Sales lead processing method, apparatus, device, medium, and product

Abstract

Embodiments of the present application provide a sales lead processing method, device, equipment and medium. The method comprises: obtaining a sales lead request, determining a target intent label corresponding to the sales lead request based on a pre-determined intent label and the association relationship information between the intent labels. Matching the target intent label with the user label in the user information, and screening out target users that meet the target intent label. This scheme can automatically understand the intent of the sales lead, accurately convert the sales lead request into the target intent label based on the pre-prepared association relationship information of the intent label, deeply understand the complex semantics and deep needs in the sales lead request, and effectively make up for the shortcomings of traditional methods in handling complex semantic relationships and understanding deep user needs. Quickly and accurately screen out the corresponding target users, greatly improve the sales lead mining efficiency and accuracy, reduce the manual processing cost and time, and improve the efficiency of sales lead processing.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a sales lead processing method, a sales lead processing device, an electronic device, a readable storage medium, and a computer program product. Background Technology

[0002] With the explosive growth of enterprise data, the efficiency and accuracy of traditional manual lead generation methods are no longer sufficient to meet the demands. Although existing technologies have developed lead screening tools based on keyword search and machine learning algorithms, these methods still have significant limitations in handling complex semantic relationships and understanding deep user needs. Summary of the Invention

[0003] The technical problem to be solved by the embodiments of the present invention is to provide a sales lead processing method, apparatus, electronic device, readable storage medium and computer program product, so as to solve the problem that there are still obvious limitations in handling complex semantic relationships and understanding deep user needs when mining sales leads.

[0004] To address the above problems, the present invention provides a sales lead processing method, the method comprising:

[0005] Request sales leads;

[0006] Based on the pre-determined intent tags and the correlation information between the intent tags, the target intent tag corresponding to the sales lead request is determined;

[0007] The target intent tags are matched with user tags in the user information to filter out target users who match the target intent tags.

[0008] Optionally, the target intent tag includes a first target intent tag and a second target intent tag. Determining the target intent tag corresponding to the sales lead request based on pre-determined intent tags and the correlation information between the intent tags includes:

[0009] Based on the sales lead request, determine the corresponding first target intent tag from the pre-determined intent tags;

[0010] The second target intent tag is determined based on the correlation information between the first target intent tag and the intent tags.

[0011] Optionally, prior to the request for obtaining sales leads, the method further includes:

[0012] Identify relevant data from the business scenario to obtain intent tags for that business scenario;

[0013] Calculate the semantic similarity between intent tags in the aforementioned business scenario;

[0014] Based on the intent tags in the business scenario and the semantic similarity, the association information between the intent tags is determined.

[0015] Optionally, the intent tags in the business scenario include key information and related tags. The process of identifying relevant data in the business scenario to obtain the intent tags in the business scenario includes:

[0016] Extract key information from relevant data in the aforementioned business scenario;

[0017] Based on the relevant data of the business scenario, the corresponding relevant tags are generated using the first content generation model.

[0018] Optionally, the association information includes the association relationships and semantic similarity between the intent tags, and determining the association relationship information between the intent tags based on the intent tags in the business scenario and the semantic similarity includes:

[0019] Based on the intent tags in the business scenario and the semantic similarity, an intent tag graph data is constructed; wherein, the intent tag graph data includes the relationships and semantic similarity between the intent tags.

[0020] Optionally, the intent tag map data further includes: multi-level classification information of the intent tags, wherein the multi-level classification information represents the category affiliation relationship between the intent tags; before determining the corresponding first target intent tag from the predetermined intent tags based on the sales lead request, the method further includes:

[0021] Training data is constructed based on the multi-level classification information of the sales lead request samples and the intent tags; wherein, the training data includes the sales lead request samples and at least one level of intent tags corresponding to the tags;

[0022] Using the training data, a multi-level intent recognition model is trained;

[0023] The step of determining the corresponding first target intent tag from the pre-determined intent tags based on the sales lead request includes:

[0024] The sales lead request is identified using a trained multi-level intent recognition model to obtain the first target intent label.

[0025] Optionally, the intent tag graph data further includes: vector data corresponding to the intent tag, and the step of determining the corresponding first target intent tag from the predetermined intent tags based on the sales lead request includes:

[0026] When the multi-level intent recognition model fails to recognize the sales lead request, the vector data corresponding to the sales lead request generated by the multi-level intent recognition model is used to perform similarity calculation with the vector data corresponding to the intent tag in the intent tag graph data to obtain a similarity result.

[0027] Based on the similarity results, the first target intent tag is determined.

[0028] Optionally, the step of matching the target intent tag with user tags in user information to filter out target users who match the target intent tag includes:

[0029] The first target intent tag is matched with the user tags in the user information to filter out target users who match the first target intent tag;

[0030] If the number of target users matching the first target intent tag is less than the number required by the sales lead request, the second target intent tag is matched with the user tags in the user information to filter out the missing number of target users matching the second target intent tag.

[0031] Optionally, the step of matching the second target intent tag with the user tags in the user information to filter out a shortage of target users matching the second target intent tag includes:

[0032] The proportion of each second target intent tag is determined based on the semantic similarity between the second target intent tag and the first target intent tag;

[0033] The second target intent tag is matched with the user tags in the user information to obtain candidate users that match each of the second target intent tags;

[0034] Based on the stated proportion, select from the candidate users corresponding to each of the second target intent tags to obtain the number of target users in short supply.

