A method for delivering product information on a digital shopping platform

By analyzing the semantic similarity and user behavior data of user search text and product information text, and calculating product placement priorities, the problem of mismatch between products and user needs in digital shopping platforms is solved, and the effectiveness of recommendations and user satisfaction are improved.

CN118521385BActive Publication Date: 2024-11-19GUANGZHOU SMARTGO TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410971696.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-19
Estimated Expiration
2044-07-19

AI Technical Summary

Technical Problem

The existing digital shopping platform's product information delivery methods have poor effectiveness, resulting in mismatch between products and user needs, excessive ineffective browsing, affecting user satisfaction and purchase conversion rate.

Method used

By analyzing the semantic similarity between user search text and product information text, obtaining demand similarity and user-triggered search redundancy, combining the number of views and communication record characteristics, the priority of product placement is calculated to improve the accuracy of recommendations and User satisfaction.

Benefits of technology

It realizes accurate understanding of user needs and matching evaluation of product information, improves the effectiveness of product recommendations and user satisfaction, reduces invalid browsing, and improves purchase conversion rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118521385B_ABST
    Figure CN118521385B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of shopping information processing technology, and specifically to a method for placing product information on a digital shopping platform. The invention obtains the user-triggered search redundancy based on the semantic similarity distribution between user search texts within the neighborhood of each search and the time-varying characteristics of the user search texts; obtains the demand matching degree of each type of product information under each search based on the semantic relevance and demand similarity between the user search text and the product information text; obtains the user browsing attention based on the number of views of each type of transaction product within different transaction time ranges, and the semantic similarity between each type of transaction product and the browsed product; obtains the user's decision-making strength for each type of transaction product based on the communication record characteristics of each type of transaction product; and then obtains the placement priority of each type of product information under each search. The present invention improves the effectiveness of platform recommendation of products by obtaining the accurate placement priority of each type of product information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of shopping information processing, and in particular to a method for delivering commodity information on a digital shopping platform. Background Art

[0002] Product information delivery on digital shopping platforms refers to the process of displaying product information to users on e-commerce platforms. Product information delivery on digital shopping platforms covers multiple aspects such as product uploading, display, promotion and personalized recommendation, aiming to increase consumers' awareness of and willingness to buy products.

[0003] In the prior art, products that meet users' search needs are placed based on traditional search engines and recommendation algorithms according to product information on shopping platforms. However, since the product information on the platform may be falsely advertised or overly beautified, and the advertisements and recommendation information are too dense and messy, traditional recommendation algorithms may not be able to accurately capture the real needs of users, and there may be a situation where the recommended products do not match the users' needs, resulting in poor effectiveness of product placement and excessive invalid browsing, which affects the complexity of product information selection. Summary of the invention

[0004] In order to solve the technical problem of poor effectiveness of commodity placement, the purpose of the present invention is to provide a commodity information placement method for a digital shopping platform, and the technical solution adopted is as follows:

[0005] The present invention proposes a method for placing commodity information on a digital shopping platform, the method comprising:

[0006] Obtaining product information text and user behavior data corresponding to the user search text in each search on the digital shopping platform; the user behavior data includes the number of product browsing times, browsing time, transaction time and communication records;

[0007] Analyze the semantic similarity between different texts to obtain the demand similarity between each type of product information text and the user search text in each search; obtain the user-triggered search redundancy based on the semantic similarity distribution between the user search texts within the neighborhood of each search and the browsing time change characteristics of the user search texts; obtain the semantic similarity between the user search texts within the neighborhood of each search, and the semantic similarity between the product information texts, and obtain the demand matching degree of each type of product information in each search based on the two semantic similarities, the user-triggered search redundancy and the demand similarity;

[0008] According to the number of views of each type of transaction product within different transaction time ranges, and the semantic similarity between each type of transaction product and the browsed product, the user's browsing attention is obtained; according to the user's browsing attention and the communication record characteristics of each type of transaction product, the user's decision-making strength for each type of transaction product is obtained;

[0009] According to the decision strength and the demand matching degree, the delivery priority of each category of product information in each search is obtained.

[0010] Furthermore, the analyzing the semantic similarity between different texts includes:

[0011] Obtain keywords of user search text and product information text;

[0012] According to the correlation between all keywords in different texts, the semantic similarity between corresponding texts is obtained.

[0013] Furthermore, obtaining the semantic similarity between corresponding texts according to the correlation between all keywords between different texts includes:

[0014] Get the word vector of the keyword, and calculate the cosine similarity between the word vectors of two keywords as the keyword similarity;

[0015] Calculate the mean of all keyword similarities between different texts as the semantic similarity between the corresponding texts.

[0016] Furthermore, the method for obtaining the user-triggered search redundancy includes:

[0017] In each search neighborhood, similar user search texts are screened out based on the semantic similarity distribution between user search texts;

[0018] The user-triggered search redundancy is obtained based on the number of search texts of similar users and the difference in browsing time between adjacent users' search texts. The number of search texts of similar users is positively correlated with the user-triggered search redundancy, and the difference in browsing time is negatively correlated with the user-triggered search redundancy.

