A product selection method, device, equipment and storage medium

Abstract

The embodiment of the present application discloses a product selection method, device, equipment and storage medium, which constructs a product database based on product data in an online platform; based on the product database, screens the relevant data of the product data to obtain key product data; based on the key product data, determines trend product words; based on the trend product words, selects target products from a preset product pool. In this way, the accuracy of activity themes and product selection can be improved, and platform resource traffic can be saved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a product selection method, device, equipment and storage medium. Background Art

[0002] In related technologies, the method of generating e-commerce activity scene themes is mainly for operators to find marketing trends and fixed holiday theme selling points based on manual experience; product selection is mainly for operators to obtain products under the labels as a product pool by selecting brands, categories or industrial attribute product labels that are similar to the scene theme; in this way, the generation of activity themes and product selection methods are highly dependent on manual experience, which can easily cause a waste of platform resources and traffic. Summary of the invention

[0003] In order to solve the above technical problems, the embodiments of the present application provide a product selection method, device, equipment and storage medium, which build a product database based on the product data in the online platform; based on the product database, filter the relevant data of the product data to obtain key product data; based on the key product data, determine the trend product words; based on the trend product words, select the target product in the preset product pool. In this way, the accuracy of the activity theme and product selection can be improved, and the platform resource flow can be saved.

[0004] To achieve the above purpose, the technical solution of this application is implemented as follows:

[0005] The present application provides a product selection method, the method comprising:

[0006] Build a product database based on the product data in the online platform;

[0007] Based on the commodity database, relevant data of the commodity data is screened to obtain key commodity data;

[0008] Determine trending commodity words based on the key commodity data;

[0009] Based on the trend commodity words, a target commodity is selected from a preset commodity pool.

[0010] The present application provides a product selection device, the device comprising:

[0011] The first building module is used to build a commodity database based on commodity data in the online platform;

[0012] A first screening module, based on the commodity database, screens the relevant data of the commodity data to obtain key commodity data;

[0013] A first processing module determines a trending commodity word based on the key commodity data;

[0014] The first determination module selects a target commodity from a preset commodity pool based on the trend commodity word.

[0015] Accordingly, an embodiment of the present application further provides an electronic device, the electronic device comprising: a processor, a memory, and a communication bus;

[0016] The communication bus is used to realize the communication connection between the processor and the memory;

[0017] The processor is used to execute the program in the memory to implement the above-mentioned product selection method.

[0018] An embodiment of the present application also provides a computer-readable storage medium, which stores one or more programs. The one or more programs can be executed by one or more processors to implement the above-mentioned product selection method.

[0019] The embodiment of the present application provides a product selection method, device, equipment and storage medium. First, a product database is constructed based on product data in an online platform; second, based on the product database, relevant data of the product data is screened to obtain key product data; third, based on the key product data, trend product words are determined; finally, based on the trend product words, target products are selected from a preset product pool. In this way, the accuracy of activity themes and product selection can be improved, and platform resource traffic can be saved. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A schematic diagram of a product selection method according to an embodiment of the present application;

[0021] Figure 2 A schematic diagram of a product selection method according to an embodiment of the present application;

[0022] Figure 3 A schematic diagram of a product selection method according to an embodiment of the present application;

[0023] Figure 4 A structural schematic diagram of a product selection method provided in an embodiment of the present application;

[0024] Figure 5 A structural schematic diagram of a product selection method provided in an embodiment of the present application;

[0025] Figure 6 A schematic diagram of the structure of a product selection device provided in an embodiment of the present application;

[0026] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0028] It should be understood that the "embodiments of the present application" or "the aforementioned embodiments" mentioned throughout the specification mean that specific features, structures or characteristics related to the embodiments are included in at least one embodiment of the present application. Therefore, "in the embodiments of the present application" or "in the aforementioned embodiments" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. In the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The above-mentioned serial numbers of the embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments.

[0029] In order to enable a more detailed understanding of the features and technical contents of the embodiments of the present application, the implementation of the embodiments of the present application is described in detail below in conjunction with the accompanying drawings. The attached drawings are for reference only and are not used to limit the present application.

[0030] In e-commerce scenarios, marketing activities play a vital role. Usually, different marketing activities are set up at different times in order to enhance the vitality of e-commerce and increase product sales. Marketing activities not only need to be used to display the event page, but more importantly, the event scene theme and the selection of products to be sold in the event (i.e., product selection). In related technologies, the method of generating event scene themes is mainly for operators to find marketing trends and fixed holiday theme selling points based on manual experience; product selection is mainly for operators to obtain products under the label as a product pool by selecting brands, categories or industrial attribute product labels that are similar to the scene theme.

[0031] The main problems with the manual scenario-based product selection method are:

[0032] 1) The business capabilities and e-commerce experience of operators are the main factors that determine the selling points and effectiveness of marketing activities, so they need to have strong business capabilities and rich e-commerce experience.

[0033] 2) The use of artificial marketing sensitivity to discover marketing trends and hot spots will have the following problems: insufficient hot spot trend power, inappropriate hot spot trend timing (too early or too late); and fixed holiday selling points will still have problems; since users pay attention to different holidays, it is difficult to capture the right opportunity manually. These problems will waste precious activity resources and traffic and affect the overall revenue of the platform.

