A sentiment understanding method based on product matching

Abstract

The present invention discloses a method for understanding emotions based on product matching, and relates to the technical field of product marketing. The present invention includes: importing product data and user behavior data in the background system; for product data, the background system quickly filters the text information in the product data through the knowledge graph to obtain the basic label information of the product; for user behavior, the background system quickly filters the user's purchase, order, and comment content to obtain the user's basic label; after obtaining all labels, the emotion label is processed; the generated more detailed emotion label is finally reviewed to determine whether the dimensional information aggregation of the emotion label is passed. The present invention presents the product in the form of emotions based on a large amount of data and rules, greatly simplifies the classification of product attributes, and facilitates users to have a better use experience in search scenarios and browsing scenarios; optimizes the process of understanding product emotions, makes the operation and maintenance process more efficient, and optimizes the efficiency of operation and maintenance.

Description

Technical Field

[0001] The present invention belongs to the technical field of product marketing, and in particular relates to a method for understanding emotions based on commodity matching. Background Art

[0002] Commodity matching plays a big role in marketing. First, the classification of commodity attributes is infinite. Among the current similar products, taking the sub-categories in a certain industry as an example, there are more than a thousand label contents, and the entire system has millions of basic labels. Users can hardly find the corresponding products without clear commodity attributes during use, resulting in traffic loss and decreased conversion rate. Moreover, the accuracy of the predefined emotional content matched with commodities is low. Among the current similar products, some manual methods are often used to supplement some emotional content on the labels, such as "cold" and "enthusiastic". These contents are often predetermined by certain rules. For example, "pure color" and "light color" commodities may be positioned as "cold"; while "bright color" and "large color block" commodities may be defined as "enthusiastic". The above rules cannot cover all commodities well. On the contrary, they may even give some non-emotional content to commodities.

[0003] The complexity of intelligent emotional management of commodities. Among the current similar products, there is no technology that can automatically manage the emotions of commodities. Intelligent emotional management of commodities is a method that manages the emotional diseases of commodities based on existing product descriptions, image recognition, outfit analysis, and semantic understanding, and then labels the commodities with corresponding emotions, so that users can search for corresponding commodities by emotions.

[0004] For different types of goods, describing them with a small amount of emotions is emotional, subjective, and complex. It is necessary to use a large amount of product data and user behavior as a basis to achieve user emotion understanding based on product data and ultimately determine the emotions expressed by the product; therefore, in response to the above problems, this technical solution provides an emotion understanding method based on product matching. Summary of the invention

[0005] The present invention provides a sentiment understanding method based on product matching, which solves the above problems.

[0006] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0007] The present invention provides an emotion understanding method based on product matching. The emotion understanding method is implemented based on a background system. The background system includes a tag engine, product tags, user tags, and a data module composed of product data and user behavior data. The emotion understanding method includes the following steps:

[0008] S01. Import product data and user behavior data into the backend system;

[0009] S02. For product data, the backend system quickly filters the text information in the product data through the knowledge graph to obtain basic product label information, and generates image labels for the image data in the product data through the image recognition engine;

[0010] S03. Based on the user behavior, the backend system quickly filters the user's purchase, order, and comment content to obtain the user's basic tags, and records the product information of the user's order as the matching product information tag;

[0011] S04, after obtaining all the labels, the emotion labels are processed, including the following steps:

[0012] S041. For the same source tags, use the tag engine to pre-generate all emotional tags;

[0013] S042. For the product information labels of matching products in the user behavior data, different products are matched with labels in the hope of obtaining more emotional labels;

[0014] S043. For different source labels, combined with dimension information, check the matching relationship between each emotion label and the basic label, and delete the conflicting emotion labels;

[0015] S05: Conduct a final review of the generated relatively detailed emotion tags to determine whether the dimension information aggregation of the emotion tags is passed, which specifically includes the following steps:

[0016] S051. The administrator randomly obtains several products for each emotion tag through the backend system and conducts the final review of the final emotion tag. If there is a situation where the review fails, it means there is a conflict. The emotion tag is deleted, reconstructed, and the model is optimized. The corresponding products are updated. If the review passes, the emotion tag will enter the tag system as a standard tag and realize the association between the product and the emotion;

[0017] S052. Administrators can manage emotion tags at any time through the background, adjust the relationship between tags and products in real time, operate quickly, and optimize user searches.

[0018] Furthermore, the product data includes product title, product description, product category, product annotation, product picture, product outfit, and product recommendation.

[0019] Furthermore, the user behavior data includes architecture records, order records, comment information, and matching product information.

