Personalized search information recommendation method based on user behavior analysis

Abstract

The invention relates to the technical field of information, in particular to a personalized search information recommendation method based on user behavior analysis, which mainly comprises the following steps: acquiring interactive behavior data of a user in real time; the method comprises the following steps: automatically extracting keyword tags of various contents through a natural language processing technology, and constructing a content keyword tag system; according to the content keyword tag system, extracting keyword tags associated with user behaviors from the interactive behavior data; in combination with the accumulated value of each type of behavior and the preset weight value, calculating the matching degree between the keyword tag and the user, and generating a matching degree set; performing variance operation on the matching degree set corresponding to the keyword tag to obtain sensitivity corresponding to the user content; and generating recommendation information of the search content of the user according to the sensitivity. According to the method, high-precision and real-time response personalized search content recommendation can be realized by analyzing user behaviors.

Description

Technical Field

[0001] This application relates to the field of information technology, and in particular, to a personalized search information recommendation method based on user behavior analysis. Background Art

[0002] With the explosive growth of the amount of information on the Internet, the scenario where users rely on search engines to obtain information has become increasingly common. However, traditional search engines usually only rely on literal keyword matching to return results. When the information input by users is incomplete or ambiguous, such as using abbreviations or vague descriptions, it is difficult to accurately capture the true intentions of users, resulting in low relevance of search results and a decline in user experience.

[0003] Although algorithms based on content and collaborative filtering in existing recommendation technologies can partially alleviate this problem, there are still significant limitations: firstly, they overly rely on historical static data and ignore the dynamic changes and scenario differences of user behavior, such as click behavior in search and non-search scenarios; secondly, they lack in-depth correlation analysis of content semantics, and the automatic generation of keyword tag systems is prone to semantic deviation and difficult to cover the diverse needs of users.

[0004] Therefore, how to achieve high-precision and real-time responsive personalized recommendation through dynamic behavior analysis and a refined tag system has become a technical problem to be solved urgently. Summary of the Invention

[0005] This application provides a personalized search information recommendation method based on user behavior analysis, which can realize high-precision and real-time responsive personalized search content recommendation through the analysis of user behavior.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a personalized search information recommendation method based on user behavior analysis, including:

[0008] S1. Real-time obtain the interactive behavior data of users, including search behavior, click behavior, and favorite behavior, and configure a preset weight value for each type of behavior;

[0009] S2. Automatically extract keyword tags of various contents through natural language processing technology to construct a content keyword tag system;

[0010] S3. Extract the keyword tags associated with user behavior from the interactive behavior data according to the content keyword tag system;

[0011] S4. Calculate the cumulative values of the search behavior, click behavior, and favorite behavior of users based on the interactive behavior data;

[0012] S5. Combine the cumulative value of each type of behavior and the preset weight value, calculate the matching degree between the keyword tag and the user, and generate a set of matching degrees;

[0013] S6. Perform variance calculation on the set of matching degrees corresponding to the keyword tag to obtain the sensitivity corresponding to the user content;

[0014] S7. Generate recommendation information for the user search content according to the sensitivity.

[0015] In a preferred example of the present application, it can be further set to further include:

[0016] Clean, de-duplicate and normalize the interaction behavior data.

[0017] In a preferred example of the present application, it can be further set that the preset weight value includes the favorite behavior weight W c , click behavior weight W s , search behavior weight W q , and the content keyword tag system includes at least one dimension of content title word splitting, service classification, functional attribute, and audience group tag.

[0018] In a preferred example of the present application, it can be further set that in step S2., the automatic extraction of keyword tags includes:

[0019] Perform word segmentation and part-of-speech tagging on multiple content texts;

[0020] Extract core keywords based on the TF-IDF algorithm;

[0021] Expand associated tags through the pre-trained semantic model BERT.

[0022] In a preferred example of the present application, it can be further set that the formula for calculating the matching degree between the keyword tag and the user is:

[0023]

[0024] Where is the matching degree, C j is the cumulative value of the number of times the user favorites the tag t j , S j is the cumulative value of the number of clicks, Q j is the cumulative value of the number of times the search statement contains the tag t j , w c , w s , w q are preset values.

[0025] In a preferred example of the present application, it can be further set to generate a matching degree set, perform variance calculation on the matching degree set corresponding to the keyword tags, and obtain the sensitivity corresponding to the user content, including:

[0026] Integrate all keywords in content c n to obtain the matching degree set of the keyword tags corresponding to content c n :

[0027]

[0028] Perform variance calculation on the matching degree set of the keyword tags to obtain the sensitivity corresponding to the single memory C n The formula is: where f(U, T) is a function defining the relationship between user behavior and keyword tags,

[0029]

[0030] represents the behavioral sensitivity of user U to the keyword tag set of content c n and k is the number.

