User grouping method and device, equipment, storage medium and program product

Abstract

The invention relates to the technical field of artificial intelligence, and provides a user grouping method and device, equipment, a storage medium and a program product. The method comprises the following steps: acquiring a plurality of keywords; based on the similarity between each keyword and each business tag, determining a target tag set; the target label set comprises a plurality of target service labels; determining the label weight of each target business label based on the modulus length of the semantic vector of each keyword, the text position of each keyword and the association relationship of all keywords; and grouping the users based on each target service label and the label weight of each target service label to obtain a target user group. Through the method, in the user grouping process, the grouping rule made by business personnel according to subjective judgment is not depended on any more, the user grouping result can be prevented from being influenced by personal subjective factors, and the accuracy of the user grouping result is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a user grouping method, device, equipment, storage medium and program product. Background Art

[0002] The labeling system plays a vital role in data processing and business operations. Its core function is to convert massive and diverse raw data into meaningful population classifications, thereby supporting accurate decision-making and application on the business side in various scenarios. This process mainly covers three core links: labeling system construction, user grouping, and application and effect evaluation of grouping data.

[0003] In the user segmentation stage, it is necessary to formulate appropriate segmentation strategies based on business needs and goals, and divide users into different user groups based on user portraits and segmentation strategies. The same user group should have similar characteristics and behavior patterns to facilitate subsequent precision marketing and services.

[0004] Common user grouping methods include rule-based label grouping methods. Rule-based label grouping methods divide users into different groups based on pre-set rules and user labels. These rules can be based on user attributes, behaviors, preferences, etc., and are suitable for situations with clear and fixed classification standards or business needs. They are simple and easy to use, and can quickly divide user groups according to business needs. At the same time, the setting of rules can be based on the professional knowledge and experience of business personnel, and have strong flexibility and interpretability.

[0005] However, since the formulation of rules often relies on the subjective judgment of business personnel, the grouping results may be affected by personal subjective factors. Therefore, existing user grouping methods may lead to inaccurate user grouping results. Summary of the invention

[0006] The embodiments of the present application provide a user grouping method, apparatus, device, storage medium and program product to solve the technical problem that the existing user grouping methods may lead to inaccurate user grouping results.

[0007] In a first aspect, an embodiment of the present application provides a user grouping method, comprising: obtaining multiple keywords; the keywords are extracted based on a demand text, and the demand text is a text that describes the user grouping needs; based on the similarity between each keyword and each business tag, a target tag set is determined; the target tag set includes multiple target business tags; based on the module length of the semantic vector of each keyword, the text position of each keyword, and the association relationship of all keywords, the tag weight of each target business tag is determined; based on each target business tag and the tag weight of each target business tag, users are grouped to obtain a target user group.

[0008] In one embodiment, based on the module length of the semantic vector of each keyword, the text position of each keyword, and the association relationship of all keywords, the tag weight of each target business tag is determined, including: performing semantic analysis on each keyword to obtain the semantic vector of each keyword, and determining the first weight of each target business tag based on the module length of each semantic vector; determining the second weight of each target business tag based on the text position of each keyword; determining the third weight of each target business tag based on the association relationship of all keywords; and determining the tag weight of each target business tag based on each first weight, each second weight, and each third weight.

[0009] In one embodiment, the similarity includes cosine similarity; based on the similarity between each keyword and each business tag, determining the target tag set includes: determining the cosine similarity between each keyword and each business tag respectively; based on each cosine similarity, performing tag screening to determine the target tag set; wherein each target business tag in the target tag set is a business tag whose cosine similarity is higher than a preset threshold.

[0010] In one embodiment, obtaining multiple keywords includes: obtaining a demand text; extracting keywords from the demand text based on a TF-IDF algorithm to obtain multiple first candidate keywords; extracting keywords from the demand text based on a TextRank algorithm to obtain multiple second candidate keywords; determining multiple candidate keywords based on each first candidate keyword and each second candidate keyword; and performing vocabulary merging processing based on the semantic similarity of each candidate keyword to obtain multiple keywords.