[0019] Furthermore, the method for obtaining the search text of the same type of users includes:

[0020] In the neighborhood of each search, the mean of the semantic similarity between the user search texts is obtained as the average search similarity;

[0021] Sort the average search similarities between all user search texts from small to large, and obtain the difference between the average similarity of the next search and the average similarity of the previous adjacent search;

[0022] The next search average similarity corresponding to the largest difference is selected, and the user search text corresponding to all subsequent search average similarities is used as the search text of the same type of user.

[0023] Furthermore, the method for obtaining the degree of demand matching includes:

[0024] In the neighborhood of each search, the average of the average similarities between all user search texts is obtained as the user search similarity;

[0025] Obtain product information similarity based on the distribution trend of semantic similarity of product information texts between search texts of different users;

[0026] According to the consistency of changes between user search similarity and product information similarity, user-triggered search redundancy and the demand similarity, the matching degree of each type of product information in each search is obtained. The consistency of changes and user-triggered search redundancy are negatively correlated with the demand matching degree, and the demand similarity is positively correlated with the demand matching degree.

[0027] Furthermore, the method for obtaining the user's browsing attention includes:

[0028] For any transaction time range, obtain the ratio of the number of views of each type of transaction product to the number of views of all products as the first focus feature of each type of transaction product;

[0029] According to the semantic similarity of the corresponding product information text between each type of transaction product and each type of browsed product, the local similarity is obtained; the average of the local similarities between each type of transaction product and all browsed products is obtained as the second focus feature of each type of transaction product;

[0030] The user's browsing attention is obtained according to the first attention feature and the second attention feature of each type of transaction goods within different transaction time ranges.

[0031] Furthermore, the method for obtaining the decision strength includes:

[0032] For any type of transaction commodity, obtain the number of communications and the communication time in the communication records;

[0033] The user's browsing attention and the number of communications are combined to obtain the user's decision depth for each type of transaction product;

[0034] According to the difference between the transaction time and the last communication time of the corresponding goods, the decision depth is weighted to obtain the user's decision strength for each type of transaction goods.

[0035] Furthermore, the method for obtaining the delivery priority includes:

[0036] Obtain the average of the user's decision strength for all traded products as the overall decision strength level;

[0037] According to the overall decision strength level, the demand matching degree of each category of product information in each search is weighted to obtain the delivery priority of each category of product information in each search.

[0038] Furthermore, the method for obtaining the neighborhood range includes:

[0039] The range between the left and right adjacent transaction times corresponding to each search is used as the neighborhood range of each search.

[0040] The present invention has the following beneficial effects:

[0041] The present invention obtains the demand similarity between each category of product information text and the user search text in each search by analyzing the semantic similarity between different texts, accurately understands the user's search intention, and evaluates the similarity of product information to user needs; obtains the user-triggered search redundancy based on the semantic similarity distribution between user search texts within the neighborhood of each search and the time change characteristics of the user search texts, more accurately understands the user's search intention and interest changes, and evaluates the repetitiveness and ineffectiveness of user search behavior; obtains the semantic similarity between user search texts within the neighborhood of each search, as well as the semantic similarity between product information texts, and obtains the product information of each category in each search based on the two semantic similarities, the user-triggered search redundancy, and the demand similarity. The degree of matching of information demand is used to comprehensively evaluate the matching degree of each type of commodity information with user demand; according to the number of views of each type of transaction commodity within different transaction time ranges, and the semantic similarity between each type of transaction commodity and the browsed commodity, the user's browsing attention is obtained, and the real-time interest changes of users in different commodities are accurately captured, and the degree of attention of users' browsing behavior to different commodities is evaluated; according to the user's browsing attention and the communication record characteristics of each type of transaction commodity, the user's decision-making strength for each type of transaction commodity is obtained, and the user's decision-making process is more deeply understood, and the user's service needs are accurately grasped; and then the delivery priority of each type of commodity information under each search is obtained, and the commodity information with a higher delivery priority is determined to be displayed to the user first, so as to improve the user's satisfaction and purchase conversion rate. The present invention improves the effectiveness of platform recommendation of commodities by obtaining the accurate delivery priority of each type of commodity information. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 A flowchart of a method for delivering product information on a digital shopping platform provided by an embodiment of the present invention;

[0044] Figure 2 A flow chart of a method for obtaining search texts of similar users provided by an embodiment of the present invention;

[0045] Figure 3 A flow chart of a method for obtaining a degree of demand matching provided by an embodiment of the present invention;

[0046] Figure 4 A flow chart of a method for obtaining user browsing attention provided by an embodiment of the present invention;

[0047] Figure 5 A flow chart of a method for acquiring decision strength provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the method for placing product information on a digital shopping platform proposed by the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0049] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0050] The following is a detailed description of a method for delivering product information on a digital shopping platform provided by the present invention in conjunction with the accompanying drawings.

