Interest label mining method and device, equipment and program product

Abstract

The invention relates to the technical field of e-commerce, in particular to an interest tag mining method and device, equipment and a program product. The method comprises the steps that unpopular commodities containing preset unpopular labels in a commodity sequence are obtained, the commodity sequence comprises commodities corresponding to at least one operation behavior in the history of a user, and all the commodities in the commodity sequence are sequentially arranged according to the operation moments of the operation behaviors; after the arrangement positions of the unpopular commodities in the commodity sequence are adjusted to the relatively recent commodities, a new commodity sequence is obtained, and the operation time corresponding to the relatively recent commodities is later than the operation time corresponding to the unpopular commodities; based on the new commodity sequence, an interest label is obtained, and the interest label is used for recommending commodities for the user. According to the method, the interest label better conforming to the interest tendency of the user can be obtained, so that the interest label mining accuracy is improved.

Description

Technical Field

[0001] The present application relates to the field of e-commerce technology, and specifically to a method, device, equipment and program product for mining interest tags. Background Art

[0002] For e-commerce systems, understanding user interests can make targeted product recommendations for users, increase the probability of user purchases, and have very high business value. However, the product categories of e-commerce systems are complex, and there are large differences in product and user behavior between categories. Especially on e-commerce platforms for second-hand goods, the unique and low-frequency characteristics of second-hand goods pose many challenges to the accurate mining of user interests. At present, user interest mining methods can be based on a well-defined user interest tag system, mainly based on statistical methods, assisted by necessary machine learning mining algorithms, to mine users' potential interest preferences. However, when setting the hyperparameters of the machine learning mining algorithm, the method based on statistical historical user behavior needs to formulate statistical formulas and set hyperparameters to avoid interference from noise behavior. The amount of user behavior for different types of tags is different. Especially when facing products with low-frequency characteristics, the existing method based on historical user behavior will lead to inaccurate interest tag mining results. Summary of the invention

[0003] Based on the above-mentioned defects and shortcomings of the prior art, the present application proposes a method, device, equipment and program product for mining interest tags, which can obtain interest tags that are more in line with user interests, thereby improving the accuracy of interest tag mining.

[0004] According to a first aspect of an embodiment of the present application, a method for mining interest tags is provided, comprising: obtaining a product sequence obtained by a user in at least one historical operation behavior, and sorting multiple products in the product sequence according to the time when the historical operation behavior occurred; obtaining unpopular products in the product sequence that contain preset unpopular tags; adjusting the order of the unpopular products in the product sequence to a preset position, and obtaining a new product sequence; and obtaining the user's interest tags based on the new product sequence.

[0005] According to the method for mining interest tags provided in the first aspect of the embodiment of the present application, adjusting the order of the unpopular products in the product sequence to a preset position includes: segmenting the product sequence to obtain one or more product sequence segments; for the products in each product sequence segment, adjusting the order of the unpopular products in the product sequence segment to a preset position.

[0006] According to the method for mining interest tags provided in the first aspect of the embodiment of the present application, the obtaining of the user's interest tags based on the new product sequence includes: adding sequence prompt information to each product in the product sequence fragment, and obtaining a new product sequence fragment with the sequence prompt information added, wherein the later the arrangement position in the new product sequence fragment is, the greater the weight of the added sequence prompt information; and inputting one or more of the new product sequence fragments into a preset large language model to obtain the user's interest tags output by the large language model.

[0007] According to the method for mining interest tags provided in the first aspect of an embodiment of the present application, after obtaining a sequence of commodities obtained by a user in at least one historical operation behavior, the method further includes: determining invalid commodities in the commodity sequence based on the correlation between commodities in the commodity sequence and the user browsing time corresponding to each commodity; and removing the invalid commodities from the commodity sequence.

[0008] According to the method for mining interest tags provided in the first aspect of the embodiment of the present application, the correlation between the commodities in the commodity sequence is calculated in the following manner: for any of the commodity sequence fragments, the comprehensive similarity between the selected commodity and other commodities is calculated.