[0011] In one embodiment, it also includes: obtaining user behavior data and market trend monitoring data; based on the behavior data and market trend monitoring data, updating each business tag according to a preset update strategy; wherein the preset update strategy includes scheduled update and / or real-time update.

[0012] In one embodiment, before obtaining a plurality of keywords, the method further includes: obtaining a plurality of tags to be described; and adding a business description to each tag to be described to obtain a plurality of business tags.

[0013] In a second aspect, an embodiment of the present application provides a user grouping device, comprising: a keyword extraction module, used to obtain multiple keywords; the keywords are extracted based on the demand text, and the demand text is a text that describes the user grouping needs; a similarity calculation module, used to determine the target tag set based on the similarity between each keyword and each business tag; the target tag set includes multiple target business tags; a weight calculation module, used to determine the tag weight of each target business tag based on the module length of the semantic vector of each keyword, the text position of each keyword, and the association relationship of all keywords; a clustering module, used to group users based on each target business tag and the tag weight of each target business tag to obtain a target user group.

[0014] In a third aspect, an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any one of the user grouping methods described above is implemented.

[0015] In a fourth aspect, an embodiment of the present application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the user grouping methods described above.

[0016] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which implements any of the above-mentioned user grouping methods when executed by a processor.

[0017] The user grouping method, apparatus, device, storage medium and program product provided in the embodiments of the present application first extract multiple keywords from the demand text, determine multiple target business tags based on the similarity between each keyword and each business tag, and then determine the tag weight of each target business tag based on the module length of the semantic vector of each keyword, the text position of each keyword and the association relationship of all keywords, and group users according to each target business tag and the tag weight of each target business tag to obtain a target user group. In the user grouping process, it is no longer dependent on the grouping rules formulated by business personnel based on subjective judgment, which can prevent the user grouping results from being affected by personal subjective factors and improve the accuracy of the user grouping results. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1This is one of the flow charts of the user grouping method provided in the embodiment of the present application.

[0020] Figure 2 This is the second flow chart of the user grouping method provided in the embodiment of the present application.

[0021] Figure 3 This is the third flow chart of the user grouping method provided in the embodiment of the present application.

[0022] Figure 4 It is an online application architecture diagram of the user grouping method provided in an embodiment of the present application.

[0023] Figure 5 It is a structural diagram of a user grouping device provided in an embodiment of the present application.

[0024] Figure 6 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0026] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, subject to the requirements of applicable laws and regulations of the country where the user is located (for example, with the user's explicit consent, effective notification to the user, etc.).

[0027] See also Figure 1 , Figure 1 This is one of the flow charts of the user grouping method provided in the embodiment of the present application. In the embodiment of the present application, the user grouping method is applied to the tag system, and the user grouping method includes steps S110 to S140, and each step is as follows: S110: Acquire multiple keywords.

[0028] Keywords are extracted based on demand text, which is text that describes the needs of user groups.

[0029] See also Figure 2 , Figure 2 This is the second flow chart of the user grouping method provided in the embodiment of the present application.

[0030] Understandably, since the labels in traditional labeling systems are created by technical personnel, the label naming is rather technical. Therefore, in order to achieve label reuse and allow business personnel to more accurately target user groups, it is necessary to businessize the labels in the existing labeling system and add business descriptions to each existing label (i.e., the label to be described) (i.e., add descriptions of the business terms in the labels).

[0031] like Figure 2 As shown, the process of adding a business description to each existing label is called label system businessization.

[0032] The steps for commercializing the label system are as follows: (1) Data integration and construction: With users at the core, we use the user's unique identifier to link data from different platforms, systems, and channels, thereby building a global data center to achieve unified management and efficient use of data.

[0033] (2) Label research and development: Based on the construction of a global data center, various labels are developed. Labels generally include factual labels (such as gender, age group, etc.), business labels (such as membership level, purchase preferences, etc.), statistical labels (such as consumption frequency, average order value, etc.) and algorithmic labels (such as user value rating, activity prediction, etc.).