[0035] Optionally, after matching the target intent tag with user tags in the user information to filter out target users who match the target intent tag, the method further includes:

[0036] Receive query requests for any of the target users;

[0037] Based on the user information of the target user, the second content generation model is used to generate the introduction information and communication strategy information of the target user;

[0038] The user details page displays the target user's introduction information and the communication strategy information.

[0039] The present invention also provides a sales lead processing device, the device comprising:

[0040] The request retrieval module is used to retrieve sales lead requests.

[0041] The tag determination module is used to determine the target intent tag corresponding to the sales lead request based on the pre-determined intent tags and the correlation information between the intent tags;

[0042] The user filtering module is used to match the target intent tags with user tags in the user information to filter out target users who match the target intent tags.

[0043] Optionally, the target intent label includes a first target intent label and a second target intent label, and the label determination module includes:

[0044] The first tag determination submodule is used to determine the corresponding first target intent tag from the pre-determined intent tags based on the sales lead request;

[0045] The second tag determination submodule is used to determine the second target intent tag based on the correlation information between the first target intent tag and the intent tag.

[0046] Optionally, the device further includes:

[0047] The identification module is used to identify relevant data of the business scenario before the request to obtain sales leads, and to obtain intent tags under the business scenario;

[0048] The similarity calculation module is used to calculate the semantic similarity between intent tags in the business scenario.

[0049] The association determination module is used to determine the association information between the intent tags based on the intent tags in the business scenario and the semantic similarity.

[0050] Optionally, the intent label in the business scenario includes key information and related labels, and the recognition module includes:

[0051] An extraction submodule is used to extract key information from relevant data in the business scenario.

[0052] The tag generation submodule is used to generate corresponding relevant tags based on the relevant data of the business scenario using the first content generation model.

[0053] Optionally, the association information includes the association relationships and semantic similarity between the intent tags, and the association determination module includes:

[0054] The graph construction submodule is used to construct intent tag graph data based on the intent tags in the business scenario and the semantic similarity; wherein, the intent tag graph data includes the correlation and semantic similarity between the intent tags.

[0055] Optionally, the intent tag map data further includes: multi-level classification information of the intent tags, wherein the multi-level classification information represents the category affiliation relationship between the intent tags; the device further includes:

[0056] The training data construction module is used to construct training data based on the sales lead request samples and the multi-level classification information of the intent tags before determining the corresponding first target intent tag from the pre-determined intent tags according to the sales lead request; wherein, the training data includes the sales lead request samples and at least one level of intent tags corresponding to the tags;

[0057] The training module is used to train a multi-level intent recognition model using the training data;

[0058] The first determining submodule includes:

[0059] The identification unit is used to identify the sales lead request using a trained multi-level intent recognition model to obtain the first target intent label.

[0060] Optionally, the intent label graph data further includes: vector data corresponding to the intent label, and the first determining submodule includes:

[0061] The similarity calculation unit is used to calculate the similarity between the vector data corresponding to the sales lead request generated by the multi-level intent recognition model and the vector data corresponding to the intent tag in the intent tag graph data when the multi-level intent recognition model cannot recognize the sales lead request, and to obtain the similarity result.

[0062] The first determining unit is used to determine the first target intent tag based on the similarity result.

[0063] Optionally, the user filtering module includes:

[0064] The matching submodule is used to match the first target intent tag with the user tags in the user information and filter out target users that match the first target intent tag.

[0065] The filtering submodule is used to match the second target intent tag with the user tags in the user information when the number of target users matching the first target intent tag is less than the number required by the sales lead request, and to filter out the missing number of target users matching the second target intent tag.

[0066] Optionally, the filtering submodule includes:

[0067] The proportion determination unit is used to determine the proportion of each second target intent tag based on the semantic similarity between the second target intent tag and the first target intent tag;

[0068] A matching unit is used to match the second target intent tag with the user tags in the user information to obtain candidate users that match each of the second target intent tags;

[0069] The selection unit is used to select from the candidate users corresponding to each of the second target intent tags according to the quantity ratio, so as to obtain the target users with the shortage quantity.

[0070] Optionally, the device further includes:

[0071] The request receiving module is configured to receive a query request for any of the target users after the target intent tag is matched with the user tags in the user information and the target users that match the target intent tag are filtered out.

[0072] The information generation module is used to generate introduction information and communication strategy information for the target user based on the user information of the target user using a second content generation model;

[0073] The display module is used to display the target user's introduction information and the communication strategy information on the user details page.

[0074] This invention also discloses an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0075] Memory, used to store computer programs;

[0076] The processor, when executing a program stored in memory, implements the method described above.

[0077] This invention also discloses a readable storage medium that, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to perform the method described above.

[0078] This invention also discloses a computer program product that, when run on an electronic device, causes a processor to execute the steps in the method described above.

[0079] According to an embodiment of the present invention, a sales lead request is first obtained. Then, based on pre-determined intent tags and the correlation information between the intent tags, a target intent tag corresponding to the sales lead request is determined. Finally, the target intent tag is matched with user tags in user information to filter out target users who match the target intent tag. This solution can automatically understand the intent of sales leads. Based on the pre-prepared correlation information of intent tags, it accurately transforms sales lead requests into target intent tags, deeply understands the complex semantics and deep needs in sales lead requests, and effectively makes up for the shortcomings of traditional methods in handling complex semantic relationships and understanding deep user needs. It not only quickly and accurately filters out the corresponding target users, but also greatly improves the efficiency and accuracy of sales lead mining, reduces manual processing costs and time, and improves the efficiency of sales lead processing. Attached Figure Description

[0080] Figure 1 A flowchart illustrating the steps of a sales lead processing method according to an embodiment of the present invention is shown.