[0034] 3) Manual product selection based on label rules requires operators to understand the connotation of the scene theme, the products included in the product label, and the relationship between the two. This is a challenge to the job responsibilities of the operator. In addition, manual selection of labels based on scene themes generally results in incorrect selection or omission, which ultimately affects user experience and sales.

[0035] Based on the above problems, the present application embodiment provides a product selection method, such as Figure 1 As shown, the method comprises the following steps:

[0036] Step S101: construct a commodity database based on the commodity data of the online platform.

[0037] In some embodiments, the online platform may be an application (Application, App), mini-program, website, etc., that can be used to purchase goods on the entire network and / or obtain user search data. The commodity data may be data such as the category name, brand name, product name, etc. corresponding to the goods on the online platform where the goods can be purchased; further, these data may represent all the goods. The commodity database stores data such as the category name, brand name, product name, etc. corresponding to the goods.

[0038] Step S102: Based on the commodity database, relevant data of the commodity data is screened to obtain key commodity data.

[0039] In some embodiments, the relevant data of the commodity data may be the hot spot data (on-site and off-site) of the online platform and the user behavior data in the online platform (on-site). Since the relevant data includes non-e-commerce related data and non-related data, it is necessary to filter the relevant data. Since the commodity database includes all commodities; therefore, the category name, brand name, product name, etc. stored in the commodity database are used as keywords to filter the relevant data to retain the commodity data in the relevant data and obtain key commodity data.

[0040] In some embodiments, key product data may include: product data in hot data, and / or product data showing an upward trend in user behavior data; for example: hot data include: hot data A, hot data B, hot data C; wherein hot data A and hot data B are e-commerce related data, and hot data C is non-e-commerce related data; user behavior data include: user behavior data of product A, user behavior data of product B, and user behavior data of product C; wherein, user behavior data of products A and product B are on an upward trend, and user behavior data of product C are on a downward trend; then, hot data A, hot data B, user behavior data of product A, and user behavior data of product B are used as key product data.

[0041] Step S103: determining trend commodity words based on the key commodity data.

[0042] In some embodiments, trending commodity words are commodity words with high user popularity selected from key commodity data. That is to say, the popularity value of each commodity word is determined by analyzing the key commodity data, and the commodity words with the highest popularity value are used as trending commodity words. Trending commodity words can be understood as trending subject words, that is, e-commerce activity subject words; for example: analyzing the above key commodity data, the heat value of hot data A is a, the heat value of hot data B is b, the heat value of commodity A is c, and the heat value of commodity B is d; wherein, a>b>c>d; the commodity words corresponding to the user behavior data of hot data A, hot data B, and commodity A ranked in the top three in terms of heat value are used as trending commodity words. In the embodiments of the present application, trending commodity words can be understood as trending subject words, that is, e-commerce activity subject words.

[0043] Step S104: Select a target product from a preset product pool based on the trend product word.

[0044] In some embodiments, the preset commodity pool is a commodity pool constructed based on commodity data in the online platform (in the site), and the commodity pool may include all commodities in the site. In the embodiment of the present application, selecting a target commodity is selecting a product. The target commodity is a commodity selected from the preset commodity pool that matches the trend commodity word (trend theme word / e-commerce activity theme word), and the commodity is used to participate in the e-commerce activities of the corresponding theme.

[0045] In some embodiments, the preset product pool can store product words; further, the similarity between the trend product words and the product words in the preset product pool can be compared, and the product words in the preset product pool with greater similarity to the trend product words can be used as target products.

[0046] The product selection method provided in the embodiment of the present application firstly builds a product database based on the product data of the online platform; secondly, based on the product database, screens the relevant data of the product data to obtain key product data; thirdly, determines the trend product words based on the key product data; finally, based on the trend product words, selects the target product from the preset product pool. In this way, there is no need to rely on manual experience, improves the accuracy of the activity subject and product selection, and saves platform resource flow.

[0047] In some embodiments, filtering the relevant data of the commodity data mainly involves filtering the hot data in the relevant data; that is, Figure 2 As shown, step S102 includes the following steps:

[0048] Step S201: Based on the commodity database, the hot data is screened to obtain hot commodity data.

[0049] In some embodiments, since the hot data of the online platform includes non-e-commerce related data and non-related data, it is necessary to filter the hot data to retain the e-commerce related data and filter out the non-e-commerce related data and non-related data. In addition, the user behavior data is e-commerce related data; therefore, it is not necessary to filter the user behavior data.

[0050] In some embodiments, category names, brand names, product names, etc. in the commodity database are used as keywords to filter hot data (within and outside the online platform) to retain commodity data in the hot data.

[0051] Step S202: Determine the sales change of the user's associated products based on the user behavior data.

[0052] In some embodiments, user behavior data may include: user purchase data, user browsing data and user order data. The above step S202 may be implemented by the following process: collect statistics of user behavior data within a period of time; analyze the user behavior data within the period of time to determine the change of the statistics within the period of time; and then determine the sales change of the user's associated products according to the change of the statistics; wherein the statistics may be: user purchase volume, user browsing volume and / or user order volume; for example: the user purchase volume, user browsing volume, and / or user order volume of product I are monotonically increasing within the period of time, then it can be determined that the sales change of product I is monotonically increasing; the user purchase volume, user browsing volume, and user order volume of product J are all monotonically decreasing within the period of time, then it can be determined that the sales change of product J is monotonically decreasing.