[0034] (3) Sorting out labels from a business perspective: From a business perspective, comprehensively sort out the entire label system, add business descriptions to each existing label, and obtain multiple business labels.

[0035] Furthermore, when business personnel need to perform user grouping, they can enter a requirement text on the label system, that is, a text describing the user grouping requirement.

[0036] For example, business personnel can manually input text (a sentence or a paragraph) that describes the characteristics of the target user group. Usually, it is a sentence, and the system will break down the input sentence into keywords. This process requires an effective text decomposition algorithm to ensure accurate extraction of key information.

[0037] Furthermore, the tag system can extract keywords from the demand text to obtain multiple keywords.

[0038] The process of keyword extraction requires an effective text decomposition algorithm to ensure accurate extraction of key information.

[0039] Optionally, the steps of keyword extraction are as follows: (1) The label system can preprocess the input demand text, split the text into individual words or phrases (e.g., using a tokenizer), remove stop words such as "de" (的) and "le" (了) that are common but contribute little to the meaning of the sentence, assign part-of-speech tags to each word or phrase (e.g., nouns, verbs, adjectives, etc.), and use programming languages and NLP libraries to achieve keyword extraction.

[0040] (2) Merge synonyms or near-synonyms, and merge the identical or similar words or phrases extracted through a dictionary or semantic similarity calculation.

[0041] S120: Determine the target label set based on the similarity between each keyword and each business label.

[0042] The target label set includes multiple target business labels.

[0043] Specifically, for each extracted keyword, perform a similarity matching calculation between the keyword and each business label in the label system. If the similarity between the keyword and a certain business label is greater than a preset threshold, it can be considered that the keyword matches the business label, and the business label can be screened and classified into the target label set.

[0044] The target label set contains all the target business labels matched by each keyword.

[0045] Understandably, the preset threshold can be adjusted according to actual application requirements to balance the matching accuracy and recall rate.

[0046] S130: Determine the label weight of each target business label based on the norm of the semantic vector of each keyword, the text position of each keyword, and the association relationship between all keywords.

[0047] Generally, a weight model can be established to assign label weights to each target business label in the label system. The label weights can be determined based on factors such as label frequency, user interaction, and relevance to business goals. The higher the label weight, the more important the target business label is in the target user group positioning. The label weight calculation for each target business label can be set based on the extracted keywords (e.g., if the keyword is "young users", then the age attribute may be given a higher weight). The label system can update the label weights regularly to adapt to changes in the market and user behavior, ensure the consistency between the label system and business goals, and enhance the accuracy of target user group matching.

[0048] In this embodiment, the label weight of each target business label is determined based on the norm of the semantic vector of each keyword, the text position of each keyword, and the association relationship between all keywords.

[0049] S140: Grouping users based on each target service tag and the tag weight of each target service tag to obtain a target user group.

[0050] The improvement of the above user grouping steps can improve the accuracy and efficiency of user grouping and lower the threshold for using the tag system. The extraction of keywords can more accurately capture the characteristics of the user group, and the optimization of the similarity comparison method can ensure effective matching with the improved tag system. Finally, the tag weight is calculated to accurately match the target user group.

[0051] The user clustering method provided in the embodiment of the present application first extracts multiple keywords from the demand text, determines multiple target business tags based on the similarity between each keyword and each business tag, and then determines the tag weight of each target business tag based on the module length of the semantic vector of each keyword, the text position of each keyword and the association relationship of all keywords, and clusters users according to each target business tag and the tag weight of each target business tag to obtain a target user group. In the user clustering process, it no longer relies on the grouping rules formulated by business personnel based on subjective judgment, which can avoid the user grouping results being affected by personal subjective factors and improve the accuracy of the user grouping results.

