means that the target comment information may be a repeated malicious comment. At this time, it can be determined that the first detection result of the target comment information is that the comment information is unreliable; otherwise, it can be determined that the first detection result of the target comment information is that the comment information is credible. Among them, the first preset proportion can be set according to actual conditions. For example, the first preset proportion can be determined based on the total number of each comment information. Among them, the first preset proportion can be set according to actual conditions. Exemplarily, the total number of each comment information is negatively correlated with the first preset proportion, that is, the more the total number of comment information of the target product, the smaller the first preset proportion.

[0044] At the same time, the target user who generated the target comment information can be determined based on the target information to obtain the historical comment information of the target user, and the target comment information is matched with the historical comment information of the target user who generated the target comment information, such as matching the target comment information with the historical comment information of the target user by text similarity. If the similarity between a certain historical comment information and the target comment information reaches a preset similarity, such as 85%, it means that the historical comment information matches the target comment information.

[0045] After matching the target comment information with the target user's historical comment information for text similarity, the second detection result of the target comment information can be obtained based on the number of historical comment information that matches the target comment information. If the number of historical comment information that matches the target comment information reaches the second preset number among the historical comment information, it means that the target comment information may be a repeated malicious comment, and at this time, the second detection result of the target comment information can be determined to be unreliable; otherwise, the second detection result of the target comment information can be determined to be credible. Among them, the second preset number can be set according to the actual situation. The total number of each comment information is positively correlated with the second preset number and the number of each historical comment information.

[0046] Alternatively, the second detection result of the target comment information can be obtained based on the proportion of the number of historical comment information that matches the target comment information. For example, the number of historical comment information that matches the target comment information can be divided by the total number of historical comment information to obtain the proportion of the number of historical comment information that matches the target comment information, and the proportion can be compared with the second preset proportion. If the proportion reaches the second preset proportion, it means that the target comment information may be a repeated malicious comment, and at this time, it can be determined that the second detection result of the target comment information is unreliable; otherwise, it can be determined that the second detection result of the target comment information is credible. Among them, the second preset proportion can be set according to actual conditions. Exemplarily, the total number of historical comment information is negatively correlated with the second preset proportion, that is, the more the total number of historical comment information of the target user, the smaller the second preset proportion.

[0047] After obtaining the first detection result and the second detection result, if both the first detection result and the second detection result are credible, it can be determined that the authenticity detection result of the target comment information is credible; otherwise, it can be determined that the authenticity detection result of the target comment information is unreliable.

[0048] After obtaining the authenticity test results of each target review information, the target review information whose authenticity test results are credible can be obtained from each target review information as credible review information, and then keywords are extracted from each credible review information to obtain the evaluation label of the target product.

[0049] By obtaining each target comment information with a positive or negative emotion type from each comment information of the target product, and matching the target comment information with each comment information for any target comment information to obtain a first detection result of the target comment information, and matching the target comment information with each historical comment information of the target user who generated the target comment information to obtain a second detection result of the target comment information, after determining the authenticity detection result of the target comment information according to the first detection result and the second detection result, according to each credible comment information with a credible authenticity detection result in each target comment information, the evaluation label of the target product is obtained. Since malicious comments usually have similar comment contents, by using each comment information of the target product and the historical comment information of the target user who generated the target comment information, the comment information with a high probability of reflecting the user's true emotion can be screened out from each target comment information reflecting the user's emotion to form the evaluation label of the product, thereby improving the credibility of the comment information used to generate the evaluation label, and then improving the credibility of the generated evaluation label.

[0050] Taking into account that there may not be corresponding rating information for the review information, in order to determine the sentiment type of the review information so as to screen the review information, in some embodiments, each target review information whose sentiment type is the target sentiment type is obtained from each review information of the target product, including: performing semantic analysis on each of the review information to obtain the sentiment type corresponding to each of the review information; and obtaining the target review information whose sentiment type is the target sentiment type from each of the review information.