[0053] In some embodiments, since the user behavior data is data generated by the behavior of all users on the site within a period of time, it is necessary to analyze the user behavior data to determine the sales trend of the associated products associated with the user.

[0054] Step S203: When the sales change increases over time, determine the commodity data corresponding to the associated commodity.

[0055] In some embodiments, after the sales change of the user's associated products is determined in the above steps, the product data corresponding to the associated products whose sales change is monotonically increasing during this period is determined; for example: the sales change of the above-mentioned product I is monotonically increasing, then the product data of product I is determined.

[0056] Step S204: Use the hot commodity data and the commodity data corresponding to the related commodities as the key commodity data.

[0057] In some embodiments, since hot data itself represents the trend over a period of time, the filtered hot data (i.e., hot commodity data) can be directly used as key commodity data. User behavior data is data generated by the behavior of all users in the site over a period of time. Therefore, it is necessary to analyze the user behavior data to determine the commodity data corresponding to the associated commodities with monotonically increasing sales changes over this period of time, and use the commodity data corresponding to the associated commodities as key commodity data.

[0058] The product selection method provided in the embodiment of the present application firstly screens the hot data based on the product database to obtain the hot product data; secondly, determines the sales change of the user's related products based on the user behavior data; thirdly, determines the product data corresponding to the related products when the sales change increases over time; finally, uses the hot product data and the product data corresponding to the related products as key product data. In this way, the key product data with an upward trend can be determined from the hot data and the user behavior data.

[0059] In some embodiments, it is necessary to determine trending commodity words from key commodity data; that is, Figure 3 As shown, step S103 includes the following steps:

[0060] Step S301: Determine the first product word corresponding to the hot product data.

[0061] In some embodiments, the hot commodity data may include: hot search commodity words and corresponding search volumes.

[0062] In some embodiments, a hot search product word in the hot product data is determined as the first product word.

[0063] Step S302: Determine the second product word of the product data corresponding to the associated product.

[0064] In some embodiments, the commodity data corresponding to the associated commodity may include: the category name, brand name and / or product name corresponding to the associated commodity.

[0065] In some embodiments, the category name, brand name and / or product name in the product data corresponding to the associated product is determined as the second product word; for example: the product data corresponding to the associated product includes: 1 brand, k categories, K products, then 1 brand, k categories, and / or K products are determined as the second product words.

[0066] Step S303: determine a trending commodity word based on the first commodity word and the second commodity word.

[0067] In some embodiments, the commodity word with a higher popularity value between the first commodity word and the second commodity word is determined as a trend commodity word; that is, step 133 can be implemented by the following steps:

[0068] Step S331: Aggregate the first product word and the second product word to obtain an aggregated product word set.

[0069] In some embodiments, in order to reduce the situation where there are semantically similar and / or repeated product words in the final trend product words, the first product word and the second product word are similarly aggregated, and the semantically similar and / or repeated product words in the first product word and the second product word are aggregated into one product word to obtain an aggregated product word set; for example: the first product word includes: skirt, one-piece dress, sun protection clothing; the second product word includes: long skirt, pleated skirt, sunglasses; then, skirt, one-piece dress, long skirt, and pleated skirt are aggregated into skirt, and the obtained aggregated word set is: skirt, sun protection clothing, sunglasses.

[0070] Step S332: Determine the popularity value of each aggregated product word in the aggregated product word set.

[0071] In some embodiments, the popularity value of each aggregated commodity word in the aggregated commodity word set is determined based on statistics and / or search volumes of different channels; that is, step S332 can be implemented by the following steps:

[0072] First, multiple source channels of each aggregated product word are determined.

[0073] In some embodiments, the multiple source channels may be online platforms (on-site and off-site platforms).

[0074] Next, determine the popularity of each source channel.

[0075] In some embodiments, if the source channel is within the site, the statistics and search volume of the aggregated product word are used as the popularity of the aggregated product word; if the source channel is outside the site, the search volume of each platform outside the site is used as the popularity of the platform.

[0076] Finally, the heat of each channel of each aggregated commodity word is merged to obtain the heat value of each aggregated commodity word.

[0077] In some embodiments, the heat of each source channel of each aggregated commodity word is normalized, and then combined according to a certain weight to obtain the heat value of each aggregated commodity word, as shown in the following formula:

[0078] T=w1t1+w2t2+…+w n t n

[0079] Where T represents the heat value; w n represents the channel weight; t n Indicates the popularity of the channel.

[0080] Step S333: Determine the aggregated commodity words whose heat values ​​are greater than the first preset threshold as the trend commodity words.

[0081] In some embodiments, the aggregated commodity words are sorted according to their popularity values, and the aggregated commodity words with popularity values ​​greater than a first preset threshold are determined as trending commodity words.