[0081] Figure 2 A schematic diagram of a multi-level intent labeling system is shown.

[0082] Figure 3 A schematic diagram of the extended intent label is shown;

[0083] Figure 4 A diagram illustrating the question-and-answer process is shown;

[0084] Figure 5 A schematic diagram of the target user list page is shown;

[0085] Figure 6 A schematic diagram of a user details page is shown;

[0086] Figure 7 A schematic diagram of the sales lead request processing system is shown.

[0087] Figure 8 The diagram shows a structural block diagram of a sales lead processing device embodiment provided by one embodiment of the present invention;

[0088] Figure 9A structural block diagram of an electronic device for processing sales leads is shown according to an exemplary embodiment. Detailed Implementation

[0089] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0090] The sales lead processing method in this embodiment of the invention can be executed by electronic devices such as computers, servers, and server clusters, or by any other applicable hardware and software. This embodiment of the invention does not impose any restrictions on this.

[0091] Reference Figure 1 The diagram illustrates a flowchart of a sales lead processing method provided by an embodiment of the present invention, which may specifically include the following steps:

[0092] Step 101: Obtain a sales lead request.

[0093] In this embodiment of the invention, a sales lead request refers to a user's (such as a salesperson or system operator) request to the system to obtain information about potential users who meet specific criteria. Such requests are typically expressed in natural language; for example, the system receives text or voice input from the user, performs preliminary standardization processing (such as speech-to-text conversion and error correction), and then obtains the sales lead request.

[0094] For example, when the system receives a user's voice message "Filter users who are interested in test driving", the system converts this voice message into a sales lead request, which is the text form "Filter users who are interested in test driving".

[0095] Step 102: Based on the pre-determined intent tags and the relationship information between intent tags, determine the target intent tag corresponding to the sales lead request.

[0096] In this embodiment of the invention, intent tags are abstract representations of user intent in sales lead requests. They are a series of tags pre-identified based on business scenarios and needs, representing different sales intents. These tags are typically refined based on business scenarios and historical data to describe potential user needs and behaviors. For example, in a car sales scenario, intent tags might include "test drive intention," "price inquiry," and "model preference."

[0097] In this embodiment of the invention, the association information describes the logical relationships, semantic similarities, and possible combinations between different intent tags. For example, some tags may have a hierarchical relationship (such as "test drive satisfaction" and "test drive experience"), or they may have certain semantic similarities. These relationships can be determined through semantic similarity calculation, business logic analysis, etc. The association information is usually stored in the form of a graph or matrix to support complex intent recognition and expansion.

[0098] In this embodiment of the invention, the intent tags and the relationships between them are predetermined. For example, based on relevant data from the business scenario, all intent tags are extracted and identified, and relationship information is created by calculating the semantic similarity between intent tags.

[0099] In this embodiment of the invention, the target intent tag is the intent tag that best matches the user's needs, determined based on the sales lead request and pre-defined intent tags and their interrelationships. It represents the system's precise understanding and abstraction of the user's request, guiding subsequent user screening and business operations.

[0100] In this embodiment of the invention, a pre-built knowledge system is utilized, including intent tags and their interrelationships. Through this pre-defined data, the system can quickly locate and understand the key intents in sales lead requests. By combining the sales lead request with the relational information, the system identifies the target intent tag that best matches the requested intent from among the pre-determined intent tags.

[0101] For example, target intent tags can be divided into two types: one is seed intent tags, which are determined directly based on sales lead requests, and the other is extended intent tags that are determined based on seed intent tags and related relationship information and are associated with seed intent tags.

[0102] In one specific implementation, natural language processing techniques (such as keyword extraction and semantic analysis) are used to analyze sales lead requests and identify key information and potential intents. The identified key information is then matched against a pre-built intent tag library, and the relationships between intent tags (such as semantic similarity and hierarchical structure) are considered to determine the target intent tag that best matches the request's intent. This may involve multiple rounds of matching and filtering to ensure the accuracy of the results.

[0103] For example, when a request is made to "find some users of enterprise users", the system analyzes and identifies the key information "enterprise users" and combines it with related tags in the intent tag library (such as "enterprise users" as a tag, and other tags that may be related, such as "company users") to determine that the corresponding target intent tag may be "enterprise users" or "company users".

[0104] Step 103: Match the target intent tags with the user tags in the user information to filter out target users who match the target intent tags.

[0105] In this embodiment of the invention, user information is a data set stored in the system regarding potential or existing users, including basic user information, behavioral data, preference data, user consultation records, sales follow-up records, online interaction data, etc. Specifically, it can include any applicable information, and this embodiment of the invention does not impose any limitations on it.

[0106] In this embodiment of the invention, user tags are identifiers extracted from user information to describe user characteristics and behavioral tendencies. They are high-level summaries of user data, used to quickly classify and identify user types and needs. For example, user tags could be "young user," "high-budget user," "prefers SUV models," etc. When determining user tags, appropriate tags can be selected from the aforementioned pre-determined intent tags as user tags.

[0107] In this embodiment of the invention, the target intent tag is compared with the user tags in the user information to identify users whose user tags match the target intent tag. Through this matching, the system can quickly locate potential users who meet specific sales lead requests from a large amount of user data, thereby achieving accurate user screening and recommendation.

[0108] For example, assuming the target intent tag is "enterprise user" or "company user", the system searches the user information database for all users tagged with "enterprise user" or "company user" and filters these users as target users, such as user A and user B, whose tags clearly include the user tag information such as "enterprise user" or "company user".