[0051] In some embodiments, after obtaining the review information of the target product, each review information can be input into a trained semantic model to obtain the emotion type corresponding to each review information. Emotion types include positive emotion types, negative emotion types, or neutral emotion types. Among them, for the training of the semantic model, each review sample with a determined actual emotion type can be input into the semantic model in turn for training. After the semantic model outputs the predicted emotion type of the review sample, each training is performed and the predicted emotion type is compared with the actual emotion type of the review sample. When the two are different, the gradient descent method is used to adjust the network parameters of the semantic model through error back propagation, and then the next training is performed until the predicted emotion type obtained by inputting the review sample into the semantic model each time is the same as the actual emotion type of the review sample, and the training of the semantic model is completed.

[0052] After obtaining the trained semantic model, the target product's review information can be input into the semantic model for semantic analysis to obtain the sentiment type corresponding to each review information, so as to obtain the target review information with a positive sentiment type or a negative sentiment type from each review information. In this way, the target review information used to participate in the generation of the target product's evaluation label can be screened out from the target product's review information, avoiding the situation where the target review information cannot be screened out for subsequent evaluation label generation due to the absence of corresponding grade evaluation information in the review information, thereby improving the reliability of evaluation label generation.

[0053] In order to improve the accuracy of the sentiment type of the acquired evaluation information, semantic analysis is performed on each of the review information of the target product to obtain the sentiment type corresponding to each of the review information, including:

[0054] Extracting keywords from the comment information to obtain at least one keyword of the comment information;

[0055] Perform similarity matching between each of the keywords and a set of sentiment terms to determine the sentiment type corresponding to each of the keywords;

[0056] The emotion type corresponding to the comment information is determined according to the number of keywords corresponding to the same emotion type.

[0057] In some embodiments, the terminal device may pre-store a set of emotion terms representing emotion types, wherein the set of emotion terms includes a plurality of emotion terms representing positive emotions and a plurality of emotion terms representing negative emotions.

[0058] After obtaining any review information of the target product, the review information can be keyword extracted to obtain all the keywords of the review information, and then these keywords are matched with the sentiment term set for similarity, and the sentiment term matching any keyword in the sentiment term set is determined, so that the sentiment type corresponding to the sentiment term matching the keyword is determined as the sentiment type corresponding to the keyword. If the keyword does not match the corresponding sentiment term in the sentiment term set, the keyword is considered to be a non-emotional keyword, and the sentiment type corresponding to the keyword can be determined to be a neutral sentiment.

[0059] After obtaining the emotion type corresponding to each keyword in the comment information, the number of keywords of different emotion types may be compared, and the emotion type with the most keywords may be selected as the emotion type corresponding to the comment information.

[0060] After obtaining the emotion type corresponding to each comment information, each target comment information with a positive emotion type or a negative emotion type can be obtained from each comment information to perform subsequent target comment information detection and evaluation label generation.

[0061] In order to improve the detection efficiency and accuracy of the target comment information, in some embodiments, the target comment information is matched with each of the comment information to obtain a first detection result of the target comment information, including:

[0062] According to the emotion type corresponding to the target review information, obtaining, from the review information of the target product, each first review information having the same emotion type as the target review information;

[0063] Matching each of the first comment information with the target comment information to determine each of the first target information that matches the target comment information in each of the first comment information;

[0064] A first detection result of the target comment information is obtained according to the quantity of each piece of the first target information and the proportion of each piece of the first target information in each piece of the first comment information.

[0065] Taking into account that the emotion types corresponding to similar malicious comments are usually the same, in some embodiments, for any target comment information, the emotion type corresponding to the target comment information can be first matched with the emotion type of each comment information of the target product, so as to obtain the comment information with the same corresponding emotion type as the target comment information from each comment information of the target product as the first comment information.

[0066] After obtaining each first comment information, each first comment information can be matched with the target comment information for text similarity, so that the first comment information that matches the target comment information among the first comment information is determined as the first target information. Since the target comment information only needs to be matched with each first comment information with the same sentiment type, the number of comments that need to be matched can be effectively reduced, thereby improving detection efficiency.