[0031]

[0031] In a second aspect, the present application provides a personalized search information recommendation device based on user behavior analysis, and the device includes:

[0032] A data collection module, configured to obtain user interaction behavior data in real time, including search behavior, click behavior, and favorite behavior, and configure a preset weight value for each type of behavior;

[0033] A keyword module, configured to automatically extract keyword tags of various contents through natural language processing technology, construct a content keyword tag system; and extract keyword tags associated with user behavior from the interaction behavior data according to the content keyword tag system;

[0034] A sensitivity module, configured to calculate the cumulative value of the user's search behavior, click behavior, and favorite behavior based on the interaction behavior data; combine the cumulative value of each type of behavior and the preset weight value, calculate the matching degree between the keyword tag and the user, and generate a matching degree set; perform variance calculation on the matching degree set corresponding to the keyword tag to obtain the sensitivity corresponding to the user content;

[0035] A recommendation module, configured to generate recommended information for the user's search content according to the sensitivity.

[0036] In a third aspect, the present application provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the personalized search information recommendation method based on user behavior analysis as described in any one of the above are implemented.

[0037] In a fourth aspect, the present application provides a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, the personalized search information recommendation method based on user behavior analysis as described in any one of the above is implemented.

[0038] In a fifth aspect, the present application provides a computer program product, including computer instructions, which implement the steps of the personalized search information recommendation method based on user behavior analysis as described in any one of the above when executed by a processor.

[0039] In summary, compared with the prior art, the beneficial effects brought by the technical solutions provided in the embodiments of the present application at least include:

[0040] The method of the present application constructs a comprehensive content keyword tag system, and real-time tracks and analyzes various interaction behaviors of users, including operations such as collection, search, and click access, and then further expands more refined interaction behaviors, thereby constructing a dynamic and comprehensive user behavior database. Then, the algorithm is used to analyze the data, and the content strongly related to the current needs of the user is displayed to the user, realizing a more accurate and personalized search recommendation experience, and at the same time ensuring the timeliness and relevance of the recommended content. Moreover, the generation and update of keyword tags are realized through automated NLP technology, replacing the manual tag configuration process. After testing and verification, the tag generation efficiency is improved. Description of the Drawings

[0041] Figure 1 It is a flowchart of a personalized search information recommendation method based on user behavior analysis provided by an embodiment of the present application.

[0042] Figure 2 It is a module diagram of a personalized search information recommendation device based on user behavior analysis provided by an embodiment of the present application. Detailed Embodiments

[0043] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0044] In one embodiment of the present application, a personalized search information recommendation method based on user behavior analysis is provided. Refer to Figure 1 as shown, the method includes:

[0045] S1. Real-time obtain the interaction behavior data of the user, including search behavior, click behavior, and favorite behavior, and configure a preset weight value for each type of behavior;

[0046] S2. Automatically extract keyword tags of various contents through natural language processing technology to construct a content keyword tag system;

[0047] S3. According to the content keyword tag system, extract the keyword tags associated with the user behavior from the interaction behavior data;

[0048] S4. Based on the interaction behavior data, calculate the cumulative values of the user's search behavior, click behavior, and favorite behavior;

[0049] S5. Combine the cumulative value of each type of behavior and the preset weight value to calculate the matching degree between the keyword tag and the user, and generate a matching degree set;

[0050] S6. Perform a variance operation on the matching degree set corresponding to the keyword tag to obtain the sensitivity of the user content;

[0051] S7. Generate recommended information for the user's search content according to the sensitivity.

[0052] Specifically, the search behavior mainly refers to the search behavior of the user from clicking the search to entering the keyword; the click behavior mainly includes the operation behavior of clicking the functional content of the page, which can be distinguished as search click (entering the search result page and clicking the functional content) and non-search click (clicking the functional content outside the search result page); the favorite behavior includes the operation of favoriting the content, which can be the favorite of the content after search or the favorite of the content browsed during non-search.

[0053] The person in charge presets the weight values of each behavior index at the management end, mainly including favorite behavior, (search / non-search) click access behavior, and search behavior. Preset the weight W c of the favorite behavior, the weight W s of the (search / non-search) click access behavior, and the weight W q of the search behavior.