[0009] According to the method for mining interest tags provided in the first aspect of the embodiment of the present application, the invalid commodities in the commodity sequence are determined based on the correlation between commodities in the commodity sequence and the user browsing time corresponding to each commodity, including: for each commodity sequence segment, if the comprehensive similarity between the selected commodity and the other commodities is the smallest in the commodity sequence segment, and the user browsing time of the selected commodity is less than a preset time threshold, then the selected commodity is the invalid commodity in the commodity sequence segment.

[0010] According to the method for mining interest tags provided in the first aspect of the embodiment of the present application, the calculation of the comprehensive similarity between the selected product and other products includes: obtaining the product vector corresponding to each product in the product sequence; based on the product vector, calculating the similarity between any two products; based on the similarity between the any two products, calculating the comprehensive similarity between the selected product and the other products.

[0011] According to a second aspect of an embodiment of the present application, a device for mining interest tags is provided, comprising: a first acquisition module, used to acquire a product sequence obtained by a user in at least one historical operation behavior, wherein a plurality of products in the product sequence are sorted according to the time when the historical operation behavior occurred; a second acquisition module, used to acquire unpopular products in the product sequence that contain preset unpopular tags; a position adjustment module, used to adjust the order of the unpopular products in the product sequence to a preset position to obtain a new product sequence; and a tag mining module, used to acquire the user's interest tags based on the new product sequence.

[0012] According to the third aspect of an embodiment of the present application, an electronic device is provided, comprising: a memory and a processor; the memory is connected to the processor and is used to store programs; the processor is used to implement the interest tag mining method as described in the first aspect by running the program in the memory.

[0013] According to a fourth aspect of an embodiment of the present application, a computer program product is provided, comprising computer program instructions; when the computer program instructions are executed by a processor, the processor executes the method for mining interest tags as described in any one of the first aspects.

[0014] In an embodiment of the present application, a product sequence containing unpopular products with preset unpopular tags is obtained, wherein the product sequence includes products corresponding to at least one operation behavior in the user's history, and each product in the product sequence is arranged in order according to the operation time when the operation behavior occurred; the arrangement position of the unpopular products in the product sequence is adjusted to after the relatively recent products, and a new product sequence is obtained, wherein the operation time corresponding to the relatively recent products is later than the operation time corresponding to the unpopular products; based on the new product sequence, interest tags are obtained, wherein the interest tags are used to recommend products to users. In the above process, by adjusting the arrangement position of the unpopular products in the product sequence, the attention paid to the unpopular tagged products in the process of obtaining interest tags is increased, and the unpopular tags are avoided from being ignored in the process of mining interest tags, so that the obtained interest tags are more in line with the user's interest tendencies, thereby improving the mining accuracy when mining interest tags. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0016] Figure 1 One of the flow charts of a method for mining interest tags provided in an embodiment of the present application;

[0017] Figure 2 A second flow chart of a method for mining interest tags provided in an embodiment of the present application;

[0018] Figure 3 A block diagram of an interest tag mining device provided in an embodiment of the present application;

[0019] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] Exemplary Methods

[0022] In order to improve the accuracy of product recommendations, this application provides a method for mining interest tags. The method for mining interest tags can be implemented by a software algorithm, and the software algorithm implementing this method can be run on any device with data processing capabilities, such as a computer, a cloud server, a smart phone, etc. The scope of protection of this application is not limited to the type of device that implements this method.

[0023] In one embodiment, if Figure 1 As shown, the process steps for implementing the mining method of interest tags are as follows:

[0024] Step 101: Obtain a commodity sequence obtained by a user in at least one historical operation behavior, and sort multiple commodities in the commodity sequence according to the time when the historical operation behavior occurs.