[0067] Considering that if the number of each first target information is used to determine whether the target comment information is a malicious comment, then when the target comment information appears in a period when there is less comment information on the target product, the number of each first target information is small, and it may appear that the target comment information is actually a malicious comment, but because the number of each first target information is less than the preset number, the target comment information is determined to be a non-malicious comment, resulting in a misjudgment of the first detection result of the target comment information. If the proportion of the number of each first target information is used to determine whether the target comment information is a malicious comment, then when the target comment information appears in a period when there is more comment information on the target product, the proportion of the number of each first target information will decrease, and it may appear that the target comment information is actually a malicious comment, but because the proportion of each first target information is less than the preset proportion, the target comment information is determined to be a non-malicious comment, resulting in a misjudgment of the first detection result of the target comment information. To this end, after obtaining each first target information that matches the target comment information, the number of each first target information and the proportion of each first target information in each first comment information can be obtained, and the number of each first target information is compared with the first preset number, and the proportion of each first target information in each first comment information is compared with the first preset proportion. If the number of each first target information is less than the first preset number, and the proportion of each first target information in each first comment information is less than the first preset proportion, it means that the target comment information is likely to be a normal comment, and at this time, the first detection result of the target comment information can be determined as the comment information is credible; otherwise, the second detection result of the target comment information can be determined as the comment information is uncredible.

[0068] When malicious comments appear in a period when there is less comment information on the target product, there is less comment information. If the target comment information is malicious information, the proportion of comment information identical to it will be relatively large. When malicious comments appear in a period when there is more comment information on the target product, there is more comment information. If the target comment information is malicious information, the number of comment information identical to it will increase. Therefore, the target comment information is jointly detected by the number of each first target information and the proportion of each first target information in each first comment information, so that the target comment information appearing in any period can be effectively detected, thereby improving the accuracy of the first detection result of the target comment information.

[0069] In order to further improve the detection efficiency and accuracy of the target comment information, in some embodiments, the target comment information is matched with each historical comment information of the target user who generated the target comment information to obtain a second detection result of the target comment information, including:

[0070] According to the emotion type corresponding to the target comment information, obtaining, from the historical comment information of the target user who generated the target comment information, respective second comment information having the same emotion type as the target comment information;

[0071] Matching each of the second comment information with the target comment information, and determining each of the second target information that matches the target comment information in each of the second comment information;

[0072] The second detection result of the target comment information is obtained according to the quantity of each piece of second target information and the proportion of each piece of second target information in each piece of second comment information.

[0073] In some embodiments, for any target comment information, the emotion type corresponding to the target comment information can be first matched with the emotion type of each historical comment information of the target user, so as to obtain the historical comment information with the same emotion type as the target comment information from the historical comment information of the target user as the second comment information.

[0074] After obtaining each second comment information, each second comment information can be matched with the target comment information for text similarity, so that the second comment information that matches the target comment information among the second comment information can be determined as the second target information. Since the target comment information only needs to be matched with each second comment information of the same sentiment type, and does not need to be matched with all historical comment information, the number of comments that need to be matched can be effectively reduced, thereby improving detection efficiency.

[0075] Considering that if the number of each second target information is used to determine whether the target comment information is a malicious comment, then when the target comment information appears in a period when the number of historical comment information of the target user is small, the number of each second target information is small. At this time, it may appear that the target comment information is actually a malicious comment, but because the number of each second target information is less than the preset number, the target comment information is determined to be a non-malicious comment, resulting in a misjudgment of the second detection result of the target comment information. If the proportion of the number of each second target information is used to determine whether the target comment information is a malicious comment, then when the target comment information appears in a period when the number of historical comment information of the target user is large, the proportion of the number of each second target information will decrease. At this time, it may appear that the target comment information is actually a malicious comment, but because the proportion of each second target information is less than the preset proportion, the target comment information is determined to be a non-malicious comment, resulting in a misjudgment of the second detection result of the target comment information. To this end, after obtaining each second target information that matches the target comment information, the number of each second target information and the proportion of each second target information in each second comment information can be obtained, and the number of each second target information is compared with the second preset number, and the proportion of each second target information in each second comment information is compared with the second preset proportion. If the number of each second target information is less than the second preset number, and the proportion of each second target information in each second comment information is less than the second preset proportion, it means that the target comment information is probably a normal comment, and at this time, the second detection result of the target comment information can be determined as the comment information is credible; otherwise, the second detection result of the target comment information can be determined as the comment information is uncredible.