[0054] Extract the keywords of the content using existing natural language processing techniques. The above-mentioned information such as keyword themes, categories, attributes, etc. can be specifically interpreted as: simple word splitting of the content title, the category to which the content belongs (including the department to which it belongs, the more refined service classification to which it belongs, etc.), and the functional attributes of the content (including specific usage functions, target audiences, specific label attributes, etc.). Manually maintaining keywords ensures that the keyword system is richer and more effective, and is more flexible and practical compared to the simple word splitting of some keyword systems.

[0055] Assign a set of keywords to each piece of content to achieve content tagging. Example: The content set C = {c1, c2,..., c n (n≥1)}, for c n (n≥1), the keyword tag set T cn can be expressed as: T cn = {t1, t2,..., t j (j≥1)}, where t j represents a certain characteristic keyword tag of the content.

[0056] After the target user generates an interaction behavior, the system identifies the user's behavior patterns and preferences by comparing the user's behavior with the preset behavior metrics (specifically, the development will perform data embedding at each function of the system interface. The search behavior is to click on the search box, enter content and click search; the click behavior is to click on a specific piece of content, such as an application, an article, etc., and the click-through is the click behavior; the favorite behavior is to click on the favorite button and then it is identified as a favorite to achieve automatic identification). At the same time, the system extracts the associated keyword tags from the content data generated by the user's behavior, including extracting the associated keywords for all the content data that generates the behavior. The keywords are manually matched with the content. One piece of content can have n keywords, and the keyword tags are sorted and recorded and stored in the behavior data of the target user.

[0057] After obtaining the user behavior data and the content keyword tags, the system calculates the matching degree between each keyword tag and the target user according to the preset recommendation strategy. Subsequently, the keyword tags that are highly relevant to the user are screened out, and the system makes personalized recommendations for the content associated with such keyword tags, thereby generating a precise content recommendation list.

[0058] The above-mentioned actions of obtaining user behavior data and content keywords are not just simple data recording, but rather further understand the relationship between user behavior and keyword tags by establishing a set of user behavior and content sensitivity. Example: Establish a sensitivity set of user behavior and content set C where represents the keyword tag set of user U for content c n ​Behavioral sensitivity: Among them, f(U, T) is a function that defines the relationship between user behavior and keyword tags, and the following is a detailed description of this function.

[0059] The recommendation strategy mentioned above is based on the user's behavior content. First, the cumulative value of the behavior indicators is calculated, and then the matching degree is calculated for each keyword tag. (First, the cumulative value data of each behavior indicator is known, and then the keyword matching degree is calculated. The top-ranked keywords are considered to have higher weights, and then the content associated with these higher-weighted keywords is selected for display)

[0060] The calculation method of the cumulative value of the above-mentioned behavioral indicators is as follows: j Represents the user's keyword tag t j The cumulative value of collection behavior; S j Represents the user's keyword tag t j The cumulative value of (search / non-search) click behaviors; Q j The search sentence contains the keyword tag t when the user searches j Cumulative value.

[0061] Specifically, C j The specific calculation method is: Among them, count(U i ,t j , collection) indicates that user U i For label t j The number of collections, n is the total number of collection behaviors of users.

[0062] Specifically, S j The specific calculation method is: Among them, count(U i ,t j , click) indicates that user U i For label t j The number of (search / non-search) clicks, n is the total number of user (search / non-search) click behaviors.

[0063] Specifically, in calculating the above Q j When accumulating values, we introduced search weights for calculation. The "search" behavior referred to here, as mentioned above, refers to the search activity of users clicking on the search box and entering keywords. When our goal is to extract tags from keywords, the search behavior can be further refined into the tags contained in the search statement when the user searches. The "search" behavior below will not be explained in detail. j The specific calculation method is: Among them, count(U i ,tj , search) indicates that user U i When searching, the search sentence contains tag t j The total number, and n is the total number of search behaviors.

[0064] Furthermore, calculate the matching degree of each keyword tag. Denote the keyword tag matching degree as m, and the calculation formula is: where w c , w s , w q are preset values, which are obtained by automatically performing statistical calculations on the user's interaction behaviors and integrating the content keywords generated behind the behaviors, so as to obtain the matching degree set corresponding to a single content c n There is a matching degree set for the corresponding keyword tag

[0065] Furthermore, perform a variance operation on the matching degree set for the keyword tag to obtain the sensitivity of the corresponding single memory C n That is: That is:

[0066] Furthermore, finally obtain the sensitivity set P of the content set C through the above operation process. Sort the values of the elements in the set P in descending order to obtain the recommended information for the search content corresponding to the current user U.