[0051] See also Figure 1 , which shows a flow chart of a method for placing product information on a digital shopping platform provided by an embodiment of the present invention. The specific method includes:

[0052] Step S1: Obtain the product information text and user behavior data corresponding to the user search text in each search on the digital shopping platform; the user behavior data includes the number of product browsing times, browsing time, transaction time and communication records.

[0053] In an embodiment of the present invention, in order to improve users' awareness and willingness to buy products, it is necessary to make personalized product recommendations to users based on their behavioral preferences, so that users can quickly find products that meet their needs; first, a large amount of user data is collected on a digital shopping platform, and the user search text corresponds to multiple categories of product information for each search; and the user's browsing and transaction status for each category of product information is recorded, and then the user's interests, preferred categories, purchasing habits, etc. are analyzed; the product information text and user behavior data corresponding to the user search text in each search on the digital shopping platform are obtained; the user behavior data includes the number of product browsing times, browsing time, transaction time and communication records.

[0054] It should be noted that in digital shopping platforms, product information text describes the text content of product name, key features and other information; by obtaining the communication records of user customer service, the number of user inquiry texts about product information in the communication records and the time information of each communication can be counted.

[0055] Step S2: Analyze the semantic similarity between different texts to obtain the demand similarity between each type of product information text and the user search text in each search; obtain the user-triggered search redundancy based on the semantic similarity distribution between user search texts within the neighborhood of each search and the time change characteristics of the user search texts; obtain the semantic similarity between user search texts within the neighborhood of each search, as well as the semantic similarity between product information texts, and obtain the demand matching degree of each type of product information in each search based on the two semantic similarities, the user-triggered search redundancy and the demand similarity.

[0056] Text content often contains information at multiple levels. The semantic similarity analysis between texts can deeply explore the internal connection of this information and better understand the characteristics of text content. Therefore, the semantic similarity between different texts is analyzed. User search text reflects the user's immediate or potential needs. Product information text provides a specific description and characteristics of the product. By analyzing the semantic similarity between texts, the user's search intention can be understood more accurately. When the similarity between the user search text and the product information text is high, it means that the product may be exactly what the user needs. Low correlation means that the product does not match the user's needs. Evaluate the degree of demand for product information for users. Obtain the demand similarity between each type of product information text and the user search text in each search.

[0057] Preferably, in one embodiment of the present invention, analyzing the semantic similarity between different texts includes:

[0058] Obtain keywords of user search text and product information text;

[0059] According to the correlation between all keywords in different texts, the semantic similarity between corresponding texts is obtained.

[0060] Preferably, in one embodiment of the present invention, obtaining the semantic similarity between corresponding texts according to the correlation relationship of all keywords between different texts includes:

[0061] Get the word vector of the keyword, and calculate the cosine similarity between the word vectors of two keywords as the keyword similarity;

[0062] Calculate the mean of all keyword similarities between different texts as the semantic similarity between the corresponding texts.

[0063] It should be noted that, in one embodiment of the present invention, keywords of user search text and product information text can be obtained based on the TF-IDF algorithm; word vectors of keywords can be obtained using a large language model; the specific TF-IDF algorithm and large language model are technical means well known to those skilled in the art and will not be elaborated here.

[0064] It should be noted that word vectors not only contain directional values. Similar words often have similar directions in the semantic space, and word vectors with different directions represent different semantics. Cosine similarity can capture the directional similarity between word vectors and analyze the correlation between keywords. In order to ensure the accuracy of calculation, for keywords with short word vectors, the length of the keyword involved in the calculation can be kept consistent by padding with 0. The specific large language model and cosine similarity are technical means well known to those skilled in the art and will not be elaborated here.

[0065] In one embodiment of the present invention, taking user search text and product information text as an example, the formula for semantic similarity is expressed as:

[0066] ,

[0067] in, Indicates product information text s and the user search text r The semantic similarity between Indicates the user search text r The j Keywords; Indicates product information text s The J Keywords; Indicates the user search text r The j Keywords and product information text s The J The keyword similarity between keywords; i A sequence number indicating the similarity of keywords; n Indicates the number of keyword similarities.

[0068] In the formula for semantic similarity, Indicates i The greater the similarity between keywords in user search text and product information text, the greater the correlation between keywords and the higher the semantic similarity between corresponding texts.

[0069] Based on this, for each search, the semantic similarity between each category of product information text and the user search text is obtained as the corresponding demand similarity, that is, ,in, Indicates a search for the next f The demand similarity between the product information text and the user search text; Indicates a search for the next f Product information text and user search text r The higher the semantic similarity, the higher the demand similarity, and the more likely it is that the product information meets the user's needs.

[0070] Within the neighborhood of each search, the semantic similarity distribution between user search texts reflects the continuity and consistency of user search behavior, can identify user groups with similar search patterns, and help to more deeply understand the user's search intentions and demand changes; the time variation characteristics of user search texts can reflect the user's search frequency and the continuity of search behavior. If the time variation of user search texts is smaller, that is, the user searches for products multiple times in a short period of time, it means that the user's demand for products is more urgent, which may also lead to an increase in search redundancy. Therefore, based on the semantic similarity distribution between user search texts within the neighborhood of each search and the time variation characteristics of user search texts, the user-triggered search redundancy is obtained.