[0076] The target comment information is detected together by the number of each second target information and the proportion of each second target information in each second comment information, so that the target comment information appearing in any period can be effectively detected, thereby improving the accuracy of the second detection result of the target comment information.

[0077] In some embodiments, after obtaining the first detection result and the second detection result, if the first detection result and the second detection result are consistent, the first detection result or the second detection result can be determined as the authenticity detection result of the target comment information. If the first detection result and the second detection result of the target comment information are both credible, the authenticity detection result of the target comment information can be determined to be credible. Otherwise, the authenticity detection result of the target comment information can be determined to be uncredible.

[0078] Considering that malicious comments may be occasional behaviors of users rather than habitual behaviors, in order to further improve the accuracy of the authenticity detection result of the target comment information, in some embodiments, determining the authenticity detection result of the target comment information according to the first detection result and the second detection result includes:

[0079] Obtaining an initial detection result of the target comment information according to the first detection result and the second detection result;

[0080] When the initial detection result shows that the review information is credible, the authenticity detection result of the target review information is obtained according to the target behavior trajectory of the target user when purchasing the target product;

[0081] The target behavior trajectory is determined according to an operation instruction associated with the target product and sent by a user terminal of the target user.

[0082] In some embodiments, after obtaining the first detection result and the second detection result, if the first detection result and the second detection result are consistent, the first detection result or the second detection result can be determined as the initial detection result of the target comment information. If the first detection result and the second detection result of the target comment information are both reliable, the initial detection result of the target comment information can be determined to be reliable; if the first detection result and the second detection result of the target comment information are both unreliable, or the first detection result or the second detection result is that the comment information is unreliable, the initial detection result of the target comment information can be determined to be unreliable.

[0083] After obtaining the initial detection result of the target comment information, if the initial detection result is that the comment information is unreliable, the initial detection result can be determined as the authenticity detection result of the target comment information; if the initial detection result is that the comment information is credible, the initial detection result is verified through the target behavior trajectory of the target user when purchasing the target product to obtain the authenticity detection result of the target comment information.

[0084] Exemplarily, if the initial detection result is that the comment information is credible, then each operation instruction initiated by the target user through his user terminal can be obtained, and each target operation instruction associated with the target product can be obtained from each operation instruction. Among them, the target operation instruction associated with the target product can be determined as the target operation instruction associated with the target product, including the keyword of the target product in each operation instruction, such as the search instruction for the keyword of the target product; determining the operation instruction including the product link of the target product in each operation instruction as the target operation instruction associated with the target product; and determining the operation instruction generated in the product interface of the target product in each operation instruction, such as the payment instruction, as the target operation instruction associated with the target product.

[0085] After obtaining each target operation instruction associated with the target product, the target operation instructions are sorted from morning to night according to the generation time of each target operation instruction, and the target behavior trajectory of the target user when purchasing the target product can be generated. If the request types of each target operation instruction are keyword search instruction, product link click instruction and payment instruction in order of time, the target behavior trajectory generated when the target user purchases the target product is search product-select product-pay.

[0086] After obtaining the target behavior trajectory of the target user when purchasing the target product, the target behavior trajectory can be compared with the historical behavior trajectory of the target user when purchasing other products. The historical behavior trajectory is determined according to the operation instructions associated with other products purchased by the target user. If the similarity between the target behavior trajectory and each historical behavior trajectory is less than the preset similarity, it means that the target behavior trajectory is an abnormal behavior trajectory that is different from the user's behavior habits. At this time, it can be determined that the authenticity detection result of the target comment information is that the comment information is unreliable; otherwise, it can be determined that the authenticity detection result of the target comment information is that the comment information is credible. This further improves the accuracy of the authenticity detection result of the target comment information.