[0067] The person in charge continuously updates the keywords and content data to ensure that the recommended content can dynamically adapt to the changes in user behaviors.

[0068] Adopt a feedback collection method to continuously optimize the keyword matching degree and the recommendation algorithm. When the target user generates each target behavior, repeat steps S3 - S7, and push personalized content to the target user according to the recommendation strategy; when the keyword tag of the content is updated, repeat steps S1 - S2, rematch the keyword tag with the behavior, the system will separately recalculate the recommendation value of the updated keyword tag, and enter step S8 to update the recommended content queue based on the obtained recommendation value, and push personalized content to the target user according to the updated keyword tag recommendation queue.

[0069] In this embodiment, by constructing a comprehensive content keyword tag system, and real-time tracking and analyzing various interaction behaviors of users, including operations such as collection, search, click access, etc., and then expanding to more refined interaction behaviors, a dynamic and comprehensive user behavior database is constructed. Then, the algorithm is used to analyze the data, and the content strongly related to the user's current needs is displayed to the user, realizing a more accurate and personalized search and recommendation experience, and at the same time ensuring the timeliness and relevance of the recommended content. Moreover, the generation and update of keyword tags are realized through automated NLP technology, replacing the manual tag configuration process. After testing and verification, the tag generation efficiency is improved.

[0070] In some embodiments, it further includes:

[0071] Cleaning, de-duplicating, and normalizing the interaction behavior data.

[0072] In some embodiments, the preset weight values include the collection behavior weight W c , click behavior weight W s , search behavior weight W q , and the content keyword tag system includes at least one dimension of content title word splitting, service classification, functional attributes, and audience group tags.

[0073] In some embodiments, in step S2., the automatic extraction of keyword tags includes:

[0074] Performing word segmentation and part-of-speech tagging on various content texts;

[0075] Extracting core keywords based on the TF-IDF algorithm;

[0076] Expanding associated tags through the pre-trained semantic model BERT.

[0077] In some embodiments, the formula for calculating the matching degree between the keyword tag and the user is:

[0078]

[0079] Wherein, is the matching degree, C j is the cumulative value of the number of times the user collects the tag t j , S j is the cumulative value of the number of clicks, Q j is the cumulative value of the number of times the search statement contains the tag t j , w c , w s , w q are preset numerical values.

[0080] In some embodiments, a matching degree set is generated, and variance operation is performed on the matching degree set corresponding to the keyword tags to obtain the sensitivity corresponding to the user content, including:

[0081] Integrate content c n All the keywords in to obtain the matching degree set of the keyword tags corresponding to content c n :

[0082]

[0083] Perform variance operation on the matching degree set of the keyword tags to obtain the sensitivity corresponding to the single memory C n The formula is: where f(U, T) is a function defining the relationship between user behavior and keyword tags,

[0084]

[0085] represents the behavioral sensitivity of user U to the keyword tag set of content c and k is the number. n of This application also provides a personalized search information recommendation device based on user behavior analysis. Please refer to

[0086] shown. The device includes: Figure 2 :

[0087] A data acquisition module 100 for obtaining the interactive behavior data of users in real time, including search behavior, click behavior, and favorite behavior, and configuring a preset weight value for each type of behavior;

[0088] A keyword module 200 for automatically extracting keyword tags of various contents through natural language processing technology to construct a content keyword tag system; extracting keyword tags associated with user behavior from the interactive behavior data according to the content keyword tag system;

[0089] A sensitivity module 300 for calculating the cumulative values of the search behavior, click behavior, and favorite behavior of the user based on the interactive behavior data; combining the cumulative values of each type of behavior and the preset weight value to calculate the matching degree between the keyword tags and the user, generating a matching degree set; performing variance operation on the matching degree set corresponding to the keyword tags to obtain the sensitivity corresponding to the user content;

[0090] A recommendation module 400 for generating recommended information for the user's search content according to the sensitivity.

[0091] The function implementation of each module in the above personalized search information recommendation device based on user behavior analysis corresponds to the steps in the embodiment of the above personalized search information recommendation method based on user behavior analysis. The functions and implementation processes will not be elaborated here one by one.

[0092] This application also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the personalized search information recommendation method based on user behavior analysis as described in any of the above embodiments are implemented.