[0071] Preferably, in one embodiment of the present invention, the method for obtaining user-triggered search redundancy includes:

[0072] In each search neighborhood, similar user search texts are screened out based on the semantic similarity distribution between user search texts;

[0073] The user-triggered search redundancy is obtained based on the number of search texts of similar users and the difference in browsing time between adjacent users' search texts. The number of search texts of similar users is positively correlated with the user-triggered search redundancy, and the difference in browsing time is negatively correlated with the user-triggered search redundancy.

[0074] Among them, a positive correlation means that the dependent variable will increase as the independent variable increases, and the dependent variable will decrease as the independent variable decreases. The specific relationship can be a multiplication relationship, an addition relationship, or the power of an exponential function, which is determined by actual application; a negative correlation means that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. It can be a subtraction relationship, a division relationship, etc., which is determined by actual application.

[0075] It should be noted that, in one embodiment of the present invention, the neighborhood range of each search is the range between the left and right adjacent transaction times corresponding to each search; in other embodiments of the present invention, the size of the neighborhood range can be set according to the specific circumstances, and is not limited or elaborated here.

[0076] The number of search texts by similar users reflects the user's continued attention to the product over a period of time. When the number of search texts by similar users is large, it may mean that the user has a high interest or demand for the product to be purchased, thus triggering search redundancy. Therefore, the search texts of similar users are selected for analysis.

[0077] Preferably, in one embodiment of the present invention, the method for obtaining the search text of the same type of user can refer to Figure 2 , which shows a flow chart of a method for obtaining search texts of similar users provided by an embodiment of the present invention, the method comprising:

[0078] Step S201: within the neighborhood range of each search, obtain the average value of the semantic similarity between the user search texts as the search average similarity.

[0079] By calculating the mean of the semantic similarity between user search texts, we can obtain a metric that reflects the overall similarity between user search texts, namely, the search average similarity. The larger the search average similarity, the more similar the user search texts are, and the more likely there is search redundancy.

[0080] Step S202: sorting the average search similarities between all user search texts from small to large, and obtaining the difference between the average similarity of the next search and the average similarity of the previous adjacent search.

[0081] By sorting the average search similarities between all user search texts, we can observe the similarity distribution. Calculating the difference can help identify points with large changes in similarity, which are often the boundaries of search texts of different categories of users.

[0082] Step S203: Select the next search average similarity corresponding to the maximum difference, and use the user search texts corresponding to all subsequent search average similarities as the same type of user search texts.

[0083] The point with the largest difference means that from this point on, there is a change in the search topic or intention, so this point can be used as the dividing line between different search categories; by selecting all the search texts after the point with the largest difference, the user search texts with similar search intentions are classified into one category, that is, the search texts of similar users.

[0084] In one embodiment of the present invention, the formula for user-triggered search redundancy is expressed as:

[0085] ,

[0086] in, R Indicates the redundancy of user-triggered searches; G Indicates the number of text searches by similar users; Indicates The browsing time of a user's search text; Indicates the adjacent The browsing time of a user's search text; m Indicates the number of browsing time differences between adjacent users' search texts; A sequence number indicating the difference in browsing time between text searches by adjacent users.

[0087] In the formula for user-triggered search redundancy, It means calculating the mean of browsing time differences between all adjacent users’ search texts, that is, the time variation trend. The larger the difference mean, the greater the time variation trend, the greater the time discreteness and the fewer search texts. On the contrary, the smaller the difference mean, the smaller the time variation trend and the more consecutive search text triggers. It represents the ratio of the number of search texts of similar users to the mean of the browsing time difference. The larger the ratio, the larger the number of search texts of similar users, the higher the number of similar user search texts, and the higher the redundancy of user-triggered searches. When the time change trend is smaller, there are more search triggers, the higher the redundancy of user-triggered searches, and the less it meets user needs.

[0088] It should be noted that in other embodiments of the present invention, the basic mathematical operation method of finding the difference after normalization may also be used, that is, , construct a correlation relationship in which the more similar user search texts there are, the smaller the time change trend is, and the higher the redundancy of user-triggered search is; the specific means are technical means well known to those skilled in the art and will not be elaborated here.

[0089] The changing trend of the semantic relevance between the user search text and the product information text reflects the similarity between the product information text and the user demand text. The closer the changing trend of the semantic relevance is, the greater the similarity is. The redundancy of user-triggered search indicates the number and frequency of users' searches on the same demand. The greater the redundancy of user-triggered search, the more likely it is that the user has a continuous and strong demand for a certain product or information, but may not have found a satisfactory product. Demand similarity indicates the degree to which product information meets the needs or expectations of users. Combining these three features to more comprehensively and accurately evaluate the demand matching degree of product information will help e-commerce platforms quickly screen out the products that best meet user needs from a large number of products, and improve the efficiency of search and recommendation. Therefore, the semantic similarity between user search texts and the semantic similarity between product information texts within the neighborhood of each search is obtained. Based on the two semantic similarities, user-triggered search redundancy, and demand similarity, the demand matching degree of each type of product information under each search is obtained.