[0087] After obtaining the authenticity test results of each target review information, each credible review information whose authenticity test results are credible review information can be obtained from each target review information, so as to perform clustering and keyword extraction through each credible review information to obtain the evaluation label of the target product.

[0088] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution in the present application will be described clearly and completely below. In some embodiments, Figure 2 As shown, the method for generating the evaluation label of the product includes:

[0089] S201, extracting keywords from any review information of a target product to obtain at least one keyword of the review information.

[0090] S202, performing similarity matching between each keyword and a set of sentiment terms to determine the sentiment type corresponding to each keyword.

[0091] S203: Determine the emotion type corresponding to the comment information according to the number of keywords corresponding to the same emotion type.

[0092] S204, according to the emotion type corresponding to each review information of the target product, obtain target review information with the emotion type of the target emotion type from each review information, wherein the target emotion type includes a negative emotion type or a positive emotion type.

[0093] S205 , according to the sentiment type corresponding to the target review information, obtaining from the review information of the target commodity respective first review information having the same corresponding sentiment type as the target review information.

[0094] S206, matching each first comment information with the target comment information, and determining each first target information in each first comment information that matches the target comment information.

[0095] S207, obtaining a first detection result of the target comment information according to the quantity of each piece of first target information and the proportion of each piece of first target information in each piece of first comment information.

[0096] S208, according to the emotion type corresponding to the target comment information, obtaining, from the credible historical comment information of each comment information of the target user who generated the credible target comment information of the comment information, each second comment information having the same corresponding emotion type as the credible target comment information of the comment information;

[0097] S209, matching the trusted second comment information of each comment information with the trusted target comment information of the comment information, and determining each second target information in the trusted second comment information of each comment information that matches the trusted target comment information of the comment information;

[0098] S210, obtaining a second detection result of the credible target comment information of the comment information according to the quantity of the credible second target information of each comment information and the quantity ratio of the credible second target information of each comment information in the credible second comment information of each comment information.

[0099] S211, determine whether the first detection result and the second detection result are both credible comment information; if so, execute S213; otherwise, execute S212.

[0100] S212, determining that the authenticity detection result of the target comment information is unreliable.

[0101] S213, obtaining a result of the authenticity detection of the target review information according to the target behavior trajectory of the target user when purchasing the target product, wherein the target behavior trajectory is determined according to the operation instruction associated with the target product and sent by the user terminal of the target user.

[0102] S214, obtaining an evaluation label of the target product according to each credible review information whose authenticity detection result is credible in each target review information.

[0103] Figure 3 The present application provides a schematic diagram of the structure of a device for generating evaluation labels for a product. It should be understood that the device is Figure 1 The method embodiment executed in corresponds to the method embodiment, and the steps involved in the aforementioned method can be executed. The specific functions of the device can be found in the description above. To avoid repetition, the detailed description is appropriately omitted here. The device includes at least one software function module that can be stored in a memory in the form of software or firmware or solidified in the operating system (OS) of the device. Specifically, the device is used to perform collision detection on a battery with a sensor on the surface. The device includes: a comment information acquisition module 210, which is used to obtain target comment information with a target emotion type from the comment information of the target product; a comment information detection module 220, which is used to match the target comment information with each of the comment information to obtain a first detection result of the target comment information, and to match the target comment information with each of the historical comment information of the target user who generated the target comment information to obtain a second detection result of the target comment information; a detection result determination module 230, which is used to determine the authenticity detection result of the target comment information according to the first detection result and the second detection result; an evaluation label generation module 240, which is used to obtain an evaluation label of the target product according to each of the target comment information whose authenticity detection result is credible; wherein the target emotion type includes a negative emotion type or a positive emotion type.