[0109] According to embodiments of the present invention, a sales lead request is first obtained. Then, based on pre-determined intent tags and the correlation information between intent tags, the target intent tag corresponding to the sales lead request is determined. Finally, the target intent tag is matched with user tags in user information to filter out target users who match the target intent tag. This solution can automatically understand the intent of sales leads. Based on the pre-prepared correlation information of intent tags, it accurately transforms sales lead requests into target intent tags, deeply understanding the complex semantics and deep needs in sales lead requests, effectively making up for the shortcomings of traditional methods in handling complex semantic relationships and understanding deep user needs. It not only quickly and accurately filters out the corresponding target users, but also greatly improves the efficiency and accuracy of sales lead mining, reduces manual processing costs and time, and improves the efficiency of sales lead processing.

[0110] In an optional embodiment of the present invention, the target intent tag includes a first target intent tag and a second target intent tag. Determining the target intent tag corresponding to the sales lead request based on the pre-determined intent tags and the association information between the intent tags includes: determining the corresponding first target intent tag from the pre-determined intent tags according to the sales lead request; and determining the second target intent tag according to the first target intent tag and the association information between the intent tags.

[0111] The primary target intent tag refers to the most direct and relevant tag found among pre-defined intent tags based on the sales lead request, through preliminary analysis and matching. It represents the system's initial understanding and positioning of the user's request intent.

[0112] The specific implementation methods for determining the corresponding first target intent tag from the pre-determined intent tags based on the sales lead request can include various methods, and the embodiments of the present invention do not limit this.

[0113] In one specific implementation, natural language processing techniques (such as synonym recognition and keyword extraction) are used to analyze sales lead requests and identify core vocabulary and semantic information. The identified key information is then compared with a pre-built intent tag library to find the tag that is semantically closest and best reflects the request's intent as the primary target intent tag. This may involve various techniques such as calculating text similarity and matching keywords. For example, advanced natural language processing models such as the ERNIE model can be used to implement this process, as these models can better understand semantics and improve matching accuracy.

[0114] For example, a sales lead request can be input into a pre-trained intent recognition model to obtain the first target intent label. Alternatively, the sales lead request can be matched against a pre-defined knowledge base within the system, and through semantic analysis and logical reasoning, the intent label that most accurately reflects the user's needs—the first target intent label—can be identified.

[0115] Secondary target intent tags refer to other relevant intent tags identified by further utilizing the correlation information between intent tags, based on the primary target intent tag. These tags have a certain semantic or business logic relationship with the primary target intent tag, and can more comprehensively cover the potential user group related to the user's request intent.

[0116] The specific implementation of the second target intent tag can be determined based on the correlation information between the first target intent tag and the intent tags. This embodiment of the invention does not limit this.

[0117] By leveraging the relationships between intent tags, and using the first target intent tag as a foundation, other related intent tags are identified and used as second target intent tags. This expands the understanding of user needs, captures more potential intents and demands, and ensures a more comprehensive and accurate selection of target users.

[0118] In one specific implementation, a pre-constructed intent tag graph (such as a knowledge graph) is used, which records information such as the relationships and semantic similarities between intent tags. Starting with the first target intent tag, other tags that are related or semantically similar to it are searched in the intent tag graph. Based on certain rules and thresholds, these related tags are determined as the second target intent tags.

[0119] For example, in a knowledge graph, the tag "enterprise user" may be associated with tags such as "company user" and "enterprise procurement," and these tags can be identified as second target intent tags. Based on the determination that the first target intent tag is "enterprise user," the system further determines that the second target intent tags may include "company user," "enterprise procurement intention," etc., according to the association information between intent tags.

[0120] By subdividing target intent tags into first target intent tags and second target intent tags, first determining the first target intent tag based on the sales lead request, and then determining the second target intent tag based on the relationship between intent tags, this method can more comprehensively explore potential user intents related to the sales lead request, expand the breadth and depth of lead sources, avoid the problem of incomplete user group coverage caused by matching a single intent tag, further improve the accuracy and comprehensiveness of sales lead screening, and enhance the flexibility and adaptability of sales strategies.

[0121] In an optional embodiment of the present invention, before obtaining a sales lead request, the method may further include: identifying relevant data of the business scenario to obtain intent tags in the business scenario; calculating the semantic similarity between intent tags in the business scenario; and determining the association information between intent tags based on the intent tags and semantic similarity in the business scenario.

[0122] Business scenarios refer to various scenarios in the sales process, such as test drives, in-store consultations, and vehicle delivery in the context of car sales. User behavior and intentions will differ across these different business scenarios.

[0123] Relevant data refers to various data related to business scenarios, including users' voice records, text records (such as chat logs, questionnaires, etc.), and behavioral data (such as browsing history, test drive appointment records, etc.).

[0124] By analyzing and processing relevant data in specific business scenarios, tags that reflect user intent are extracted, providing a foundation for subsequent intent understanding and user filtering. Any applicable implementation method can be adopted, and this embodiment of the invention does not impose any limitations on this.

[0125] In one specific implementation, natural language processing techniques (such as Automatic Speech Recognition (ASR) to convert speech into text, text error correction to fix text errors, and Named Entity Recognition (NER) and Key Information Extraction (KIE) to extract specific entities and key information) are used to process text data in business scenarios and identify key information. Then, using intent summarization techniques from large models, user intents are organized and summarized, and relevant tags are extracted for the car sales scenario, achieving preliminary structured processing of the natural language data.

[0126] For example, from the voice recording of a test drive appointment, the text can be converted by ASR and then NER can be used to identify key information such as the car model and time mentioned by the user, thereby generating intent tags such as "test drive appointment" and "interested in car model XX".