[0052] In one embodiment, based on the module length of the semantic vector of each keyword, the text position of each keyword, and the association relationship of all keywords, the tag weight of each target business tag is determined, including: performing semantic analysis on each keyword to obtain the semantic vector of each keyword, and determining the first weight of each target business tag based on the module length of each semantic vector; determining the second weight of each target business tag based on the text position of each keyword; determining the third weight of each target business tag based on the association relationship of all keywords; and determining the tag weight of each target business tag based on each first weight, each second weight, and each third weight.

[0053] In order to further improve the accuracy and depth of keyword analysis and assign reasonable label weights to each target business label, in this embodiment, a weight calculation strategy that comprehensively considers semantics, lexical position, and the relationship between words is adopted. This strategy can not only accurately capture the importance of each keyword, but also ensure that the label weight of the keyword mapping can truly reflect its role in constructing the user portrait.

[0054] Specifically, based on the similarity between each keyword and each business tag, the target tag is determined, semantic analysis is performed on each keyword to obtain the semantic vector of each keyword, and based on the modulus length of each semantic vector, the first weight of each target business tag is determined.

[0055] Using semantic analysis technology, such as semantic analysis models such as Word2Vec or BERT, each keyword is semantically analyzed and converted into a corresponding semantic vector. These semantic vectors can capture the subtle differences in the meaning of each keyword, so that the label weight calculation is not only based on the word frequency, but also on the deep meaning of the word.

[0056] For a keyword and a target business tag corresponding to the keyword, the tag weight of the target business tag corresponding to the keyword can be evaluated by the modulus of the semantic vector of the keyword, the distance between vectors, or other vector characteristics. In this embodiment, the modulus of the semantic vector is used to determine the first weight of the target business tag. The first weight The calculation formula is as follows: ; in, is the semantic vector of keyword i; d is the dimension of the semantic vector; Represents the vector representation of the kth dimension of keyword i.

[0057] Furthermore, based on the text position of each keyword, a second weight of each target business tag is determined.

[0058] Taking into account the influence of the text position of a keyword (ie, the position of the keyword in the requirement text) on its importance, in this embodiment, a position weight factor is introduced.

[0059] Generally speaking, keywords at the beginning and end of a text may be more important than keywords in the middle of the text. Based on this, the position weight can be adjusted using the inverse formula.

[0060] For a keyword and a target business tag corresponding to the keyword, the second weight of the target business tag can be determined according to the text position of the keyword. The calculation formula is as follows: ; in, is the position of the ith keyword in the requirement text.

[0061] Furthermore, based on the association relationship of all keywords, a third weight of each target business tag is determined.

[0062] The association between multiple keywords can be analyzed by constructing a co-occurrence network or dependency tree. In a co-occurrence network or dependency tree, the weight of the edge between two keywords can reflect the strength or distance of the relationship between the two keywords, which can be used to further adjust the weight of the keywords. For example, if the co-occurrence frequency of two keywords in the co-occurrence network is high, the two keywords may have a higher joint importance to the topic of the demand text, so in the co-occurrence matrix, the weight can be a function of the co-occurrence frequency of the two keywords.

[0063] For a target business tag in all keywords and target tag sets, the third weight of the target business tag can be determined according to the association relationship of all keywords. The calculation formula is as follows: ; in, is the number of co-occurrences of keywords i and j in the demand text; is the total number of occurrences in the requirement text.

[0064] Further, based on each first weight, each second weight and each third weight, a label weight of each target service label is determined, and the label weight The calculation formula is as follows: .

[0065] By comprehensively considering semantics, lexical position, and the relationship between words, the label weight of the target business label corresponding to each keyword can be calculated more accurately to ensure that it reflects the actual value and influence of the keyword in constructing the user portrait.

[0066] The user clustering method provided in the embodiment of the present application further refines the analysis of keywords. By examining the semantics of the keywords in the demand text, the text position and the association relationship with different keywords, the label weight of the target business label mapped to each keyword is accurately calculated. This process aims to evaluate the accuracy and influence of the labeled keywords in matching the target user group, and ensure that the label weight of the target business label corresponding to the keyword reflects the true value of the keyword to the construction of the user portrait.