[0104] In the technical solution of the embodiment of the present application, each target comment information with a positive or negative emotion type is obtained from each comment information of the target product, and for any target comment information, the target comment information is matched with each comment information to obtain a first detection result of the target comment information, and the target comment information is matched with each historical comment information of the target user who generates the target comment information to obtain a second detection result of the target comment information, and after determining the authenticity detection result of the target comment information according to the first detection result and the second detection result, the evaluation label of the target product is obtained according to each credible comment information whose authenticity detection result is credible in each target comment information. Since malicious comments usually have similar comment contents, by using each comment information of the target product and the historical comment information of the target user who generates the target comment information, it is possible to screen out the comment information with a high probability that can reflect the user's true emotion from each target comment information that reflects the user's emotion to form the evaluation label of the product, thereby improving the credibility of the comment information used to generate the evaluation label, and then improving the credibility of the generated evaluation label.

[0105] According to some embodiments of the present application, the comment information acquisition module 210 is specifically used to: perform semantic analysis on each of the comment information to obtain the emotion type corresponding to each of the comment information; and obtain the target comment information whose emotion type is the target emotion type from each of the comment information.

[0106] According to some embodiments of the present application, the comment information acquisition module 210 is specifically used to: extract keywords from the comment information to obtain at least one keyword of the comment information; perform similarity matching on each of the keywords with a set of sentiment terms to determine the sentiment type corresponding to each of the keywords; and determine the sentiment type corresponding to the comment information based on the number of keywords corresponding to the same sentiment type.

[0107] According to some embodiments of the present application, the comment information detection module 220 is specifically used to: obtain, from the comment information of the target product, first comment information having the same emotion type as the target comment information according to the emotion type corresponding to the target comment information; match each first comment information with the target comment information to determine each first target information in each first comment information that matches the target comment information; and obtain a first detection result of the target comment information according to the number of each first target information and the proportion of each first target information in the number of each first comment information.

[0108] According to some embodiments of the present application, the comment information detection module 220 is specifically used to: obtain, according to the emotion type corresponding to the target comment information, each second comment information having the same emotion type as the target comment information from each historical comment information of the target user who generated the target comment information; match each second comment information with the target comment information to determine each second target information in each second comment information that matches the target comment information; and obtain a second detection result of the target comment information based on the number of each second target information and the proportion of each second target information in each second comment information.

[0109] According to some embodiments of the present application, the detection result determination module 230 is specifically used to: obtain an initial detection result of the target comment information based on the first detection result and the second detection result; and when the initial detection result is that the comment information is credible, obtain the authenticity detection result of the target comment information based on the target behavior trajectory of the target user when purchasing the target product.

[0110] According to some embodiments of the present application, the detection result determination module 230 is further used to: when the initial detection result is that the comment information is unreliable, determine the initial detection result as the authenticity detection result of the target comment information.

[0111] According to some embodiments of the present application, the clock rate is less than or equal to a preset rate, and the preset rate is determined according to computing resources currently available for data sampling.

[0112] According to some embodiments of the present application, Figure 4As shown, the present application provides an electronic device 300, including: a processor 301 and a memory 302, the processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanisms (not marked), the memory 302 stores a computer program executable by the processor 301, and when the computing device is running, the processor 301 executes the computer program to execute a method executed by an external terminal in any optional implementation method, for example: from each review information of the target product, obtain each target review information with a target emotional type; match the target review information with each of the review information to obtain a first detection result of the target review information, and match the target review information with each of the historical review information of the target user who generated the target review information to obtain a second detection result of the target review information; determine the authenticity detection result of the target review information based on the first detection result and the second detection result; obtain the evaluation label of the target product based on each of the target review information whose authenticity detection result is credible; wherein the target emotional type includes a negative emotional type or a positive emotional type.

[0113] The present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method in any of the aforementioned optional implementations is executed.