[0082] The product selection method provided in the embodiment of the present application first determines the first product word corresponding to the hot product data; secondly, determines the second product word of the product data corresponding to the associated product; thirdly, aggregates the first product word and the second product word to obtain an aggregated product word set; then, determines the heat value of each aggregated product word in the aggregated product word set; finally, determines the aggregated product word with a heat value greater than a first preset threshold as the trend product word. In this way, the trend product word popular with users can be determined from the key product data with an upward trend.

[0083] In some embodiments, if the above trend commodity words are directly used to select products, it is difficult to select suitable commodities due to too little information. Therefore, it is necessary to perform intent analysis on the trend commodity words (trend scenarios) to find more words or commodity tags to express the trend commodity words; that is, step S104 includes the following steps:

[0084] Step S141: semantically expand the trend product words to obtain a first expanded word set.

[0085] In some embodiments, the intention analysis of trend scenarios is realized by combining two semantic similarity expressions, text vector and text word importance; for example, semantic analysis of "dress" is performed by using text vector to obtain semantically similar words such as long skirt, A-line skirt, pleated skirt, dress, etc.; semantic analysis of "dress" is performed by using the importance of important words to obtain semantically similar words such as versatile dress, retro dress, and Hepburn style dress; that is, step S141 includes the following steps:

[0086] Step S1411: Determine the text vector and importance of the trend product word.

[0087] In some embodiments, a text vector of a trending commodity word may be generated by inputting the trending commodity word into a word vector model.

[0088] In some embodiments, the importance of a word may be determined by calculating a term frequency-inverse document frequency (TF-IDF) value of the word.

[0089] Step S1412: Expand the trend product words based on the text vector and / or the importance to obtain the first expanded word set.

[0090] In some embodiments, based on the text vector, the trend product word is expanded to obtain the first expanded word set; that is, the intention of the trend scenario is parsed using the text vector; this can be achieved through the following process:

[0091] First, the commodity words in the preset commodity pool are input into a word vector model to obtain a word vector database.

[0092] In some embodiments, the word vector model can be obtained by training the word vector Word2vec model using the product name after word segmentation (in the commodity database). The word vector model is used to generate word vectors. Of course, the word vector GloVe model, the word vector ELMo model, and the word vector BERT model can also be trained to generate word vectors; no further details are given here.

[0093] In some embodiments, since the preset commodity pool is a commodity pool constructed based on commodity data in the online platform (in the site), the preset commodity pool may include all commodities in the site. Then, the commodity words in the preset commodity pool are input into the word vector model, and the obtained word vector database includes the word vectors of all commodities in the site.

[0094] Secondly, the trend product word is input into the word vector model to obtain the trend product word vector.

[0095] In some embodiments, there may be one or more trend product words, and accordingly, there may be one or more trend product word vectors, depending on the actual situation.

[0096] Again, determine the word vectors in the word vector database whose similarity with the trend product word vector is greater than a second preset threshold, and obtain a first extended word vector set.

[0097] In some embodiments, the similarity between each word vector in the word vector database and the trend product word vector is calculated, and the word vector in the word vector data whose similarity is greater than a second preset threshold is used as a word vector (first extended word vector) with similarity to the trend product word, thereby obtaining a first extended word vector set.

[0098] In some embodiments, the similarity may be calculated using algorithms such as cosine similarity, Euclidean distance, Manhattan distance, Pearson correlation coefficient, Spearman (rank) correlation coefficient, and Jaccard distance.

[0099] Finally, the first extended word vector set is converted to obtain the first extended word set.

[0100] In some embodiments, the above-mentioned similarity calculation algorithm is used to determine the words closest to each word vector in the first extended word vector set, and these words are used as the first extended word set.

[0101] In some embodiments, based on the importance, the trend product words are expanded to obtain the first expanded word set; that is, the intention of the trend scenario is parsed using the importance of text words; this can be achieved through the following process:

[0102] First, the similarity between the commodity words in the commodity database and the trend commodity words is determined.

[0103] In some embodiments, the similarity between each product word in the product database and the trend product word is calculated, and the product words in the product database whose similarity is greater than a third preset threshold are determined.

[0104] In some embodiments, a product database may be used to build a text word-based search service through a distributed search engine, Elastic search, and then the trend product words may be searched in the search service to obtain search results.

[0105] Secondly, the importance of each product word among the product words whose similarity is greater than a third preset threshold is determined.

[0106] In some embodiments, the importance of each product word among the product words whose similarity is greater than a third preset threshold value can be determined by the TF-IDF value of the product word (search result); wherein, the larger the TF-IDF value, the more important the word.

[0107] Finally, the commodity words whose importance is greater than a fourth preset threshold are used as the first extended word set.

[0108] In some embodiments, the commodity words are sorted according to the TF-IDF values ​​calculated above, and the commodity words with TF-IDF values ​​greater than a fourth preset threshold are used as the first expanded word set.

[0109] Step S142: Select a target product that matches the first set of expanded words from the preset product pool.

[0110] In some embodiments, if the expansion words in the first expansion word set are category names and / or brand names, an exact match is used to directly select a product that matches the category name and / or brand name in the preset product pool as the target product; if the expansion words in the first expansion word set are product names, a fuzzy search is used to match the product name and business details in the preset product pool using the product name, and then select the target product. Step S142 includes the following steps:

[0111] Step S1421: Determine the similarity between each extended word in the first extended word set and the commodity words in the preset commodity pool.