[0090] Preferably, in one embodiment of the present invention, the method for obtaining the degree of demand matching is as follows: Figure 3 , which shows a flow chart of a method for obtaining a demand matching degree provided by an embodiment of the present invention, the method comprising:

[0091] Step S301: within the neighborhood of each search, obtain the average of the average search similarities between all user search texts as the user search similarity.

[0092] In order to quantify the similarity between user search texts in the search, that is, user search similarity, to reflect the concentration and consistency of user search intentions, high user search similarity indicates that the user group's search needs in the neighborhood are more concentrated, which is conducive to the recommendation system to provide more accurate results.

[0093] Step S302: obtaining the similarity of product information according to the distribution trend of the semantic similarity of product information texts between search texts of different users.

[0094] It should be noted that the first mean is obtained by calculating the mean of the semantic similarity of the product information texts between the user search texts; the product information similarity is obtained by calculating the mean of the first means between all user search texts; and by measuring the similarity of the product information texts between different user search texts, it helps to evaluate the accuracy and relevance of the search results.

[0095] Step S303: According to the consistency of changes between user search similarity and product information similarity, user-triggered search redundancy and demand similarity, the matching degree of each type of product information in each search is obtained. The consistency of changes and user-triggered search redundancy are negatively correlated with the demand matching degree, and the demand similarity is positively correlated with the demand matching degree.

[0096] Among them, a positive correlation means that the dependent variable will increase as the independent variable increases, and the dependent variable will decrease as the independent variable decreases. The specific relationship can be a multiplication relationship, an addition relationship, or the power of an exponential function, which is determined by actual application; a negative correlation means that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. It can be a subtraction relationship, a division relationship, etc., which is determined by actual application.

[0097] The changes between user search similarity and product information similarity reflect the concentration of user needs in the search and the difference in the relevance of search results, and reflect the overall matching level between user search text and product information text. The greater the change, the smaller the matching degree.

[0098] It should be noted that, in some embodiments of the present invention, the consistency of changes between user search similarity and product information similarity can be achieved by calculating the ratio or difference between the user search similarity and the product information similarity. The smaller the ratio or difference, the more consistent the changes in user search similarity and product information similarity, and the more likely a match is to be made; the larger the ratio or difference, the more inconsistent the changes in user search similarity and product information similarity, and the smaller the matching relationship.

[0099] In one embodiment of the present invention, the formula for the degree of demand matching is expressed as:

[0100] ,

[0101] in, Indicates a search for the next f The degree of demand matching of similar product information; It represents the mean of the average similarity between all user search texts in the neighborhood, that is, user search similarity; Indicates the similarity of product information; R Indicates the redundancy of user-triggered searches; Indicates f The demand similarity between the product information text and the user search text; Represents an exponential function with a natural constant as its base.

[0102] In the formula for demand matching, It represents the ratio between the user search similarity and the product information similarity, that is, the change. The larger the ratio, the greater the change, which means that when the user triggers multiple similar search texts, the product information similarity is smaller, that is, the same search demand gets different results, and the demand matching degree is smaller. For negative correlation mapping, the smaller the redundancy of user-triggered search, the greater the demand similarity, the smaller the ratio, and the greater the demand matching degree.

[0103] It should be noted that, in other embodiments of the present invention, The method of negative correlation mapping of the function can also be achieved by finding The reciprocal of is used to realize the negative correlation relationship in which the larger the value is, the smaller the matching degree is; when realizing the negative correlation relationship, if the reciprocal is used, it is necessary to add an artificially set threshold to the denominator of the fraction, such as a value of 0.01, to avoid the situation where the denominator of the fraction is 0. The specific means are well known to those skilled in the art and will not be described in detail here.

[0104] Step S3: Obtain the user's browsing attention based on the number of views of each type of transaction product within different transaction time ranges, and the semantic similarity between each type of transaction product and the browsed product; obtain the user's decision-making strength for each type of transaction product based on the user's browsing attention and the communication record characteristics of each type of transaction product.

[0105] The number of times a user browses a certain transaction product directly reflects the user's interest and attention to the product. The more times the user browses, the higher the user's demand or interest in the product. The higher the correlation between the transaction product and the browsed product, the more attention the user pays to the transaction product and the higher the user's browsing attention. The user's browsing attention is obtained based on the number of views of each type of transaction product within different transaction time ranges, and the semantic similarity between each type of transaction product and the browsed product.

[0106] Preferably, in one embodiment of the present invention, the method for obtaining the user's browsing attention degree can be found in Figure 4 , which shows a flow chart of a method for obtaining user browsing attention provided by an embodiment of the present invention, the method comprising:

[0107] Step S401: For any transaction time range, the ratio of the number of views of each type of transaction product to the number of views of all browsed products is obtained as the first focus feature of each type of transaction product.