[0114] Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

[0115] The present application provides a computer program product. When the computer program product is run on a computer, the computer is enabled to execute the method in any optional implementation manner.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should all be included in the scope of the claims and specification of the present application. In particular, as long as there is no structural conflict, the various technical features mentioned in the various embodiments can be combined in any way. The present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions that fall within the scope of the claims.

Claims

1. A method for generating evaluation labels for commodities, It is characterized in that include: From each review information of the target product, obtain each target review information whose sentiment type is the target sentiment type; Matching the target comment information with each of the comment information to obtain a first detection result of the target comment information, and matching the target comment information with each of the historical comment information of the target user that generated the target comment information to obtain a second detection result of the target comment information; Determining a truthfulness test result of the target review information according to the first test result and the second test result; According to each of the target review information, the authenticity detection result of each of the credible review information is credible, obtaining the evaluation label of the target product; The target emotion type includes a negative emotion type or a positive emotion type.

2. The method according to claim 1, It is characterized in that From each review information of the target product, obtain each target review information with the sentiment type being the target sentiment type, including: Performing semantic analysis on each of the comment information to obtain the emotion type corresponding to each of the comment information; From each of the comment information, the target comment information whose emotion type is the target emotion type is obtained.

3. The method according to claim 2, It is characterized in that Performing semantic analysis on each of the review information of the target product to obtain the sentiment type corresponding to each of the review information includes: Extracting keywords from the comment information to obtain at least one keyword of the comment information; Perform similarity matching between each of the keywords and a set of sentiment terms to determine the sentiment type corresponding to each of the keywords; The emotion type corresponding to the comment information is determined according to the number of keywords corresponding to the same emotion type.

4. The method according to any one of claims 1 to 3, It is characterized in that Matching the target comment information with each of the comment information to obtain a first detection result of the target comment information includes: According to the emotion type corresponding to the target review information, obtaining, from the review information of the target product, each first review information having the same emotion type as the target review information; Matching each of the first comment information with the target comment information to determine each of the first target information that matches the target comment information in each of the first comment information; A first detection result of the target comment information is obtained according to the quantity of each piece of the first target information and the proportion of each piece of the first target information in each piece of the first comment information.

5. The method according to any one of claims 1 to 3, It is characterized in that Matching the target comment information with each historical comment information of the target user who generated the target comment information to obtain a second detection result of the target comment information includes: According to the emotion type corresponding to the target comment information, obtaining, from the historical comment information of the target user who generated the target comment information, respective second comment information having the same emotion type as the target comment information; Matching each of the second comment information with the target comment information, and determining each of the second target information that matches the target comment information in each of the second comment information; The second detection result of the target comment information is obtained according to the quantity of each piece of second target information and the proportion of each piece of second target information in each piece of second comment information.

6. The method according to claim 1, It is characterized in that Determining the authenticity detection result of the target review information according to the first detection result and the second detection result includes: Obtaining an initial detection result of the target comment information according to the first detection result and the second detection result; When the initial detection result indicates that the review information is credible, the authenticity detection result of the target review information is obtained according to the target behavior trajectory of the target user when purchasing the target product.

7. The method according to claim 6, It is characterized in that Also includes: When the initial detection result indicates that the comment information is unreliable, the initial detection result is determined as the authenticity detection result of the target comment information.

8. A device for generating evaluation labels for commodities, It is characterized in that include: A review information acquisition module, used to acquire target review information of a target sentiment type from the review information of the target product; A comment information detection module, used for matching the target comment information with each of the comment information to obtain a first detection result of the target comment information, and matching the target comment information with each of the historical comment information of the target user who generated the target comment information to obtain a second detection result of the target comment information; A detection result determination module, used to determine the authenticity detection result of the target comment information according to the first detection result and the second detection result; An evaluation label generation module, used to obtain an evaluation label of the target product according to each credible review information whose authenticity detection result is credible in each target review information; The target emotion type includes a negative emotion type or a positive emotion type.

9. An electronic device comprising a processor and a memory storing a computer program, It is characterized in that When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images