[0093] This application also provides a computer-readable storage medium. A program is stored on the computer-readable storage medium. Herein, the computer-readable storage medium refers to a carrier for storing data, which may include, but is not limited to, a floppy disk, an optical disc, a hard disk, a flash memory, a USB flash drive, and / or a Memory Stick, etc. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. For the working process, working details, and technical effects of the computer-readable storage medium provided in this embodiment, reference may be made to the embodiment of the personalized search information recommendation method based on user behavior analysis in the foregoing text, which will not be elaborated here.

[0094] This application also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, the steps of the personalized search information recommendation method based on user behavior analysis as described in any of the above embodiments are implemented.

[0095] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM).

[0096] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification. The above-described embodiments merely represent several implementation manners of the present application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A personalized search information recommendation method based on user behavior analysis, characterized in that, Including: S1. Real-time obtain the user's interaction behavior data, including search behavior, click behavior, and favorite behavior, and configure a preset weight value for each type of behavior; S2. Automatically extract keyword tags of various contents through natural language processing technology to construct a content keyword tag system; S3. According to the content keyword tag system, extract the keyword tags associated with the user's behavior from the interaction behavior data; S4. Based on the interaction behavior data, calculate the cumulative values of the user's search behavior, click behavior, and favorite behavior; S5. Combine the cumulative value of each type of behavior and the preset weight value to calculate the matching degree between the keyword tag and the user, and generate a matching degree set; S6. Perform a variance operation on the matching degree set corresponding to the keyword tag to obtain the sensitivity corresponding to the user content; S7. Generate recommended information for the user's search content according to the sensitivity.

2. The personalized search information recommendation method for user behavior analysis according to claim 1, wherein It also includes: Clean, de-duplicate, and normalize the interaction behavior data.

3. The personalized search information recommendation method based on user behavior analysis according to claim 2, characterized in that, The preset weight value includes a collection behavior weight W c , a click behavior weight W s , a search behavior weight W q , and the content keyword tag system includes at least one dimension among word segmentation of the content title, service classification, functional attributes, and audience group tags.

4. The personalized search information recommendation method based on user behavior analysis according to claim 3, wherein In step S2., the automatic extraction of keyword tags includes: Perform word segmentation and part-of-speech tagging on various content texts; Extract core keywords based on the TF-IDF algorithm; Expand associated tags through the pre-trained semantic model BERT.

5. The personalized search information recommendation method based on user behavior analysis according to claim 3, characterized in that, The formula for calculating the matching degree between the keyword tag and the user is: Among them, is the matching degree, C j is the cumulative value of the number of times the user collects the label t j , S j is the cumulative value of the number of clicks, Q j is the cumulative value of the number of times the search statement contains the label t j , w c , w s , w q are preset values.

6. The personalized search information recommendation method based on user behavior analysis according to claim 5, characterized in that Generate a matching degree set, perform a variance operation on the matching degree set corresponding to the keyword tag to obtain the sensitivity corresponding to the user content, including: Integrate all keywords in content c to obtain content c n n The matching degree set of the corresponding keyword tags:​ Set of matching degrees for keyword tags Perform variance calculation to obtain the corresponding single memory C n Sensitivity The formula is: Among them, f(U, T) is a function that defines the relationship between user behavior and keyword tags. represents the behavioral sensitivity of user U to content c n of the set of keyword tags where k is the number.

7. A personalized search information recommendation device based on user behavior analysis, characterized in that Including: A data collection module for real-time obtaining the user's interaction behavior data, including search behavior, click behavior, and favorite behavior, and configuring a preset weight value for each type of behavior; A keyword module for automatically extracting keyword tags of various contents through natural language processing technology to construct a content keyword tag system; According to the content keyword tag system, extract the keyword tags associated with the user's behavior from the interaction behavior data; A sensitivity module for calculating the cumulative values of the user's search behavior, click behavior, and favorite behavior based on the interaction behavior data; Combine the cumulative value of each type of behavior and the preset weight value to calculate the matching degree between the keyword tag and the user, and generate a matching degree set; Perform a variance operation on the matching degree set corresponding to the keyword tag to obtain the sensitivity corresponding to the user content; A recommendation module for generating recommended information for the user's search content according to the sensitivity.

8. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the personalized search information recommendation method based on user behavior analysis according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, A program is stored on the computer-readable storage medium, and when the program is executed by the processor, it implements the personalized search information recommendation method based on user behavior analysis according to any one of claims 1 to 6.

10. A computer program product, comprising computer instructions, characterized in that, When the computer instruction is executed by the processor, it implements the steps of the personalized search information recommendation method based on user behavior analysis according to claims 1 to 6.

Images