[0112] In some embodiments, the similarity between each expansion word in the first expansion word set and the commodity words in the preset commodity pool is calculated.

[0113] In some embodiments, if the expanded words in the first expanded word set are category names and / or brand names, exact matching is used; if the expanded words in the first expanded word set are product names, fuzzy retrieval is used.

[0114] Step S1422: Determine the target product words whose similarity is greater than a fifth preset threshold.

[0115] In some embodiments, exact matching is used to directly select product words that match the category name and / or brand name in the preset product pool as target products; fuzzy retrieval is used to use the product name to match the product name and business details data in the preset product pool, and then select the target product word.

[0116] In some embodiments, a commodity word in the preset commodity pool whose similarity with the extended word (in the first extended set) is greater than a fifth preset threshold is determined as a target commodity word.

[0117] Step S1423: Based on the target product word, determine the target product in the preset product pool.

[0118] In some embodiments, in order to enrich the final target products, step S1423 can be implemented by the following process:

[0119] First, semantic expansion is performed on the target product word to obtain a second expanded word set.

[0120] In some embodiments, the process of obtaining the second expanded word set refers to the process of obtaining the first expanded word set described above.

[0121] Secondly, the commodities in the preset commodity pool that match the second extended word set are used as the target commodities.

[0122] In some embodiments, the second extended word set is matched in the preset commodity pool, and commodities matching the second extended word set are used as target commodities. In the embodiment of the present application, the preset commodity pool stores commodities and corresponding commodity word vectors and commodity words.

[0123] The product selection method provided in the embodiment of the present application first determines the text vector and importance of the trend product word; secondly, based on the text vector and / or importance, the trend product word is expanded to obtain a first expansion word set; thirdly, the similarity between each expansion word in the first expansion word set and the product words in the preset product pool is determined; then, a target product word whose similarity is greater than a fifth preset threshold is determined; finally, based on the target product word, a target product is determined in the preset product pool. In this way, the trend product word and the first expansion word set can be expanded, so that the target product (selected product) finally selected is richer.

[0124] Based on this, Figure 4 FIG. 1 is a schematic diagram of a product selection method provided in an embodiment of the present application; Figure 4 The modules shown are described below:

[0125] The basic product word module 401 is used to determine the basic product word.

[0126] Here, the basic commodity words include at least: brand name, category name and product name, etc.

[0127] The word search service module 402 is used to provide word search services.

[0128] Here, an Elasticsearch search service (word search service) is built based on the words in the basic commodity words (such as brand name, category name, product name). The search results can be sorted by the TF-IDF value corresponding to the word, and then the words with similar semantics can be obtained; this can be achieved through the following process:

[0129] Step a: Build an Elastic search retrieval service based on basic product terms;

[0130] Step b: Based on the trend keywords, use the Elastic search service to search for basic product keywords and obtain search results;

[0131] Step c, calculating the TF-IDF value of the search results;

[0132] Step d: Sort the search results according to the TF-IDF value, and use the search results with a TF-IDF value greater than a preset threshold as product-related words (corresponding to the first extended word set mentioned above)

[0133] Here, if Figure 5As shown, combined Figure 5 The module shown illustrates the process of obtaining trend keywords:

[0134] The first step is to conduct off-site data mining.

[0135] Here, the data obtained through off-site data mining include: A industry ranking 51, A industry hot search data 52, B industry hot search data 53, C industry hot search hot list 54, shopping platform hot list 55 and D industry trend research 56, etc.; among them, A industry ranking 51 and A industry hot search data 52 are obtained through full crawling; B industry hot search data 53, C industry hot search hot list 54, shopping platform hot list 55 are obtained through full crawling and keyword selection; D industry trend research 56 is obtained through keyword crawling.

[0136] The second step is to perform keyword screening on the data mined outside the site.

[0137] While executing the first and second steps, execute the third and fourth steps to conduct on-site mining and implement keyword screening.

[0138] Here, since A industry ranking 51, A industry hot search data 52, B industry hot search data 53, C industry hot search hot list 54, shopping platform hot list 55 and A industry trend research 56 are recent trend data open to the outside of the site and represent recent trends, the words in the basic data words are used as keywords to filter the data mined outside the site, and the obtained trend keywords can be directly used as trend words, and the corresponding search volume can be used as trend heat.

[0139] The third step is to conduct data mining within the site.

[0140] Here, the data obtained by performing on-site data mining includes: search data 57, user behavior data 58, hot search list 59, etc.; among them, search data 57 and user behavior data 58 are obtained by keyword crawling; hot search list 59 is obtained by full crawling and keyword screening.

[0141] Here, the user behavior data 58 may be user order data, user browsing data, and user purchase data. The search data 57 and the hot search list 59 may be hot search text words and corresponding search volumes.

[0142] The fourth step is to filter the search data and hot search lists by keywords.

[0143] Since search data 57 and hot search list 59 are recent trend data open to the public within the site and represent recent trends, the words in the basic data words are used as keywords to filter search data 57 and hot search list 59. The obtained trend keywords can be directly used as trend words, and the corresponding search volume can be used as trend heat.

[0144] Here, the trend keywords obtained in the second and fourth steps correspond to the first product word mentioned above.