[0025] In this embodiment, when a user browses products using an e-commerce system, different operations will be performed on the products, and each operation corresponds to a product, for example, clicking on product a, browsing product b, etc. Historical operation behavior refers to the operation behavior performed by the user in history. According to the time when multiple historical operation behaviors occurred, the products corresponding to each historical operation behavior are sorted to obtain a product sequence. Each product includes at least one label, such as color, model, brand, function, etc. For example, the following is the generation process of a product sequence:

[0026] First, the user clicks on product a, and the corresponding tags of product a include: model is smartphone A, color is red;

[0027] Next, the user clicks on product b, and the labels corresponding to product b include: model is smartphone A, color is pink;

[0028] Then, the user clicks on product c, and the labels corresponding to product c include: model is watch, color is red;

[0029] Then, the user clicks on product d, and the labels corresponding to product d include: model is smartphone B, color is red;

[0030] Finally, the user clicks on product e, and the labels corresponding to product e include: model is smartphone C, color is red;

[0031] Based on the above user operation behavior, obtain the product sequence {a, b, c, d, e}.

[0032] Step 102, obtaining unpopular commodities containing preset unpopular labels in the commodity sequence.

[0033] In this embodiment, the preset unpopular label refers to a label that is pre-set according to actual conditions and needs. For example, the preset unpopular label may be watch, green, etc. A product containing a preset unpopular label is an unpopular product. For example, if product C in the above product sequence {a, b, c, d, e} includes the preset unpopular label watch, then product C is an unpopular product. Furthermore, the preset unpopular label may be pre-set manually or obtained by analyzing a large amount of historical data of different users.

[0034] Step 103, after adjusting the order of the unpopular products in the product sequence to a preset position, a new product sequence is obtained.

[0035] In this embodiment, according to the actual e-commerce environment, the products that users click more recently are more likely to be the products that users actually want to buy. Therefore, in the process of mining interest tags, the closer the products corresponding to the historical operation behaviors are to the current moment, the greater the impact on the mining of interest tags, and the mined interest tags will be closer to the products needed at the current moment. Based on the above situation, for unpopular products, users may have only clicked on the products of this category once or several times in history, and because unpopular tags cannot be pushed to users too much, even if the user is more interested in unpopular products, when mining interest tags based on the product sequence, accurate interest tags cannot be obtained based on the unpopular tags of unpopular products in the future. Therefore, the arrangement position of unpopular products in the product sequence can be adjusted. According to actual business conditions and needs, a preset position can be pre-set, which can increase the influence of unpopular products when mining interest tags based on new product sequences. Optionally, the arrangement position of unpopular products in the product sequence is adjusted to after relatively recent products, wherein the operation time corresponding to relatively recent products is later than the operation time corresponding to unpopular products. Specifically, relatively recent products make products that are arranged further back in the product sequence than relatively unpopular products. For example, in the product sequence {a, b, c, d, e}, if c is a rare product, then d and e are relatively recent products of rare product c. When adjusting the ranking position of rare product c, rare product c can be adjusted to after product d or product e, and the new product sequence can be {a, b, d, c, e} or {a, b, d, e, c}. The closer the position of the rare product is to the end of the product sequence, the greater the influence of the adjusted new product sequence on the interest tag mining process.

[0036] Step 104: Obtain user interest tags based on the new product sequence.

[0037] In this embodiment, obtaining interest tags based on the new product sequence can be achieved by a pre-set mining model. Specifically, the new product sequence is input into the preset mining model to obtain the interest tags output by the mining model. The mining model can be a pre-trained deep learning model, such as a large language model. At the same time, obtaining interest tags based on the new product sequence can also be obtained by other methods such as vector calculation.

[0038] In this embodiment, when mining interest tags based on a product sequence, the order of each product in the product sequence will respond to the mining process of interest tags. The arrangement position of unpopular products in the product sequence is adjusted to after relatively recent products to obtain a new product sequence. The process of mining interest tags based on the new product sequence does not ignore the influence of unpopular tags, thereby obtaining more accurate interest tags. The more accurate interest tags can be further used to recommend products to users, making the accuracy of product recommendations higher and more in line with the actual needs of users.