[0020] Furthermore, the tag engine includes a text recognition engine capable of generating text tags, an image recognition engine capable of generating image tags, and a user behavior recognition engine capable of generating user tags.

[0021] Furthermore, the product label includes a basic product label and a picture label.

[0022] Furthermore, the user tags include user basic tags and matching product information tags.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. The present invention presents products in an emotional way based on a large amount of data and rules, greatly simplifies the classification of product attributes, and extracts products from hundreds of thousands of subcategories. This facilitates users to have a better experience in search and browsing scenarios;

[0025] 2. The present invention optimizes the process of understanding product emotions, making the operation and maintenance process more efficient. The process of labeling products in this method is not limited to full-text matching of labels. It combines multiple rules to complete information and uses product information, dressing information, user feedback, product pictures, comment pictures and other content to deeply understand the products. The most suitable emotional words are automatically analyzed through the model, which optimizes the efficiency of operation and maintenance, and saves a lot of operation and maintenance costs on both the basic labels and emotional labels of the products.

[0026] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0028] Figure 1 This is a schematic diagram of the structure of a backend system corresponding to a method for understanding emotions based on product matching according to the present invention;

[0029] Figure 2 This is a step diagram of a method for understanding emotions based on product matching according to the present invention;

[0030] Figure 3 for Figure 2 The detailed step diagram of step S05;

[0031] Figure 4 for Figure 2 The overall step schematic diagram. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] This method is mainly based on the fact that most retail companies adopt a simple hierarchical management for commodity management, and do not carry out anthropomorphic emotional management for commodities. Through the description of commodities and the analysis of dressing styles in the knowledge graph, based on the understanding of contextual semantics, emotional types are assigned to commodities to meet the requirements of commodity emotional management. Through this emotion, it can connect with people's actual mood at the moment to stimulate consumers' secondary consumption.

[0034] Referring to the existing commodity classification methods, most of them are based on commodity attributes, and commodity labels are classified and classified by predetermined dimensions. For example, commodity classification on platforms such as Taobao and JD.com. The core is to continuously subdivide commodity attributes and establish corresponding classification relationships to help users understand commodity attributes.

[0035] The technical problem that the present invention needs to solve is a method of emotion understanding based on product matching, which helps platform operators to quickly and accurately label products with emotions, so that consumers can search for products based on emotions. To this end, the present invention adopts a method of combining product data with user behavior to build a model training model to perform emotion understanding. At the same time, the present invention also provides a solution for product search and product recommendation for emotion understanding. Specifically, this technical solution is implemented through the following technologies:

[0036] See also Figure 1-4 As shown, the present invention provides an emotion understanding method based on product matching, which is implemented based on a background system. The background system includes a tag engine, product tags, user tags, and a data module composed of product data and user behavior data. The emotion understanding method includes the following steps:

[0037] S01. Import product data and user behavior data into the backend system;

[0038] S02. For product data, the backend system quickly filters the text information in the product data through the knowledge graph to obtain basic product label information, and generates image labels for the image data in the product data through the image recognition engine;

[0039] S03. Based on the user behavior, the backend system quickly filters the user's purchase, order, and comment content to obtain the user's basic tags, and records the product information of the user's order as the matching product information tag;

[0040] S04, after obtaining all the labels, the emotion labels are processed, including the following steps:

[0041] S041. For the same source tags, use the tag engine to pre-generate all emotional tags;

[0042] S042. For the product information labels of matching products in the user behavior data, different products are matched with labels in the hope of obtaining more emotional labels;

[0043] S043. For different source labels, combined with dimension information, check the matching relationship between each emotion label and the basic label, and delete the conflicting emotion labels;

[0044] For example, a product with the basic label "seaside" is basically impossible to be defined as "passionate mountains" at the same time. In this case, the system should actively remove the emotion label "passionate mountains".

[0045] S05: Conduct a final review of the generated relatively detailed emotion tags to determine whether the dimension information aggregation of the emotion tags is passed, which specifically includes the following steps:

[0046] S051. The administrator randomly obtains several products for each emotion tag through the backend system and conducts the final review of the final emotion tag. If there is a situation where the review fails, it means there is a conflict. The emotion tag is deleted, reconstructed, and the model is optimized. The corresponding products are updated. If the review passes, the emotion tag will enter the tag system as a standard tag and realize the association between the product and the emotion;

[0047] S052. Administrators can manage emotion tags at any time through the background, adjust the relationship between tags and products in real time, operate quickly, and optimize user searches.