[0127] Semantic similarity measures the degree of semantic similarity between two intent tags, that is, their similarity in meaning and expressed intent. For example, "test drive appointment" and "appoint test drive" have a high degree of semantic similarity.

[0128] By calculating the semantic similarity between intent tags, the degree of semantic association between them is quantified, providing a basis for subsequently determining the relationship between intent tags.

[0129] In one specific implementation, text vectorization techniques (such as embedding) are used to convert intent tags into computer-processable vector forms. Then, semantic similarity is obtained by calculating the similarity between vectors (such as cosine similarity). For example, converting "test drive appointment" and "appoint test drive" into vector forms and calculating their cosine similarity yields a value close to 1, indicating that they are semantically very similar.

[0130] By comprehensively considering the intent tags themselves and their semantic similarity, a system of relationships between intent tags is established to enable a more comprehensive understanding and processing of user intent.

[0131] Based on the semantic similarity of intent tags, similar tags are grouped together, and similarity associations are established between them. Simultaneously, by combining business logic and knowledge graph technology, other relationships between intent tags are determined, such as hierarchical relationships and causal relationships. For example, in a knowledge graph, "test drive satisfaction" may be associated with tags such as "test drive experience" and "test drive intention," where "test drive satisfaction" might be a result of "test drive experience."

[0132] For example, in the context of car sales, it can be determined that there is a sequential relationship in the business process between "test drive appointment" and "test drive at the store", a causal relationship between "test drive satisfaction" and "purchase intention", and a semantic similarity relationship between "test drive appointment" and "appointment for test drive". These relationships together constitute the association information between intent tags.

[0133] Before acquiring sales lead requests, relevant data from the business scenario is identified to obtain intent tags, and the semantic similarity between these tags is calculated to determine their correlation information. This series of operations lays a solid foundation for more accurate understanding and identification of the intent in sales lead requests, enabling the system to handle complex and diverse sales leads more intelligently, improving the intelligence level and accuracy of intent identification in the entire sales lead processing process, reducing misjudgments of intent due to insufficient data understanding, and improving the quality and efficiency of sales lead processing.

[0134] In an optional embodiment of the present invention, the intent tag in the business scenario includes key information and related tags. Identifying the relevant data of the business scenario to obtain the intent tag in the business scenario includes: extracting key information from the relevant data of the business scenario; and generating corresponding related tags using a first content generation model based on the relevant data of the business scenario.

[0135] Key information refers to entities and keywords in relevant data. Entities refer to objectively existing things with independent meaning, such as specific car models, times, and locations mentioned by users. Keywords refer to important information that can reflect user behavior, preferences, and intentions, such as vehicle configurations that users are interested in and their intention to test drive.

[0136] We filter and refine large amounts of data from business scenarios to identify the most valuable and representative information, providing a foundation for subsequent tag generation.

[0137] In one specific implementation, natural language processing (NLP) technology is used to process text data in a business scenario and identify key information. For example, from a user's test drive appointment phone call record, ASR (Automatic Speech Recognition) is used to convert it into text, and then NER (Natural Language Errata) is used to identify key information such as the car model and time mentioned by the user.

[0138] For example, when processing a user's test drive appointment call record, the system uses ASR to convert the speech into text: "I would like to book a test drive of the Li Auto L7 next Tuesday," and then uses NER to identify the key information "next Tuesday" and "Li Auto L7."

[0139] Relevant tags refer to labels generated based on relevant data from business scenarios, which are related to user behavior and intent, and are used to more comprehensively describe user characteristics and behaviors. These tags can be direct reflections of key information, or they can be the result of semantic understanding and summarization.

[0140] The first content generation model refers to a certain type of large existing model that has powerful natural language understanding and generation capabilities and can generate corresponding labels based on the input data.

[0141] By leveraging existing large models, the extracted key information is transformed into structured labels, enabling automated annotation of user behavior and intent.

[0142] Relevant data from the business scenario is input into the first content generation model. Based on the input data, combined with existing knowledge and semantic understanding capabilities, the model generates tags related to the business scenario.

[0143] For example, given the relevant data "users are concerned about battery life and price", the first content generation model can generate relevant tags "concerned about battery life" and "concerned about price".

[0144] This method extracts key information from relevant data in business scenarios and uses a first content generation model to generate corresponding relevant tags. It can automatically and efficiently mine valuable information from a large amount of business data and transform it into a structured tag form. This not only enriches the intent tag system but also ensures the accuracy and timeliness of the tags. It provides a more comprehensive and accurate basis for subsequent intent recognition and user screening, improves the system's ability to understand business scenarios and the degree of automation in sales lead processing, and reduces manual intervention and data processing costs.

[0145] In an optional embodiment of the present invention, the association information includes the association relationship and semantic similarity between intent tags. Determining the association relationship information between intent tags based on the intent tags and semantic similarity in the business scenario includes: constructing intent tag graph data based on the intent tags and semantic similarity in the business scenario; wherein, the intent tag graph data includes the association relationship and semantic similarity between intent tags.

[0146] Intent tag graph data is a structured data format used to present information such as the relationships and semantic similarity between intent tags. It uses intent tags as nodes and the relationships and semantic similarity between them as edges, forming a graph structure.

[0147] By integrating and structuring intent tags, their relationships, and semantic similarity, an intent tag graph is constructed to more intuitively display and utilize this information, supporting subsequent intent recognition and expansion. For example, ... Figure 2 The diagram shows a multi-level intent labeling schema and, as shown, a multi-level intent labeling schema. Figure 3 The diagram shown is a schematic representation of the extended intent label. Figure 2 and Figure 3 This shows a portion of the intent tags in the intent tag knowledge graph and their interrelationships.