[0108] The number of views is one of the important indicators to measure user interest. The more times a user views a transaction product, the higher the interest in the product, the more it meets user needs, and the higher the attention it pays to the product. By calculating the ratio of all views of each type of transaction product, we can find out the relative attention of users to different transaction products, thereby identifying users' preferences and interests.

[0109] It should be noted that, in one embodiment of the present invention, the transaction time range is a range between two adjacent transaction times.

[0110] Step S402: obtaining local similarity according to the semantic similarity of the corresponding product information text between each type of transaction product and each type of browsed product; obtaining the average of the local similarities between each type of transaction product and all browsed products as the second focus feature of each type of transaction product.

[0111] It should be noted that the local similarity is calculated by calculating the mean of the semantic similarity of the corresponding product information texts between each type of transaction products and each type of browsed products.

[0112] By comparing the semantic similarity between traded goods and browsed goods, we can understand the degree of correlation between the content browsed by users and traded goods. Semantic similarity can more accurately reflect users' understanding of the relationship between different goods. The greater the semantic similarity, the greater the similarity between goods. The more specific the search is for a specific product within the transaction time range, the greater the second attention feature.

[0113] Step S403: obtaining the user's browsing attention degree according to the first attention feature and the second attention feature of each type of transaction commodity within different transaction time ranges.

[0114] The greater the attention feature of the transaction product during the transaction period, the more the user tends to browse in a targeted manner, and the greater the user's browsing attention. In one embodiment of the present invention, the formula for the user's browsing attention is expressed as:

[0115] ,

[0116] in, Indicates the user's browsing attention; Indicates the serial number of the transaction commodity within the transaction time range; Indicates the trading time range Number of views of trading products; Indicates the number of views of all products within the transaction time range; Indicates the serial number of the product viewed within the transaction time range; Indicates the number of items viewed during the transaction time range.

[0117] In the formula of user browsing attention, Indicates the first The ratio of the number of views of the transaction product of this type to the total number of views is the first attention feature. The larger the ratio, the greater the first attention feature. The greater the number of views of a trading product, the greater the attention paid to the trading product; It represents the mean of the first concern feature of all trading commodities in all trading time ranges. The larger the mean, the greater the first concern feature of each type of trading commodity, and the greater the relationship degree of the trading commodity in each type of trading time range; Indicates the calculation Trading commodities and The mean value of the semantic similarity of keywords between similar browsed products, i.e., the local similarity. The larger the mean value, the greater the local similarity, and the more similar the transaction products and browsed products are. Indicates The mean of the local similarity between the transaction items and all browsed items, i.e. the second focus feature, It represents the mean of the second attention feature of all transaction commodities within all transaction time ranges. The larger the mean, the larger the second attention feature, that is, there are more similar commodities within the transaction time range. The more targeted the user's attention is to a category of commodities, the higher the user's browsing attention is.

[0118] The higher the user's browsing attention, the higher the user's interest and attention to the product; the communication record characteristics can fully describe the user's interaction with customer service during the purchase decision-making process, reflecting the user's acquisition, understanding and evaluation process of product information; when the user's browsing attention is high and the communication record is also more concerned about the product information, it can be considered that the user has a strong interest in the product and is willing to invest time and energy to obtain more information and conduct in-depth research in order to make a more informed purchase decision. Based on the user's browsing attention and the communication record characteristics of each type of transaction product, the user's decision-making strength for each type of transaction product is obtained.

[0119] Preferably, in one embodiment of the present invention, the method for obtaining the decision strength is as follows: Figure 5 , which shows a flow chart of a method for obtaining decision strength provided by an embodiment of the present invention, the method comprising:

[0120] Step S501: For any type of transaction commodity, obtain the communication frequency and communication time in the communication record.

[0121] The number of communications directly reflects the frequency of contact between users and customer service. A higher number of communications may mean that the user has a strong interest in the product; the communication time reflects the user's attention to the product.

[0122] Step S502: Integrate the user's browsing attention and communication times to obtain the user's decision depth for each type of transaction product.

[0123] User browsing attention measures the user's interest in the trading products. The more frequently the user communicates with the merchant's customer service, the more likely the user is to invest a lot of time and energy in researching the products of interest. The user also further confirms the product information and details through communication, thereby strengthening the user's determination to buy. The higher the user's decision-making depth on the trading products.

[0124] It should be noted that, in some embodiments of the present invention, the user's browsing attention and communication times may be integrated by multiplication or addition to obtain the user's decision depth for each type of transaction product.

[0125] Step S503: weighting the decision depth according to the difference between the corresponding commodity transaction time and the last communication time to obtain the user's decision strength for each type of transaction commodity.

[0126] The deeper the user's decision-making depth on the trading goods, the more time and energy the user invests in researching the goods, thus making a more informed purchasing decision; the difference between the corresponding product transaction time and the last communication time reflects the user's process of acquiring, understanding and confirming product information. The greater the difference, it may mean that the user has spent more time and energy in the decision-making process, conducted more in-depth product research and comparison, and the weaker the decision-making strength; the smaller the difference, the stronger the user's decision-making strength on the trading goods he is interested in.