[0145] The fifth step is to determine the trend based on the user behavior data mined from the site and determine the corresponding trend keywords.

[0146] Here, the output of the user behavior data 58 is used as the input of the trend determination module 501 to perform trend word determination.

[0147] Here, the sales trend of the product is determined by analyzing the user behavior data; the product words corresponding to the products whose sales trend is monotonically increasing within a preset period of time are used as trend keywords (corresponding to the second product words mentioned above); the fifth step can be implemented by the following steps:

[0148] First, analyze user behavior data to obtain corresponding trends;

[0149] Table 1 User behavior data statistics

[0150]

[0151] Here, the specific implementation process is: according to the change of the statistical quantity of the user behavior data within the preset time, the corresponding trend of the user behavior data is obtained; take the preset time of three days and the statistical quantity as the user order quantity in the user order data as an example for explanation: the user behavior data within three days is shown in Table 1, where the user order quantity x 11 <User order quantity x 12 <User order quantity x 13 ; User order quantity x 21 >User order quantity x 22 >User order quantity x 23 ; User order quantity x 31 <User order quantity x 32 >User order quantity x 33 ; User order quantity x 41 >User order quantity x 42 <User order quantity x 43 Then, the trend of commodity A within the preset time period is monotonically increasing, the trend of commodity B within the preset time period is monotonically decreasing, the trend of commodity C within the preset time period is first increasing and then decreasing, and the trend of commodity D within the preset time period is first decreasing and then increasing. Of course, the statistical quantity can be the user order quantity in the user order data, the user browsing quantity in the user browsing data, and / or the user added purchase quantity in the user added purchase data, which will not be repeated here.

[0152] Secondly, based on the trends, get trending keywords.

[0153] In some embodiments, the product words corresponding to the products with a monotonically increasing trend within a preset time period are used as trend keywords, and the statistics of the products (such as user order volume, user browsing volume or user added purchase volume) are used as their corresponding trend heat.

[0154] Step 6: Based on the trend keywords obtained in steps 2, 4 and 5, trend word fusion is performed;

[0155] like Figure 5 As shown, the A industry ranking 51, A industry hot search data 52, B industry hot search data 53, C industry hot search hot list 54, shopping platform hot list 55, A industry trend research 56, search data 57, hot search list 59 and the output of the trend determination module 501 are used as the input of the trend word fusion module 502 to perform trend word fusion.

[0156] Step 7: Sort the integrated trend words.

[0157] Here, the output of the trend word fusion module 502 is used as the input of the trend word sorting module 503 to sort the trend words.

[0158] Here, the trend power (corresponding to the above heat value) of the fused trend words is calculated, and then the trend words are sorted according to the trend power. The calculation of the trend power is the same as the calculation of the heat value, which will not be described in detail here.

[0159] Step 8: Determine trend keywords based on the output of the trend keyword ranking module.

[0160] Here, the trend theme word is determined based on the output of the trend word sorting module 503. The trend word corresponding to the trend force greater than the preset threshold is used as the trend theme word.

[0161] The word vector retrieval service module 403 is used to provide word vector retrieval services.

[0162] Here, we segment the product names in the product pool, extract the word vectors of the segmented product names, and build the Vearch word vector retrieval service based on the word vectors. This can be achieved through the following process:

[0163] Step A: Segment all product names on the site;

[0164] Step B: Use the segmented product names as training sets;

[0165] Step C, using the training set to train the word vector Word2vec model to obtain a trained word vector Word2vec model;

[0166] Step D: Input the product names in the commodity pool into the word vector model to obtain a word vector database;

[0167] Step E: Input the trend topic word into the word vector model to obtain the trend topic word vector;

[0168] Step D: Building Vearch search service based on word vector database;

[0169] Step E: Based on the trend theme word vector, use Vearch search service to search the word vector database to obtain product related words (corresponding to the first extended word set mentioned above)

[0170] The search result sorting module 404 is used to sort the word search results and the word vector search results to determine the product related words.

[0171] Here, various similarities may be superimposed with certain weights to obtain new similarities for sorting, or they may be used separately according to the application scenario.

[0172] The product search_1 module 405 is used to match product-related words in the product pool to determine a co-occurring product group.

[0173] Here, an exact match or fuzzy search is performed in the product pool according to the type of product-related words. Exact matching means: when the product-related words are category names or brand names, the brand name and / or category name are directly used to select products in the product pool to obtain a co-occurring product group. Fuzzy search means: when the product-related words are product names, the product name and / or its corresponding business details are used to select products in the product pool to obtain a co-occurring product group.

[0174] The product search_2 module 406 is used to perform word vector search based on the product word vectors of the co-occurring product group to obtain selected products. This can be achieved by the following process:

[0175] First, the co-occurring product group is input into the word vector model to obtain the co-occurring product word vector group.

[0176] Secondly, based on the co-occurring product word vector group, the Vearch search service is used to search the word vector database to obtain the target product.