[0039] In one embodiment, the order of unpopular products in a product sequence is adjusted to a preset position as follows: the product sequence is segmented to obtain one or more product sequence segments; for each product in the product sequence segment, the order of the unpopular products in the product sequence segment is adjusted to a preset position in the product sequence segment.

[0040] In this embodiment, when adjusting the order of unpopular products in the product sequence, the adjustment can be made based on the entire product sequence, and the order of unpopular products can be adjusted to a preset position in the entire product sequence. Furthermore, in order to avoid excessive focus on unpopular tags in interest tag mining, the product sequence can be divided into one or more product sequence segments, and the position of unpopular products can be adjusted in each product sequence segment to limit the range of adjustment of the position of unpopular products, so as to avoid excessive adjustment of the position of unpopular products in the entire product sequence, which will lead to excessive focus on unpopular tags in interest tag mining.

[0041] In this embodiment, the product sequence is segmented to obtain one or more product sequence segments, and there are no duplicate products between any two product sequence segments. When the product sequence is segmented in advance, it can be segmented according to time period, number of products or semantics according to actual conditions and needs, and the number of products contained in each product sequence segment can be the same or different.

[0042] In this embodiment, the order of the unpopular commodities in the commodity sequence segment is adjusted to a preset position in the commodity sequence segment. At this time, the preset position can be set based on the actual situation and needs of the segmentation of the commodity sequence segment. Optionally, the arrangement position of the unpopular commodities in the commodity sequence segment can be adjusted to after the commodity corresponding to the most recent operation behavior. Furthermore, when a commodity sequence segment includes two or more unpopular commodities, when adjusting the arrangement position of the unpopular commodities, the original relative position relationship between each unpopular commodity is maintained. For example, in the commodity sequence segment {a, b, c, d}, if a and c are unpopular commodities, then when adjusting the arrangement position of the unpopular commodities a and c, the new segment sequence can be {b, d, a, c}.

[0043] In this embodiment, in order to prevent the interest tag acquisition process from paying too much attention to unpopular tags, the product sequence is divided into one or more product sequence segments, and then the arrangement position of the unpopular tags in each small product sequence segment is adjusted, that is, the influence of unpopular products on the interest tag mining process is adjusted within a reasonable range, thereby further improving the accuracy of interest tags.

[0044] In one embodiment, based on a new product sequence, user interest tags are obtained as follows: sequence prompt information is added to each product in a product sequence segment to obtain a new product sequence segment with added sequence prompt information, wherein the later the arrangement position in the new product sequence segment is, the greater the weight of the added sequence prompt information; one or more new product sequence segments are input into a preset large language model to obtain the user's interest tags output by the large language model.

[0045] In this embodiment, since the amount of user behavior corresponding to unpopular products is relatively small, in order to further enhance the influence of unpopular products, sequence prompt information can be added to each product in each product sequence segment to obtain a new product sequence segment with added sequence prompt information. Optionally, adding sequence prompt information to each product in each product sequence segment can be implemented based on prompt technology. Through the prompt technology, sequence prompt information with greater weight is added to products with a later arrangement position in the product sequence segment, and the large language model (LLM) pays more attention to products with greater sequence prompt information weight when mining interest tags. Since the arrangement position of unpopular products in the product sequence segment has been adjusted, unpopular products can add sequence prompt information with greater weight. Therefore, even unpopular products with less user behavior can still get more attention from the large language model, so that the large language model outputs more accurate interest tags.

[0046] In this embodiment, the large language model can be a pre-set model. Due to the complex internal structure of the large language model, it is difficult to adjust the internal parameters. Therefore, the new product sequence fragments obtained after adjusting the order of the unpopular products and adding sequence prompt information are input into the large language model, so that the large prediction model can pay enough attention to the unpopular products and output more accurate interest tags.

[0047] In one embodiment, after obtaining a product sequence obtained by a user in at least one historical operation behavior, the method further includes: determining invalid products in the product sequence based on the correlation between products in the product sequence and the user browsing time corresponding to each product; and removing the invalid products from the product sequence.