[0067] In one embodiment, the similarity comprises cosine similarity.

[0068] Based on the similarity between each keyword and each business tag, a target tag set is determined, including: respectively determining the cosine similarity between each keyword and each business tag; based on each cosine similarity, performing tag screening to determine a target tag set; wherein each target business tag in the target tag set is a business tag whose cosine similarity is higher than a preset threshold.

[0069] Cosine similarity is a commonly used method to calculate the similarity between two vectors. It evaluates the similarity between them by measuring the cosine value of the angle between the two vectors. In text processing, each keyword and each tag can be represented as a vector, and the elements in the vector may be word frequency or TF-IDF value, etc. The cosine similarity value ranges from -1 to 1, where 1 means that the vectors are exactly the same, 0 means that the two are completely unrelated, and -1 means that the two are completely opposite. The formula for calculating cosine similarity is as follows: ; Among them, A and B represent two vectors respectively; Represents the dot product of vector A and vector B; represents the vector modulus of vector A, represents the vector modulus of vector B, and The calculation formula is as follows: ; ; The calculation formula is as follows: ; in, and are the elements of vector A and vector B at position i respectively.

[0070] Specifically, in this embodiment, each keyword and each service label are converted into a vector form to obtain each keyword vector and each service label vector.

[0071] For example, a weight can be assigned to each keyword in the demand text through word frequency, TF-IDF, etc. to form a vector.

[0072] For each keyword vector and the corresponding business label vector, the cosine similarity between the keyword and the business label is calculated using the above cosine similarity calculation formula.

[0073] After obtaining the cosine similarity between each keyword and each business tag, tag screening is performed based on each cosine similarity to determine a target tag set, in which each target business tag in the target tag set is a business tag whose cosine similarity is higher than a preset threshold.

[0074] The user clustering method provided in the embodiment of the present application calculates the similarity between each keyword and each business tag through the cosine similarity formula, and quantifies the proximity between the keyword and the business tag in a mathematical way, so that the tag system can assign the business tag that best matches the text entered by each user, thereby improving the accuracy and efficiency of tag selection.

[0075] In one embodiment, obtaining multiple keywords includes: obtaining a demand text; extracting keywords from the demand text based on a TF-IDF algorithm to obtain multiple first candidate keywords; extracting keywords from the demand text based on a TextRank algorithm to obtain multiple second candidate keywords; determining multiple candidate keywords based on each first candidate keyword and each second candidate keyword; and performing vocabulary merging processing based on the semantic similarity of each candidate keyword to obtain multiple keywords.

[0076] By making full use of advanced automatic text recognition technology, we can carefully analyze the text information contained in the user input content in order to make a more refined division of user groups.

[0077] In this embodiment, advanced NLP (Natural Language Processing) technology, especially the TF-IDF (Term Frequency-Inverse Document Frequency) algorithm and the TextRank algorithm, is used to extract keywords from the interactive content between the tag system and business personnel (i.e., the text entered by the user).

[0078] Specifically, obtain the requirement text.

[0079] Furthermore, based on the TF-IDF algorithm, keywords are extracted from the demand text to obtain a plurality of first candidate keywords.

[0080] The TF-IDF algorithm combines two statistical indicators, term frequency (TF) and inverse document frequency (IDF), to evaluate the importance of a word in a set of texts or a document. The calculation formula is as follows: ; ; Among them, IDF reflects the general importance of a word. If a word appears frequently in many documents, its IDF value will be reduced; D is the total number of documents; is the number of documents containing word i; TF refers to the frequency of a word appearing in a document, which is calculated by dividing the number of times the word appears in the document by the total number of words in the document; the higher the TF-IDF, the more common the word is in the document and can provide a lot of information, and is a keyword in the text.

[0081] Furthermore, based on the TextRank algorithm, keywords are extracted from the demand text to obtain a plurality of second candidate keywords.