[0177] The product selection method provided in the embodiment of the present application firstly uses the basic product word module to determine the basic product word; secondly, uses the word search service module to provide word search services for trend keywords; thirdly, uses the word vector search service module to provide word vector search services for trend keywords; then, uses the search result sorting module to sort the word search results and the word vector search results to determine the product-related words; then, uses the product search_1 module to match the product-related words in the product pool to determine the co-occurring product group; finally, uses the product search_2 module to perform word vector search based on the product word vectors of the co-occurring product group to obtain the selected products. In this way, the following technical problems are solved: 1) Automatic mining of trend keywords is realized, which solves the problem of insufficient control of trends by operators; 2) The scene trend theme analysis is used to accurately and comprehensively express the scene trend theme, which solves the problem of insufficient expression of the scene theme; 3) The combination of traditional search logic and product vector search is used to solve the problem of insufficient selection level; through the above three points, the effect of marketing activities is improved, and the problem of waste of platform resources and traffic is solved.

[0178] Based on the above embodiments, the present application also provides a product selection device 600, which can be applied to Figures 1 to 5 In a product selection method provided in a corresponding embodiment, referring to Figure 6 As shown, the product selection device 600 includes: a first construction module 601, a first screening module 602, a first processing module 603 and a first determination module 604, wherein:

[0179] The first construction module 601 is used to construct a commodity database based on commodity data in the online platform;

[0180] A first screening module 602, based on the commodity database, screens the relevant data of the commodity data to obtain key commodity data;

[0181] The first processing module 603 determines a trending commodity word based on the key commodity data;

[0182] The first determination module 604 selects a target product from a preset product pool based on the trend product word.

[0183] In other embodiments of the present application, the first screening module 602 also includes: a first screening submodule, used to screen the hot data based on the product database to obtain hot product data; a first determination submodule, used to determine the sales change of the user's associated products based on the user behavior data; a second determination submodule, used to determine the product data corresponding to the associated products when the sales change increases over time; and a third determination submodule, used to use the hot product data and the product data corresponding to the associated products as the key product data.

[0184] In other embodiments of the present application, the first processing module 603 also includes: a fourth determination submodule, used to determine the first product word corresponding to the hot product data; a fifth determination submodule, used to determine the second product word of the product data corresponding to the associated product; and a sixth determination submodule, used to determine the trend product word based on the first product word and the second product word.

[0185] In other embodiments of the present application, the sixth determination submodule also includes: a first aggregation submodule, used to aggregate the first product word and the second product word to obtain an aggregated product word set; a seventh determination submodule, used to determine the popularity value of each aggregated product word in the aggregated product word set; an eighth determination submodule, used to determine the aggregated product word whose popularity value is greater than a first preset threshold as the trend product word.

[0186] In other embodiments of the present application, the seventh determination submodule is also used to determine multiple source channels of each aggregated commodity word; determine the popularity of each source channel; and merge the popularity of each channel of each aggregated commodity word to obtain the popularity value of each aggregated commodity word.

[0187] In other embodiments of the present application, the first determination module 604 further includes: a first expansion submodule, used to semantically expand the trend product word to obtain a first expansion word set; a first selection submodule, used to select a target product that matches the first expansion word set in the preset product pool.

[0188] In other embodiments of the present application, the first expansion submodule also includes: a ninth determination submodule, used to determine the text vector and importance of the trend commodity word; and a tenth determination submodule, based on the text vector and / or the importance, to expand the trend commodity word to obtain the first expansion word set.

[0189] In other embodiments of the present application, the tenth determination submodule is further used to input the product words in the preset product pool into the word vector model to obtain a word vector database; input the trend product words into the word vector model to obtain a trend product word vector; determine the word vectors in the word vector database whose similarity with the trend product word vector is greater than a second preset threshold to obtain a first extended word vector set; convert the first extended word vector set to obtain the first extended word set; and / or determine the similarity between the product words in the product database and the trend product words; determine the importance of each product word among the product words whose similarity is greater than a third preset threshold; and use the product words whose importance is greater than a fourth preset threshold as the first extended word set.

[0190] In other embodiments of the present application, the first selection submodule also includes: an eleventh determination submodule, used to determine the similarity between each extended word in the first extended word set and the product words in the preset product pool; a twelfth determination submodule, used to determine the target product word whose similarity is greater than the fifth preset threshold; and a thirteenth determination submodule, used to determine the target product in the preset product pool based on the target product word.

[0191] In other embodiments of the present application, the thirteenth determination submodule is further used to perform semantic expansion on the target product word to obtain a second expansion word set; and use the product in the preset product pool that matches the second expansion word set as the target product.

[0192] Based on the above embodiments, the present application also provides an electronic device 700, which can be applied to Figures 1 to 5 In a product selection method provided in a corresponding embodiment, referring to Figure 7 As shown, the electronic device 700 includes: a processor 701, a memory 702 and a communication bus 703, wherein:

[0193] The communication bus 703 is used to implement the communication connection between the processor 71 and the memory 72 .

[0194] The processor 701 is used to execute the program of the product selection method stored in the memory 702 to achieve reference Figures 1 to 5 The corresponding embodiment provides a product selection method.

[0195] Based on the foregoing embodiments, the embodiments of the present application further provide a computer-readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the following. Figures 1 to 5 The corresponding embodiment provides a product selection method.

[0196] It should be understood that the "one embodiment" or "one embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in one embodiment" appearing throughout the specification may not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in various embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application. The above-mentioned sequence number of the embodiment of the present application is only for description and does not represent the advantages and disadvantages of the embodiment. It should be noted that in this article, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or device including the element.