[0048] In this embodiment, a series of operations continuously generated by users in the past may include invalid behaviors caused by misoperations, and the products corresponding to the invalid behaviors are defined as invalid products. Therefore, if any product in the product sequence has a small correlation with other products in the product sequence, and it is determined based on the user's browsing time that the user has not browsed the product for a long time, then the product is determined to be an invalid product, and the invalid product is removed from the product sequence, thereby avoiding the interference of invalid products on the interest tag mining process. At the same time, removing invalid products from the product sequence can also reduce the amount of data in the product sequence and improve the efficiency of the interest tag mining process.

[0049] In this embodiment, when determining and eliminating invalid commodities, if the commodity sequence has been segmented to obtain one or more commodity sequence segments, the invalid commodities can be determined and eliminated based on the commodity sequence segments. For each commodity sequence segment, the invalid commodities in the commodity sequence segment are determined based on the correlation between commodities in the commodity sequence segment and the user browsing time corresponding to each commodity; the invalid commodities are eliminated from the commodity sequence segment, and further sequence prompt information can be directly added to each commodity sequence segment after the invalid commodities are eliminated, to obtain a new commodity sequence segment with added sequence prompt information.

[0050] In one embodiment, the correlation between commodities in a commodity sequence is calculated in the following manner: for any commodity sequence segment, the comprehensive similarity between the selected commodity and other commodities is calculated.

[0051] In this embodiment, the correlation between commodities in the commodity sequence can be characterized by the comprehensive similarity between commodities. Each commodity in the commodity sequence is respectively regarded as a selected commodity. When any commodity is selected as a selected commodity, the other unselected commodities in the commodity sequence are other commodities. The comprehensive similarity refers to the comprehensive similarity between the selected commodity and each other commodity. The comprehensive similarity can be calculated based on any method such as information vectorization and information matrixization.

[0052] In one embodiment, invalid commodities in a commodity sequence are determined based on the correlation between commodities in the commodity sequence and the user browsing time corresponding to each commodity, including: for each commodity sequence segment, if the comprehensive similarity between the selected commodity and other commodities is the smallest in the commodity sequence segment, and the user browsing time of the selected commodity is less than a preset time threshold, then the selected commodity is an invalid commodity in the commodity sequence segment.

[0053] In this embodiment, in any product sequence segment, if the comprehensive similarity corresponding to any selected product is the smallest in the product sequence segment, it indicates that the selected product may be a product that the user has mistakenly operated. If at the same time, the user's browsing time of the selected product is less than the preset time threshold, it indicates that the user browsed the selected product for a very short time, and the selected product is determined to be an invalid product. Further, the preset time threshold can be a time threshold set manually in advance, or a time threshold calculated in advance based on historical data. For example, the preset time threshold can be the average browsing time of products in the same category.

[0054] In one embodiment, calculating the comprehensive similarity between the selected product and other products includes: obtaining a product vector corresponding to each product in the product sequence; calculating the similarity between any two products based on the product vector; and calculating the comprehensive similarity between the selected product and other products based on the similarity between any two products.

[0055] In this embodiment, the comprehensive similarity between the selected product and other products is calculated based on the product sequence segments obtained in advance. Specifically, for each product sequence segment, the product vectors corresponding to each product in the product sequence segment are obtained; based on the product vectors, the similarity between any two products in the product sequence segment is calculated; based on the similarity between any two products in the product sequence segment, the comprehensive similarity between the selected product and other products in the product sequence segment is calculated. By traversing each product sequence segment, the corresponding invalid products can be eliminated for each product sequence segment.

[0056] In this embodiment, the product vector corresponding to each product can be obtained through a pre-trained vector calculation model. The labels of any product are input into the vector calculation model, and the vector calculation model outputs the product vector of the product. The product vectors corresponding to each product calculated by the vector calculation model are all equal-length vectors, which facilitates subsequent calculations. Of course, when obtaining the product vector corresponding to each product, other semantic analysis methods can also be used to parse and integrate the product labels to obtain the corresponding product vectors.