[0148] Collect all intent tags from various business scenarios and calculate their semantic similarity. Using a graph database or similar storage structure, construct an intent tag graph data, treating intent tags as nodes and their relationships and semantic similarity as edges. For example, use a graph database to store and manage the intent tag graph data.

[0149] For example, in a car sales scenario, the constructed intent tag graph data might include the following: Nodes: Intent tags such as "Test drive satisfied," "Test drive dissatisfied," "Test drive appointment successful," etc. Edges: Represent the relationships and semantic similarity between intent tags. For example, there is a semantically similar edge between "Test drive satisfied" and "Test drive experience good," with a similarity of 0.8.

[0150] By constructing an intent tag graph, the relationships and semantic similarities between intent tags are structured, providing rich semantic and relational information for intent recognition in sales lead requests. After constructing the intent tag graph based on intent tags and semantic similarity within a business scenario, the system can more intelligently understand and identify complex intents in sales lead requests. When processing sales lead requests, the system can leverage the relationships and semantic similarities in the graph data to quickly find the tags that best match the request intent, improving the accuracy and efficiency of intent recognition. This structured tag graph not only enhances the system's adaptability to diverse and complex sales leads but also provides a solid foundation for subsequent user screening and marketing strategy development, thereby improving the overall quality and effectiveness of sales lead processing.

[0151] In an optional embodiment of the present invention, the intent tag map data further includes: multi-level classification information of intent tags, wherein the multi-level classification information represents the category affiliation relationship between intent tags; before determining the corresponding first target intent tag from the predetermined intent tags according to the sales lead request, it may further include: constructing training data based on the sales lead request sample and the multi-level classification information of intent tags; wherein the training data includes the sales lead request sample and at least one level of intent tags corresponding to the tag; and training a multi-level intent recognition model using the training data. Accordingly, determining the corresponding first target intent tag from the predetermined intent tags according to the sales lead request includes: using the trained multi-level intent recognition model to recognize the sales lead request to obtain the first target intent tag.

[0152] Multi-level classification information represents the category affiliation relationship between intent tags. That is, some intent tags may belong to higher-level categories, and some categories may belong to even higher-level categories, forming a hierarchical structure. For example, "test drive satisfied" and "test drive dissatisfied" may both belong to the category "test drive tag", and "test drive tag" may belong to the higher-level category "user feedback".

[0153] Sales lead request samples refer to examples of sales lead requests used to train the model. These samples include various forms of sales lead requests.

[0154] Training data refers to the dataset used to train a multi-level intent recognition model, which includes sales lead request samples and intent labels of at least one level corresponding to the tags.

[0155] By collecting sales lead request samples and labeling them with multi-level classification information of intent tags, a dataset suitable for training a multi-level intent recognition model is constructed, providing a foundation for model training.

[0156] Collect a large number of sales lead request samples, which can come from real-world business scenarios or be manually constructed. For each sample, label it with at least one level of intent tag based on its content and intent. For example, for a request "find some users who were satisfied after a test drive," label it with a first-level category "test drive tag," a second-level category "test drive effect," and a third-level category "test drive satisfaction." Combine these labeled samples and their corresponding intent tags to form a training dataset.

[0157] A multi-level intent recognition model is a model capable of identifying multiple levels of intent in sales lead requests. It can output intent labels for one or more levels corresponding to the request, thereby gaining a more detailed understanding of the user's intent.

[0158] Using the constructed training data, a model can be trained through machine learning or deep learning algorithms to accurately identify multi-level intents in sales lead requests.

[0159] Specifically, select a suitable machine learning or deep learning algorithm, such as a Transformer-based model (e.g., ERNIE). Input training data into the algorithm to train the model. During training, the model learns how to map sales lead requests to corresponding multi-level intent labels. Evaluate and optimize the trained model to ensure it can accurately identify intents in real-world applications.

[0160] For example, a large amount of collected training data, including various sales lead request samples and their corresponding multi-level intent labels, is input into the ERNIE model for training. After multiple rounds of training and adjustments, the model can learn how to accurately identify multi-level intents in sales lead requests. For example, for the request "find some users who are satisfied after test driving", the model can output the corresponding multi-level intent labels "test drive label - test drive effect - test drive satisfaction".

[0161] The user's sales lead request is input into a pre-trained multi-level intent recognition model, which then identifies the sales lead request. Leveraging the model's semantic understanding and intent recognition capabilities, the first target intent label that best matches the request is obtained.

[0162] The user's sales lead request is taken as input and fed into a multi-level intent recognition model. The model performs semantic analysis and intent recognition on the request, and outputs the corresponding first-target intent label. For example, for the request "find some users who are satisfied after test driving", the model outputs the first-target intent label "test drive satisfied".

[0163] By constructing an intent tag graph containing multi-level classification information of intent tags, and building training data based on sales lead request samples and multi-level classification information, the trained multi-level intent recognition model can more meticulously and accurately identify multi-level intents in sales lead requests. When determining the first target intent tag based on a sales lead request, this model can effectively improve the depth and accuracy of intent recognition, helping to more comprehensively understand user needs and behaviors, providing strong support for developing personalized sales strategies, enhancing the sales team's ability to handle complex sales scenarios, and improving the conversion efficiency and closing rate of sales leads.