[0082] The TextRank algorithm is a graph-based ranking algorithm used for keyword extraction and document summary generation in text processing. The algorithm constructs a graph in which the nodes are words in the text and the edges reflect the co-occurrence or semantic relationship between words. The score of each node is calculated iteratively, and nodes with high scores are regarded as keywords in the text. The TextRank algorithm does not rely on an external corpus and can automatically discover key information from the text itself.

[0083] Further, based on each first candidate keyword and each second candidate keyword, a plurality of candidate keywords are determined.

[0084] Optionally, the candidate keywords include a first candidate keyword and each second candidate keyword.

[0085] The first candidate keyword and the second candidate keyword not only describe the core theme of the demand text, but also deeply reflect the core information such as the user's interests, behavioral characteristics and preferences. The TF-IDF algorithm can identify unique and important words in the text, while the TextRank algorithm starts from the structure of the text and identifies keywords that have important roles and relevance in the text. Combining these two algorithms can efficiently extract the most representative and valuable elements from the huge text data, ensuring the accuracy and depth in the analysis and application process.

[0086] Furthermore, based on the semantic similarity of each candidate keyword, a vocabulary merging process is performed to obtain multiple keywords.

[0087] The user grouping method provided in the embodiment of the present application is an optimization of the existing labeling system. It identifies the needs of business personnel based on an automated text recognition method, so that business personnel do not need to understand the labeling system and related technologies. They only need to input the required text to obtain the corresponding user grouping, which greatly reduces the operating threshold of the labeling system and improves the accuracy of user grouping.

[0088] In one embodiment, it also includes: obtaining user behavior data and market trend monitoring data; based on the behavior data and market trend monitoring data, updating each business tag according to a preset update strategy; wherein the preset update strategy includes scheduled update and / or real-time update.

[0089] In order to maintain the real-time and accuracy of the label system, this application introduces an innovative dynamic label update mechanism, which aims to automatically update and optimize the label library by real-time monitoring and analysis of user behavior data and market trends.

[0090] Specifically, obtain user behavior data.

[0091] The tag system can track and analyze users' interactive content and behaviors in real time, such as browsing history, search queries and purchasing behaviors, and obtain users' behavioral data. Through users' behavioral data, the tag system can capture subtle changes in users' interests and needs.

[0092] Further, obtain market trend monitoring data.

[0093] In parallel with user behavior analysis, the labeling system can continuously monitor and analyze external factors such as industry dynamics, hot events and seasonal trends to obtain market trend monitoring data. The market trend monitoring data can assist the labeling system in predicting market changes and guiding label updates.

[0094] Furthermore, based on the behavior data and market trend monitoring data, each business tag is updated according to a preset update strategy.

[0095] The preset update strategy includes scheduled update and / or real-time update.

[0096] Specifically, the label system can combine user behavior data and market trend monitoring data, and use a series of algorithm models to determine the addition, modification or deletion of business labels, and update each business label according to the preset update strategy.

[0097] Optionally, the algorithm model includes but is not limited to cluster analysis, trend prediction models and natural language processing technology, which work together to ensure that the business tags of the tag system can reflect the latest user needs and market conditions.

[0098] Preferably, a preset update strategy combining scheduled updates and real-time updates is adopted: the label system adopts a dual-track mechanism of scheduled updates and real-time updates. The scheduled update strategy is used to capture long-term trends and cyclical changes, while the real-time update strategy is used to target emergencies and rapidly changing trends, ensuring the sensitivity and adaptability of the label system.

[0099] The label system has set up a feedback mechanism to evaluate the effect of label updates by monitoring the application effects of business labels (such as user click-through rate, conversion rate and other indicators). This feedback information is used to optimize subsequent preset update strategies, forming a cycle of continuous improvement.

[0100] The user grouping method provided in the embodiment of the present application adopts a dynamic label update mechanism, which can not only flexibly respond to changes in the market and user needs, but also continuously improve the relevance and accuracy of the label system. This highly adaptive and real-time update feature ensures that the results of user portrait grouping are both accurate and practical, greatly enhancing the company's competitiveness in the market and user satisfaction.