[0057] In this embodiment, the similarity between any two commodities is calculated based on the commodity vectors, and any similarity measurement value such as cosine distance, Euclidean distance or Mahalanobis distance between any two commodity vectors can be calculated. When calculating the comprehensive similarity between the selected commodity and other commodities, the sum of similarities, average similarity or similarity mean square error between each selected commodity and other commodities in the commodity sequence can be calculated based on the similarity between any two commodities.

[0058] In this embodiment, the calculation of the comprehensive similarity between the selected product and other products can also be completed based on the product sequence. Specifically, for the entire product sequence, the product vectors corresponding to each product in the product sequence are obtained; based on the product vectors, the similarity between any two products in the product sequence is calculated; based on the similarity between any two products in the product sequence, the comprehensive similarity between the selected product and other products in the product sequence is calculated. Based on the comprehensive similarity calculated based on the entire product sequence, the corresponding invalid products are eliminated for each product sequence segment.

[0059] In a specific embodiment, Figure 2 As shown in FIG, the detailed process of implementing the mining method of interest tags is as follows:

[0060] Step 201, obtaining a product sequence;

[0061] Step 202, based on a preset time period window, divide the product sequence into at least one product sequence segment;

[0062] Step 203, based on the product vector of each product, calculate the comprehensive similarity between the selected product and other products;

[0063] Step 204: if the comprehensive similarity corresponding to the selected product is the smallest in the product sequence segments, and the user's browsing time for the selected product is less than a preset time threshold, the selected product is determined to be an invalid product;

[0064] Step 205, removing invalid commodities respectively determined in each commodity sequence segment;

[0065] Step 206, for each product sequence segment, adjusting the arrangement position of the unpopular product in the product sequence segment to a preset position;

[0066] Step 207, based on the prompt technology, adding sequence prompt information to the commodities in each commodity sequence segment to obtain a new commodity sequence segment;

[0067] Step 208, input each new product sequence segment into a preset large language model to obtain interest tags output by the large language model;

[0068] Step 209, obtaining recommended products based on the interest tags.

[0069] In this embodiment, the user's operation behavior within a time window is coherent, and the interests reflected are the same or similar. Therefore, it is more effective to filter invalid products by comprehensive similarity. At the same time, in order to avoid large deviations in eliminating invalid products based only on comprehensive similarity, the user's browsing time is considered at the same time. If the browsing time of users who select the product at the same time is less than the preset time threshold (for example, the average browsing time of users in the category), the invalid products will be determined and eliminated. Eliminating invalid products from the product sequence ensures that the products in the product sequence are all corresponding to the user's effective operation behavior, reducing the noise interference caused by invalid products, which is conducive to improving the accuracy of interest tag mining. For unpopular products, by adjusting the arrangement order of unpopular products and cooperating with prompt technology, the effect of focusing on the large language model is achieved. At the same time, in order to avoid the large language model from paying too much attention to unpopular products, only the position of unpopular products in the product sequence fragment is adjusted, rather than the position of the entire product sequence. And because the content richness of product details pages of different categories is inconsistent, the user browsing time of each product is not added to the product sequence used by the large language model to mine interest tags, so as not to introduce more noise. The large language model can extract multiple labels at one time, such as model and color. Compared with the traditional method that requires a corresponding statistical task for each label, the large language model can greatly improve development efficiency.

[0070] In an embodiment of the present application, a product sequence containing unpopular products with preset unpopular tags is obtained, wherein the product sequence includes products corresponding to at least one operation behavior in the user's history, and each product in the product sequence is arranged in order according to the operation time when the operation behavior occurred; the arrangement position of the unpopular products in the product sequence is adjusted to after the relatively recent products, and a new product sequence is obtained, wherein the operation time corresponding to the relatively recent products is later than the operation time corresponding to the unpopular products; based on the new product sequence, interest tags are obtained, wherein the interest tags are used to recommend products to users. In the above process, by adjusting the arrangement position of the unpopular products in the product sequence, the attention paid to the unpopular tagged products in the process of obtaining interest tags is increased, and the unpopular tags are avoided from being ignored in the process of mining interest tags, so that the obtained interest tags are more in line with the user's interest tendencies, thereby improving the mining accuracy when mining interest tags.