[0164] In an optional embodiment of the present invention, the intent tag graph data further includes: vector data corresponding to intent tags. Based on the sales lead request, a first target intent tag is determined from the predetermined intent tags, including: when the multi-level intent recognition model cannot recognize the sales lead request, using the vector data corresponding to the sales lead request generated by the multi-level intent recognition model to perform similarity calculation with the vector data corresponding to the intent tag in the intent tag graph data to obtain a similarity result; and determining the first target intent tag based on the similarity result.

[0165] When the intent of a sales lead request is too complex, ambiguous, or differs significantly from the training data, the multi-level intent recognition model cannot accurately identify the corresponding intent label. For example, if the confidence level of the multi-level intent recognition model in recognizing a sales lead request is lower than a set threshold, it will be judged as unable to recognize it.

[0166] The vector data corresponding to a sales lead request refers to the data in vector form that is converted from a sales lead request through a multi-level intent recognition model. This vector can represent the semantic features of the request.

[0167] When the model cannot directly identify the intent, it uses the vector data generated by the model to indirectly identify the intent, and finds the closest intent label by calculating the similarity in the vector space.

[0168] The vector data corresponding to the intent tags in the intent tag graph data refers to the data in which each intent tag is converted into a vector form when constructing the intent tag graph data. These vector data are stored in the graph data and are used to represent the semantic features of each intent tag.

[0169] By calculating the similarity between the vector data corresponding to the sales lead request and the vector data of each intent tag, we can measure their semantic similarity and find the tag that is most similar to the request intent.

[0170] Specifically, the process involves retrieving the vector data corresponding to all intent tags in the intent tag graph. Then, a similarity calculation method (such as cosine similarity) is used to calculate the similarity between the sales lead request vector and each intent tag vector. This results in a similarity list containing the similarity values ​​between the sales lead request and each intent tag.

[0171] The similarity result refers to the list of similarity values ​​between the sales lead request and each intent tag, calculated using similarity metrics. Based on the similarity result, the intent tag that is semantically closest to the sales lead request is selected as the first target intent tag. Even when the model cannot directly recognize it, a reasonable intent tag can be found through vector similarity.

[0172] The similarity results are sorted, and one or more intent tags with the highest similarity are selected as candidates. Based on a set threshold or rule, the final first target intent tag is determined from the candidate tags. For example, the tag with the highest similarity and a similarity greater than a certain threshold is selected.

[0173] For example, the vector data generated from a sales lead request "find users who are very concerned about vehicle range and price" is compared with the vector data of each intent tag (such as "concerned about range," "concerned about price," "concerned about configuration," etc.) in the intent tag graph data. The results might show a similarity of 0.75 for "concerned about range" and 0.78 for "concerned about price," etc. In the similarity results, "concerned about price" has a similarity of 0.78, "concerned about range" has a similarity of 0.75, and other tags have lower similarities. Based on a set threshold of 0.7, and selecting the tag with the highest similarity, "concerned about price" is determined as the primary target intent tag.

[0174] When a multi-level intent recognition model fails to identify the intent, the first target intent label is determined by calculating the similarity between the vector data of the sales lead request and the vector data in the intent label graph. This provides an effective supplementary means for intent recognition in sales lead requests. When direct model recognition is limited, this method can accurately find the label that best matches the request intent, ensuring the robustness and stability of the system. By calculating vector similarity, not only is the accuracy of intent recognition improved, but the system can also flexibly respond to complex and ever-changing sales scenarios, avoiding the loss of sales opportunities due to model recognition failures, thereby improving the overall efficiency and reliability of sales lead processing.

[0175] In an optional embodiment of the present invention, matching the target intent tag with the user tag in the user information to filter out target users who match the target intent tag includes: matching the first target intent tag with the user tag in the user information to filter out target users who match the first target intent tag; if the number of target users who match the first target intent tag is less than the number required for the sales lead request, matching the second target intent tag with the user tag in the user information to filter out the missing number of target users who match the second target intent tag.

[0176] The semantically understood and transformed initial target intent tags are applied to actual user data. A matching operation identifies users who meet the requirements, thereby enabling the precise targeting of potential sales leads from the user group. For details, please refer to the description in the preceding embodiments; further elaboration is not provided here.

[0177] After the first step of matching, the total number of users who match the first target intent tag is obtained.

[0178] The number of leads requested refers to the number of potential users a user expects to acquire when submitting a lead request, or the default requested number. For example, a user might want to "find 100 users of enterprise users".

[0179] The difference between the number of sales lead requests required and the number of target users matching the first target intent tag is the shortage quantity.

[0180] If the number of target users matching the first target intent tag is insufficient, the second target intent tag is used to match user tags to find additional users matching these tags, in order to meet the number required for the sales lead request.

[0181] When the number of initial matching results is insufficient, supplementary matching is performed using extended second target intent tags to meet the user's needs for the number of potential users as much as possible, thereby improving the comprehensiveness and effectiveness of sales leads.

[0182] Specifically, the number of target users matching the first target intent tag is compared with the number required for the sales lead request to calculate the shortage quantity. The previously determined second target intent tag is matched with user tags in the user information to identify users who match these tags. From these users matching the second target intent tag, the shortage quantity is selected according to certain rules (such as similarity, user activity, etc.) and added to the target user list.

[0183] For example, suppose a sales lead request is "find 100 users who care about battery life and price." After the first matching step, only 50 users matching the first target intent tag (caring about battery life and price) are found. At this point, the system uses second target intent tags, such as "concerned about battery life" or "concerned about price," to match these user tags and find users who match these tags. Suppose 80 users matching the second target intent tags are found, the system might select 50 of them based on similarity rules and add them to the target user list, ultimately resulting in 100 target users.