[0101] In one embodiment, before obtaining a plurality of keywords, the method further includes: obtaining a plurality of tags to be described; and adding a business description to each tag to be described to obtain a plurality of business tags.

[0102] The present application also provides a specific example of a user grouping method. Figure 3 and Figure 4 , Figure 3 This is the third flow chart of the user grouping method provided in the embodiment of the present application. Figure 4 It is an online application architecture diagram of the user grouping method provided in an embodiment of the present application.

[0103] like Figure 3 As shown in the figure, the user grouping method is applied to the management of the online work of survey and design personnel by managers. Based on their understanding of the business, managers enter the demand text used to describe the characteristics of the personnel to be surveyed in this work into the label system.

[0104] The labeling system performs data preprocessing on the input requirement text.

[0105] like Figure 4 As shown, the input management and data preprocessing modules of the labeling system can perform data cleaning, noise removal, word segmentation and frequency analysis on the demand text, remove noise, irrelevant characters and repeated content in the demand text, and ensure the quality of the data.

[0106] Furthermore, the demand text is segmented and word frequency statistics are performed to count the frequency of occurrence of each word, providing a data basis for subsequent keyword extraction.

[0107] Preferably, before obtaining the keyword, the dynamic tag database (the dynamic tag database stores all business tags) is dynamically updated to ensure that the business tags in the dynamic tag database are the latest data.

[0108] Furthermore, keyword extraction and label mapping are performed on the demand text through the TF-IDF algorithm and the TextRank algorithm to extract keywords and obtain multiple keywords, which can usually reflect the core content of the text; based on the similarity between each keyword and each business label, the target label set is determined; the target label set includes multiple target business labels.

[0109] Furthermore, a weight calculation is performed: based on the module length of the semantic vector of each keyword, the text position of each keyword, and the association relationship of all keywords, the tag weight of each target business tag is determined.

[0110] Furthermore, based on each of the target service tags and the tag weight of each of the target service tags, the users are grouped to obtain a target user group.

[0111] The present application also provides a user grouping device. Figure 5 , Figure 5 5 is a schematic diagram of the structure of the user grouping device provided in the embodiment of the present application. In the embodiment of the present application, the user grouping device includes a keyword extraction module 510, a similarity calculation module 520, a weight calculation module 530 and a grouping module 540.

[0112] The keyword extraction module 510 is used to obtain multiple keywords.

[0113] Keywords are extracted based on demand text, which is text that describes the needs of user groups.

[0114] The similarity calculation module 520 is used to determine a target tag set based on the similarity between each keyword and each business tag.

[0115] The target tag set includes multiple target business tags.

[0116] The weight calculation module 530 is used to determine the tag weight of each target business tag based on the module length of the semantic vector of each keyword, the text position of each keyword, and the association relationship of all keywords.

[0117] The grouping module 540 is used to group users based on each target service tag and the tag weight of each target service tag to obtain a target user group.

[0118] In some embodiments, the weight calculation module 530 is used to perform semantic analysis on each keyword, obtain the semantic vector of each keyword, and determine the first weight of each target business tag based on the module length of each semantic vector; determine the second weight of each target business tag based on the text position of each keyword; determine the third weight of each target business tag based on the association relationship of all keywords; determine the tag weight of each target business tag based on each first weight, each second weight and each third weight.

[0119] In some embodiments, the similarity comprises cosine similarity.

[0120] The similarity calculation module 520 is used to determine the cosine similarity between each keyword and each business tag respectively; based on each cosine similarity, perform tag screening to determine a target tag set; wherein each target business tag in the target tag set is a business tag whose cosine similarity is higher than a preset threshold.

[0121] In some embodiments, the keyword extraction module 510 is used to obtain the demand text; based on the TF-IDF algorithm, perform keyword extraction on the demand text to obtain multiple first candidate keywords; based on the TextRank algorithm, perform keyword extraction on the demand text to obtain multiple second candidate keywords; based on each first candidate keyword and each second candidate keyword, determine multiple candidate keywords; based on the semantic similarity of each candidate keyword, perform vocabulary merging processing to obtain multiple keywords.