[0127] In one embodiment of the present invention, the formula for decision strength is expressed as:

[0128] ,

[0129] ,

[0130] in, Indicates the user's The decision-making strength of similar trading commodities; Indicates the user's Decision depth of trading commodities; Indicates the user's browsing attention; Indicates Number of communications on similar trading products; Indicates the trading time of the product; Indicates the last communication time.

[0131] In the formula of decision-making strength, the greater the user's browsing attention, the more times the information of the transaction product is communicated, which means that the user seriously understands the product and is more likely to buy the product, and the user's decision depth on the transaction product is greater; It indicates the difference in time between the transaction and the last communication. The larger the difference, the longer the time between the transaction and the last communication. There may be other browsing behaviors, and the user's decision-making power is weaker. The smaller the difference, the shorter the time between the transaction and the last communication, and the stronger the user's decision-making power.

[0132] It should be noted that, in other embodiments of the present invention, other basic mathematical operations such as normalization and addition can also be used to construct a correlation relationship in which the greater the decision depth, the smaller the difference in time nodes, and the greater the decision strength. The specific means are technical means well known to those skilled in the art and will not be elaborated here.

[0133] Step S4: According to the decision strength and the degree of demand matching, the delivery priority of each category of product information in each search is obtained.

[0134] The decision-making strength of users on different trading products reflects the consistency of their purchasing habits and preferences. The greater the decision-making strength, the greater the user's interest in the product and the more likely they are to buy it. Product information with a high degree of demand matching is usually highly consistent with the user's search intent and can meet the user's needs, so it is displayed to the user first to improve the accuracy of the recommendation. Therefore, according to the user's decision-making strength on different trading products and the degree of demand matching, the priority of each type of product information is obtained for each search.

[0135] Preferably, in one embodiment of the present invention, the method for obtaining the delivery priority includes:

[0136] The average of the user's decision strength for all traded products is obtained as the overall decision strength level; based on the overall decision strength level, the demand matching degree of each category of product information in each search is weighted to obtain the delivery priority of each category of product information in each search.

[0137] In one embodiment of the present invention, the formula for delivery priority is expressed as:

[0138] ,

[0139] in, Indicates a search for the next f The priority of placing product information of this category; Indicates the user's The decision-making strength of similar trading commodities; Indicates a search for the next f The degree of demand matching of similar product information; W Indicates the quantity of traded goods; represents the hyperbolic tangent function.

[0140] In the formula for delivery priority, It represents the average of the user's decision-making strength for all trading commodities, that is, the overall decision-making strength level. The larger the average, the greater the overall decision-making strength level, indicating that the user has a higher decision-making strength for each type of trading commodity; It indicates the calculation of the overall decision-making strength level, which is the product of the demand matching degree of each type of product information in each search. The larger the product, the greater the overall decision-making strength level. f The higher the demand matching degree of the product information, the more important it is to f The product information of the first f The higher the priority of product information of this category, the higher the priority of product information.

[0141] It should be noted that in other embodiments of the present invention, a positive correlation can also be established by the addition method, that is, the greater the overall decision-making strength level, the greater the demand matching degree, and the greater the delivery priority; the specific means are technical means well known to those skilled in the art and will not be elaborated here.

[0142] It should be noted that, in one embodiment of the present invention, since the shopping platform will display a lot of product information under the user search text each time, some are relevant and some are not, but the relevant product information is definitely displayed first, so in order to reduce the amount of calculation and avoid analysis of unnecessary product information, the implementer can take the first 20 categories of product information in each search for analysis; in other embodiments of the present invention, the analysis type of product information can also be set according to the specific situation, which is not limited or elaborated here.

[0143] After obtaining the delivery priority of each type of product information under the user search text of each search, the user search text of the digital shopping platform is matched, and the products with higher delivery priorities will be displayed earlier. Product information with higher delivery priorities can better meet user needs, which can facilitate transaction completion more quickly to a certain extent, reduce users' invalid browsing, and improve the market exposure and user experience of the digital shopping platform.

[0144] In summary, the present invention analyzes the semantic similarity between different texts, obtains the demand similarity between each category of commodity information text and the user search text in each search; obtains the user-triggered search redundancy according to the semantic similarity distribution between the user search texts within the neighborhood of each search and the time change characteristics of the user search texts; obtains the demand matching degree of each category of commodity information in each search in combination with the change trend of the semantic similarity between the user search text and the commodity information text within the neighborhood of each search; obtains the user browsing attention according to the number of views of each category of transaction commodities within different transaction time ranges, and the semantic similarity between each category of transaction commodities and the browsed commodities; obtains the user's decision-making strength for each category of transaction commodities according to the user browsing attention and the communication record characteristics of each category of transaction commodities; obtains the placement priority of each category of commodity information in each search. The present invention improves the effectiveness of platform recommendation of commodities by obtaining the accurate placement priority of each category of commodity information.