[0048] Among them, product data includes product title, product description, product category, product annotations, product pictures, product outfits, and product recommendations.

[0049] Among them, user behavior data includes architecture records, order records, comment information, and matching product information.

[0050] Among them, the label engine includes a text recognition engine that can generate text labels, an image recognition engine that can generate image labels, and a user behavior recognition engine that can generate user labels.

[0051] Among them, product tags include basic product tags and image tags.

[0052] Among them, user tags include user basic tags and matching product information tags.

[0053] like Figure 1 As shown, the present invention uses a complete labeling system based on knowledge graphs, combined with image recognition and multi-dimensional analysis, to quickly define emotional labels for products.

[0054] Among them, through the entity relationship of the knowledge graph, sentiment labels can be defined quickly and directly according to the rules; image recognition can identify the picture information in the product data, feedback the corresponding labels, and define sentiment labels according to the knowledge graph; after the labels are produced, they will be aggregated according to the dimensions of the labels, further refine the product emotions, and optimize the accuracy of the emotions.

[0055] Beneficial effects:

[0056] 1. The present invention presents products in an emotional way based on a large amount of data and rules, greatly simplifies the classification of product attributes, and extracts products from hundreds of thousands of subcategories. This facilitates users to have a better experience in search and browsing scenarios;

[0057] 2. The present invention optimizes the process of understanding commodity sentiment, making the operation and maintenance process more efficient. The method is not limited to full-text matching of labels in the process of labeling commodities. It combines multiple rules to complete information, uses product information, outfit information, user feedback, commodity pictures, comment pictures and other contents to deeply understand commodities, and automatically analyzes the most suitable sentiment words through the model, thereby optimizing the efficiency of operation and maintenance, and saving a lot of operation and maintenance costs on both the basic labels and sentiment labels of commodities;

[0058] 3. Compared with manual definition, the emotion content defined by the model in the present invention is more accurate. The emotion labels defined by the model are not only based on simple rules. After an emotion label is generated, image recognition and dimension combination are often used to verify the output emotion label. The mutual verification and optimization between multiple models can continuously improve the accuracy of the emotion content and realize unattended and unsupervised work.

[0059] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A sentiment understanding method based on product matching, characterized in that: The emotion understanding method is implemented based on a background system, which includes a tag engine, product tags, user tags, and a data module consisting of product data and user behavior data; the emotion understanding method includes the following steps: S01. Import product data and user behavior data into the backend system; S02. For product data, the backend system quickly filters the text information in the product data through the knowledge graph to obtain basic product label information, and generates image labels for the image data in the product data through the image recognition engine; S03. Based on the user behavior, the backend system quickly filters the user's purchase, order, and comment content to obtain the user's basic tags, and records the product information of the user's order as the matching product information tag; S04, after obtaining all the labels, the emotion labels are processed, including the following steps: S041. For the same source tags, use the tag engine to pre-generate all emotional tags; S042. For the product information labels of matching products in the user behavior data, different products are matched with labels in the hope of obtaining more emotional labels; S043. For different source labels, combined with dimension information, check the matching relationship between each emotion label and the basic label, and delete the conflicting emotion labels; S05: Conduct a final review of the generated relatively detailed emotion tags to determine whether the dimension information aggregation of the emotion tags is passed, which specifically includes the following steps: S051. The administrator randomly obtains several products for each emotion tag through the backend system and conducts the final review of the final emotion tag. If there is a situation where the review fails, it means there is a conflict. The emotion tag is deleted, reconstructed, and the model is optimized. The corresponding products are updated. If the review passes, the emotion tag will enter the tag system as a standard tag and realize the association between the product and the emotion; S052. Administrators can manage emotion tags at any time through the background, adjust the relationship between tags and products in real time, operate quickly, and optimize user searches.

2. The emotion understanding method based on product matching according to claim 1, characterized in that: The product data includes product title, product description, product category, product annotation, product image, product outfit, and product recommendation.

3. The emotion understanding method based on product matching according to claim 1, characterized in that: The user behavior data includes architecture records, order records, comment information, and matching product information.

4. The emotion understanding method based on product matching according to claim 1, characterized in that: The tag engine includes a text recognition engine capable of generating text tags, an image recognition engine capable of generating image tags, and a user behavior recognition engine capable of generating user tags.

5. The emotion understanding method based on product matching according to claim 1, characterized in that: The product label includes a basic product label and a picture label.

6. The emotion understanding method based on product matching according to claim 1, characterized in that: The user tags include user basic tags and matching product information tags.

Images