[0122] In one embodiment, the user grouping device further includes an acquisition module.

[0123] The acquisition module is used to obtain user behavior data and market trend monitoring data; based on the behavior data and market trend monitoring data, each business tag is updated according to a preset update strategy; wherein the preset update strategy includes scheduled update and / or real-time update.

[0124] In one embodiment, the acquisition module is used to acquire multiple tags to be described; and to add a service description to each tag to be described to obtain multiple service tags.

[0125] The present application also provides an electronic device, Figure 6 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the user grouping method.

[0126] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0127] An embodiment of the present application also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the user grouping method provided by the above methods is implemented.

[0128] An embodiment of the present application also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the user grouping methods provided by the above methods.

[0129] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0130] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A user grouping method, It is characterized in that include: Acquire multiple keywords; the keywords are extracted based on the demand text, and the demand text is a text describing the user grouping needs; Determine a target tag set based on the similarity between each of the keywords and each of the business tags; the target tag set includes a plurality of target business tags; Determining a tag weight of each target service tag based on the modulus of the semantic vector of each keyword, the text position of each keyword, and the association relationship of all the keywords; Based on each of the target service tags and the tag weight of each of the target service tags, users are grouped to obtain a target user group.

2. The user grouping method according to claim 1, It is characterized in that The determining of the tag weight of each target service tag based on the module length of the semantic vector of each keyword, the text position of each keyword, and the association relationship of all the keywords includes: Performing semantic analysis on each of the keywords to obtain a semantic vector of each of the keywords, and determining a first weight of each of the target service tags based on a modulus length of each of the semantic vectors; Determining a second weight of each of the target business tags based on a text position of each of the keywords; Determining a third weight of each of the target business tags based on the association relationship of all the keywords; A label weight of each of the target service labels is determined based on each of the first weights, each of the second weights, and each of the third weights.

3. The user grouping method according to claim 1, It is characterized in that The similarity includes cosine similarity; The determining of the target tag set based on the similarity between each keyword and each business tag includes: Determining the cosine similarity between each of the keywords and each of the business tags respectively; Based on each of the cosine similarities, label screening is performed to determine a target label set; Each of the target service tags in the target tag set is a service tag whose cosine similarity is higher than a preset threshold.

4. The user grouping method according to claim 1, It is characterized in that The obtaining of multiple keywords includes: Obtaining the requirement text; Based on the TF-IDF algorithm, extract keywords from the demand text to obtain multiple first candidate keywords; Based on the TextRank algorithm, keywords are extracted from the demand text to obtain multiple second candidate keywords; determining a plurality of candidate keywords based on each of the first candidate keywords and each of the second candidate keywords; Based on the semantic similarity of each candidate keyword, a vocabulary merging process is performed to obtain a plurality of keywords.

5. The user grouping method according to claim 1, It is characterized in that Also includes: Obtain user behavior data and market trend monitoring data; Based on the behavior data and the market trend monitoring data, updating each of the service tags according to a preset update strategy; The preset update strategy includes scheduled update and / or real-time update.

6. The user grouping method according to claim 1, It is characterized in that Before obtaining the multiple keywords, the method further includes: Get multiple tags to be described; For each of the tags to be described, a service description is added to obtain multiple service tags.

7. A user grouping device, It is characterized in that include: A keyword extraction module is used to obtain multiple keywords; the keywords are extracted based on the demand text, and the demand text is a text describing the user grouping needs; A similarity calculation module, used to determine a target tag set based on the similarity between each of the keywords and each of the business tags; the target tag set includes a plurality of target business tags; A weight calculation module, used to determine the tag weight of each target business tag based on the module length of the semantic vector of each keyword, the text position of each keyword, and the association relationship of all the keywords; The grouping module is used to group users based on each target service tag and the tag weight of each target service tag to obtain a target user group.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the user grouping method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the user grouping method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the user grouping method according to any one of claims 1 to 6 is implemented.

Images