[0197] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0198] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0199] In addition, all functional units in the embodiments of the present application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional units. A person of ordinary skill in the art can understand that all or part of the steps of the above method embodiments can be completed by hardware related to program instructions, and the above program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above method embodiments are executed; and the above storage medium includes: mobile storage devices, read-only memory (ROM), disks or optical disks, etc. Various media that can store program codes.

[0200] Alternatively, if the above-mentioned integrated unit of the present application is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiment of the present application can be embodied in the form of a software product in essence or in other words, the part that contributes to the prior art can be stored in the form of a software product, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks, or optical disks. The above is only a specific implementation method of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or replacements within the technical scope disclosed in the present application, and they should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A product selection method, characterized in that: The method comprises: Build a product database based on the product data of the online platform; Based on the commodity database, relevant data of the commodity data is screened to obtain key commodity data; Determine trending commodity words based on the key commodity data; Based on the trending commodity words, select a target commodity from a preset commodity pool; The step of selecting a target product from a preset product pool based on the trend product word includes: Performing semantic expansion on the trend product words to obtain a first expanded word set; In the preset commodity pool, a target commodity matching the first extended word set is selected.

2. The method according to claim 1, characterized in that The relevant data includes: hot spot data of the online platform and user behavior data in the online platform; the relevant data of the commodity data is screened based on the commodity database to obtain key commodity data, including: Based on the commodity database, the hot data is screened to obtain hot commodity data; Determine the sales change of the user's associated products based on the user behavior data; In a case where the sales change amount increases over time, determining commodity data corresponding to the associated commodity; The hot commodity data and the commodity data corresponding to the related commodities are used as the key commodity data.

3. The method according to claim 2, characterized in that The determining of trend commodity words based on the key commodity data includes: Determine a first product word corresponding to the hot product data; Determine a second product word of the product data corresponding to the associated product; Based on the first commodity word and the second commodity word, a trend commodity word is determined.

4. The method according to claim 3, characterized in that The determining of a trend commodity word based on the first commodity word and the second commodity word includes: Aggregating the first product word and the second product word to obtain an aggregated product word set; Determine the popularity value of each aggregated product word in the aggregated product word set; The aggregated commodity words whose heat values ​​are greater than the first preset threshold are determined as the trend commodity words.

5. The method according to claim 4, characterized in that The determining the popularity value of each aggregated commodity word in the aggregated commodity word set includes: Determining multiple source channels for each aggregated product word; Determine the popularity of each source channel; The heat of each channel of each aggregated commodity word is integrated to obtain the heat value of each aggregated commodity word.

6. The method according to claim 1, characterized in that The semantic expansion of the trend product words to obtain a first expanded word set includes: Determining the text vector and importance of the trending commodity word; Based on the text vector and / or the importance, the trend product words are expanded to obtain the first expanded word set.

7. The method according to claim 6, characterized in that The step of expanding the trend product words based on the text vector to obtain the first expanded word set includes: Inputting the commodity words in the preset commodity pool into a word vector model to obtain a word vector database; Inputting the trend product word into the word vector model to obtain the trend product word vector; Determine a word vector in the word vector database whose similarity to the trend product word vector is greater than a second preset threshold, and obtain a first extended word vector set; The first extended word vector set is converted to obtain the first extended word set.

8. The method according to claim 6, characterized in that The step of expanding the trend product words based on the importance to obtain the first expanded word set includes: Determining the similarity between the product words in the product database and the trend product words; Determining the importance of each product word among the product words whose similarity is greater than a third preset threshold; The commodity words whose importance is greater than a fourth preset threshold are used as the first expanded word set.

9. The method according to claim 1, characterized in that: The step of selecting a target commodity matching the first set of extended words from the preset commodity pool includes: Determining the similarity between each extended word in the first extended word set and the commodity words in the preset commodity pool; Determine the target product word whose similarity is greater than a fifth preset threshold; Based on the target commodity word, the target commodity is determined in the preset commodity pool.

10. The method according to claim 9, characterized in that The determining the target product in the preset product pool based on the target product word includes: Performing semantic expansion on the target product word to obtain a second expanded word set; The commodities in the preset commodity pool that match the second extended word set are used as the target commodities.

11. A product selection device, characterized in that: The device comprises: The first building module is used to build a commodity database based on commodity data in the online platform; A first screening module, configured to screen relevant data of the commodity data based on the commodity database to obtain key commodity data; A first processing module, configured to determine a trend commodity word based on the key commodity data; A first determination module, configured to select a target commodity from a preset commodity pool based on the trend commodity word; The first determination module is specifically configured to perform semantic expansion on the trend commodity words to obtain a first expanded word set; and select a target commodity matching the first expanded word set from the preset commodity pool.

12. An electronic device, characterized in that: The electronic device comprises: a processor, a memory and a communication bus; The communication bus is used to realize the communication connection between the processor and the memory; The processor is used to execute the program in the memory to implement the product selection method as described in any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that: The computer readable storage medium stores One or more programs are stored, and the one or more programs can be executed by one or more processors, To realize the product selection method as described in any one of claims 1 to 10.

Images