[0071] Exemplary Devices

[0072] Accordingly, the present application embodiment also provides a device for mining interest tags, such as Figure 3 As shown, the device may include:

[0073] The first acquisition module 301 is used to acquire a commodity sequence obtained by the user in at least one historical operation behavior, and multiple commodities in the commodity sequence are sorted according to the occurrence time of the historical operation behavior;

[0074] The second acquisition module 302 is used to acquire unpopular commodities containing preset unpopular labels in the commodity sequence;

[0075] A position adjustment module 303 is used to adjust the order of the unpopular products in the product sequence to a preset position to obtain a new product sequence;

[0076] The tag mining module 304 is used to obtain user interest tags based on the new product sequence.

[0077] The device for mining interest tags provided in this embodiment belongs to the same application concept as the method for mining interest tags provided in the above embodiments of this application, and can execute the method for mining interest tags provided in any of the above embodiments of this application, and has the corresponding functional modules and beneficial effects of the execution method. For technical details not fully described in this embodiment, please refer to the specific processing content of the method for mining interest tags provided in the above embodiments of this application, and will not be repeated here.

[0078] Exemplary Electronic Devices

[0079] The present application also provides an electronic device, such as Figure 4 As shown, the electronic device includes: a memory 400 and a processor 401 .

[0080] The memory 400 is connected to the processor 401 and is used to store programs.

[0081] The processor 401 is used to implement the method for mining interest tags in the above embodiment by running the program stored in the memory 400 .

[0082] Specifically, the electronic device may further include: a communication interface 402 , an input device 403 , an output device 404 and a bus 405 .

[0083] The processor 401, the memory 400, the communication interface 402, the input device 403 and the output device 404 are connected to each other via a bus.

[0084] Bus 405 may include a pathway for transferring information between various components of the computer system.

[0085] Processor 401 may be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the scheme of the present invention. It may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0086] The processor 401 may include a main processor, and may also include a baseband chip, a modem, etc.

[0087] The memory 400 stores a program for executing the technical solution of the present invention, and may also store an operating system and other key services. Specifically, the program may include a program code, and the program code includes a computer operation instruction. More specifically, the memory 400 may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk storage, a flash, and the like.

[0088] The input device 403 may include a device for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor.

[0089] Output device 404 may include a device that allows information to be output to a user, such as a display screen, a printer, a speaker, etc.

[0090] The communication interface 402 may include any transceiver or the like to communicate with other devices or communication networks, such as Ethernet, a radio access network (RAN), a wireless local area network (WLAN), etc.

[0091] The processor 401 executes the program stored in the memory 400 and calls other devices, which can be used to implement the various steps of the interest tag mining method provided in the above embodiments of the present application.

[0092] Exemplary computer program products and storage media

[0093] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps in the interest tag mining method described in the embodiment of the present application.

[0094] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0095] In addition, an embodiment of the present application may also be a storage medium on which a computer program is stored, and the computer program is executed by a processor to execute the steps of the interest tag mining method described in the embodiment of the present application.

[0096] For the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0097] It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0098] The steps in the methods of each embodiment of the present application can be adjusted in order, combined and deleted according to actual needs, and the technical features recorded in each embodiment can be replaced or combined.

[0099] The modules and sub-modules in the devices and terminals provided in the embodiments of the present application can be merged, divided and deleted according to actual needs.