[0145] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0146] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for delivering commodity information on a digital shopping platform, characterized in that: The method comprises: Obtaining product information text and user behavior data corresponding to the user search text in each search on the digital shopping platform; the user behavior data includes the number of product browsing times, browsing time, transaction time and communication records; Analyze the semantic similarity between different texts to obtain the demand similarity between each type of product information text and the user search text in each search; obtain the user-triggered search redundancy based on the semantic similarity distribution between the user search texts within the neighborhood of each search and the browsing time change characteristics of the user search texts; obtain the semantic similarity between the user search texts within the neighborhood of each search, and the semantic similarity between the product information texts, and obtain the demand matching degree of each type of product information in each search based on the two semantic similarities, the user-triggered search redundancy and the demand similarity; According to the number of views of each type of transaction product within different transaction time ranges, and the semantic similarity between each type of transaction product and the browsed product, the user's browsing attention is obtained; according to the user's browsing attention and the communication record characteristics of each type of transaction product, the user's decision-making strength for each type of transaction product is obtained; According to the decision strength and the demand matching degree, the delivery priority of each category of product information in each search is obtained.

2. The method for placing product information on a digital shopping platform according to claim 1, characterized in that: The analysis of semantic similarity between different texts includes: Obtain keywords of user search text and product information text; According to the correlation between all keywords in different texts, the semantic similarity between corresponding texts is obtained.

3. The method for placing product information on a digital shopping platform according to claim 2, characterized in that: The obtaining of semantic similarity between corresponding texts according to the correlation between all keywords between different texts includes: Get the word vector of the keyword, and calculate the cosine similarity between the word vectors of two keywords as the keyword similarity; Calculate the mean of all keyword similarities between different texts as the semantic similarity between the corresponding texts.

4. The method for placing product information on a digital shopping platform according to claim 1, characterized in that: The method for obtaining the user-triggered search redundancy includes: In each search neighborhood, similar user search texts are screened out based on the semantic similarity distribution between user search texts; The user-triggered search redundancy is obtained based on the number of search texts of similar users and the difference in browsing time between adjacent users' search texts. The number of search texts of similar users is positively correlated with the user-triggered search redundancy, and the difference in browsing time is negatively correlated with the user-triggered search redundancy.

5. The method for placing product information on a digital shopping platform according to claim 4, characterized in that: The method for obtaining the search text of the same type of users includes: In the neighborhood of each search, the mean of the semantic similarity between the user search texts is obtained as the average search similarity; Sort the average search similarities between all user search texts from small to large, and obtain the difference between the average similarity of the next search and the average similarity of the previous adjacent search; The next search average similarity corresponding to the largest difference is selected, and the user search text corresponding to all subsequent search average similarities is used as the search text of the same type of user.

6. A method for placing commodity information on a digital shopping platform according to claim 5, characterized in that: The method for obtaining the demand matching degree includes: In the neighborhood of each search, the average of the average similarities between all user search texts is obtained as the user search similarity; Obtain product information similarity based on the distribution trend of semantic similarity of product information texts between search texts of different users; According to the consistency of changes between user search similarity and product information similarity, user-triggered search redundancy and the demand similarity, the matching degree of each type of product information under each search is obtained, the consistency of changes and user-triggered search redundancy are negatively correlated with the demand matching degree, and the demand similarity is positively correlated with the demand matching degree; The method for obtaining the change consistency is to calculate the ratio of the user search similarity and the product information similarity, that is, ,in, Indicates the user search similarity, Indicates the similarity of product information.

7. The method for placing product information on a digital shopping platform according to claim 1, characterized in that: The method for obtaining the user's browsing attention includes: For any transaction time range, obtain the ratio of the number of views of each type of transaction product to the number of views of all products as the first focus feature of each type of transaction product; According to the semantic similarity of the corresponding product information text between each type of transaction product and each type of browsed product, the local similarity is obtained; the average of the local similarities between each type of transaction product and all browsed products is obtained as the second focus feature of each type of transaction product; The user's browsing attention is obtained according to the first attention feature and the second attention feature of each type of transaction goods within different transaction time ranges.

8. The method for placing product information on a digital shopping platform according to claim 1, characterized in that: The method for obtaining the decision strength includes: For any type of transaction commodity, obtain the number of communications and the communication time in the communication records; The user's browsing attention and the number of communications are combined to obtain the user's decision depth for each type of transaction product; According to the difference between the transaction time and the last communication time of the corresponding goods, the decision depth is weighted to obtain the user's decision strength for each type of transaction goods.

9. The method for placing product information on a digital shopping platform according to claim 1, characterized in that: The method for obtaining the delivery priority includes: Obtain the average of the user's decision strength for all traded products as the overall decision strength level; According to the overall decision strength level, the demand matching degree of each category of product information in each search is weighted to obtain the delivery priority of each category of product information in each search.

10. The method for placing commodity information on a digital shopping platform according to claim 1, characterized in that: The method for obtaining the neighborhood range includes: The range between the left and right adjacent transaction times corresponding to each search is used as the neighborhood range of each search.