[0100] In the several embodiments provided in the present application, it should be understood that the disclosed terminals, devices and methods can be implemented in other ways. For example, the terminal embodiments described above are only schematic, for example, the division of modules or submodules is only a logical function division, and there may be other division methods in actual implementation, for example, multiple submodules or modules can be combined or integrated into another module, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0101] The modules or submodules described as separate components may or may not be physically separated, and the components of the modules or submodules may or may not be physical modules or submodules, that is, they may be located in one place, or they may be distributed on multiple network modules or submodules. Some or all of the modules or submodules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0102] In addition, each functional module or submodule in each embodiment of the present application may be integrated into one processing module, or each module or submodule may exist physically separately, or two or more modules or submodules may be integrated into one module. The above-mentioned integrated modules or submodules may be implemented in the form of hardware or in the form of software functional modules or submodules.

[0103] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0104] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly by hardware, software units executed by a processor, or a combination of the two. The software units may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0105] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0106] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for mining interest tags, characterized in that: include: Acquire a commodity sequence obtained by the user in at least one historical operation behavior, wherein multiple commodities in the commodity sequence are sorted according to the time when the historical operation behavior occurred; Acquire unpopular products in the product sequence that contain preset unpopular labels; After adjusting the order of the unpopular products in the product sequence to a preset position, a new product sequence is obtained; Based on the new product sequence, the user's interest tags are obtained.

2. The method for mining interest tags according to claim 1, characterized in that: The step of adjusting the order of the unpopular product in the product sequence to a preset position includes: Segmenting the product sequence to obtain one or more product sequence segments; For each commodity in a commodity sequence segment, the order of the unpopular commodities in the commodity sequence segment is adjusted to a preset position.

3. The method for mining interest tags according to claim 2, characterized in that: The acquiring the user's interest tags based on the new product sequence includes: For each commodity in the commodity sequence segment, sequence prompt information is added respectively to obtain a new commodity sequence segment with the sequence prompt information added thereto, wherein the later the arrangement position in the new commodity sequence segment is, the greater the weight of the added sequence prompt information is; One or more of the new product sequence segments are input into a preset large language model, and the user's interest tags output by the large language model are obtained.

4. The method for mining interest tags according to claim 2, characterized in that: After obtaining the commodity sequence obtained by the user in at least one historical operation behavior, the method further includes: Determine invalid commodities in the commodity sequence according to the correlation between commodities in the commodity sequence and the user browsing time corresponding to each commodity; The invalid commodities are removed from the commodity sequence.

5. The method for mining interest tags according to claim 4, characterized in that: The correlation between the commodities in the commodity sequence is calculated as follows: For any of the product sequence segments, the comprehensive similarity between the selected product and other products is calculated.

6. The method for mining interest tags according to claim 5, characterized in that: The determining, based on the correlation between commodities in the commodity sequence and the user browsing time corresponding to each commodity, invalid commodities in the commodity sequence includes: For each of the product sequence segments, if the comprehensive similarity between the selected product and the other products is the smallest in the product sequence segment, and the user browsing time of the selected product is less than a preset time threshold, then the selected product is the invalid product in the product sequence segment.

7. The method for mining interest tags according to claim 5, characterized in that: The calculating of the comprehensive similarity between the selected product and other products includes: Obtain the product vector corresponding to each product in the product sequence; Based on the product vectors, calculating the similarity between any two products; Based on the similarity between the any two commodities, a comprehensive similarity between the selected commodity and the other commodities is calculated.

8. A device for mining interest tags, characterized in that: include: A first acquisition module is used to acquire a commodity sequence obtained by a user in at least one historical operation behavior, wherein multiple commodities in the commodity sequence are sorted according to the time when the historical operation behavior occurs; A second acquisition module is used to acquire unpopular commodities in the commodity sequence that contain preset unpopular labels; A position adjustment module, used to adjust the order of the unpopular products in the product sequence to a preset position to obtain a new product sequence; The tag mining module is used to obtain the user's interest tags based on the new product sequence.

9. An electronic device, characterized in that: include: Memory and processor; The memory is connected to the processor and is used to store programs; The processor is used to implement the method for mining interest tags as described in any one of claims 1-7 by running the program in the memory.

10. A computer program product, characterized in that includes computer program instructions; When the computer program instructions are executed by a processor, the processor executes the method for mining interest tags as described in any one of claims 1 to 7.

Images