[0184] By matching target intent tags with user tags in user information, target users matching the first target intent tag are first filtered out, ensuring the accuracy and relevance of user screening. When the number of target users is insufficient, the second target intent tag is further matched with user tags to filter out the missing number of target users matching the second target intent tag, effectively solving the problem of insufficient target users and improving the flexibility and comprehensiveness of sales lead mining. This phased, multi-level screening strategy not only ensures the quality of target users but also fully considers the quantity requirements of sales lead requests, thereby more effectively supporting the sales team's precision marketing and user expansion efforts, improving sales conversion rates and market competitiveness.

[0185] In an optional embodiment of the present invention, matching the second target intent tag with user tags in user information to filter out a shortage of target users that match the second target intent tag includes: determining the proportion of each second target intent tag based on the semantic similarity between the second target intent tag and the first target intent tag; matching the second target intent tag with user tags in user information to obtain candidate users that match each second target intent tag; and selecting from the candidate users corresponding to each second target intent tag according to the proportion of the number of users to obtain a shortage of target users.

[0186] The percentage of users refers to the proportion of each secondary target intent tag used to supplement the target user list. It is used to determine how many users to select from the users corresponding to each secondary target intent tag to supplement the target user list.

[0187] Based on the semantic similarity between the second target intent tag and the first target intent tag, the number of each second target intent tag is reasonably allocated when supplementing target users to ensure the relevance of the supplemented users to the original intent.

[0188] Calculate the semantic similarity between each second target intent tag and the first target intent tag. Determine the proportion of each second target intent tag based on the semantic similarity. Tags with higher similarity may have a larger proportion.

[0189] For example, suppose the first target intent label is "concerned about battery life and price," and the second target intent labels include "concerned about battery life" and "concerned about price," with similarities of 0.7 and 0.8 to the first target intent label, respectively. Based on the similarity, the proportion of "concerned about battery life" is determined to be 40%, and the proportion of "concerned about price" is determined to be 60%.

[0190] Candidate users are those who may meet the requirements, obtained by matching the second target intent tag with the user tags in the user information. These users need to be further screened to determine whether they should be added to the target user list.

[0191] By matching the second target intent tag with the user tag, all users who may match these tags are identified as candidate users for further screening.

[0192] For example, for the second target intent label "concern about battery life", the system searches the user information database for all users labeled "concern about battery life" and obtains a batch of candidate users, such as user C, user D, etc.

[0193] Users are selected from the candidate users of each second target intent tag according to their proportion of numbers, ensuring that the supplementary users not only meet the quantity requirements, but also have a reasonable relevance to the original intent.

[0194] Specifically, based on the proportion determined in Step 1, calculate the number of users to be selected for each second target intent tag. From the candidate users corresponding to each second target intent tag, select the appropriate number of users according to certain rules (such as random selection, selection after sorting by user activity, etc.). Add the selected users to the target user list until the required number of sales leads is met.

[0195] For example, suppose we need to add 50 users. Based on the proportion of users, 20 users need to be provided for "Focusing on Battery Life" and 30 users need to be provided for "Focusing on Price". The system selects 20 users from the candidate users for "Focusing on Battery Life" and 30 users from the candidate users for "Focusing on Price", and adds them to the target user list, ultimately resulting in 100 target users.

[0196] By determining the proportion of target users based on the semantic similarity between the second and first target intent tags, and then selecting the required number of target users from the matched candidate users according to this proportion, precise supplementation is achieved when the number of target users is insufficient. This method not only considers the relevance of each relevant intent tag to the original intent, ensuring the overall quality of the target user group and its match with the sales lead request, but also avoids the waste of resources and reduced sales efficiency caused by blindly supplementing users, further optimizing the effectiveness and accuracy of sales lead processing.

[0197] In an optional embodiment of the present invention, after matching the target intent tag with the user tag in the user information and filtering out the target users that match the target intent tag, the method further includes: receiving a query request for any target user; generating introduction information and communication strategy information of the target user using a second content generation model based on the user information of the target user; and displaying the introduction information and communication strategy information of the target user on the user details page.

[0198] A query request is a request from a user (such as a salesperson) who wants to obtain more detailed information about a target user.

[0199] For example, such as Figure 4 The diagram illustrates the question-and-answer process. Upon receiving a sales lead request, the system analyzes the question and understands the intent to obtain target intent tags. Then, a site-wide search is performed based on these tags to identify the target user. The user information of the target user is then processed, and finally, the processed content is output to complete the answer.

[0200] The system needs to be able to receive user requests for detailed information about a specific target user so that it can provide the corresponding user details in the future.

[0201] The system provides a user interface or API that allows users to input the identifier of the target user (such as a user ID) to initiate a query request. Figure 5 The illustration shown depicts a target user list page where the sales lead request is: "Find some users of enterprise users." AI (Artificial Intelligence) is used to identify and display the primary target intent tag, "Enterprise Users." The target users are then determined using the above method. A list of target users is displayed on the page, including user ID, tags, and the date of the last follow-up. A summary text can also be displayed at the bottom of the page. Clicking on any target user in the user list initiates a query request for that user.

[0202] For example, a salesperson enters the user ID "123456" into the system to initiate a query request.

[0203] The second content generation model refers to an existing large-scale model, which can be a generative model based on natural language processing. The second content generation model can be the same as the first content generation model, or it can be a different model.

[0204] Introductory information is a summary and description of the target user's basic information, behavioral characteristics, preferences, etc., such as a piece of information describing a user story.

[0205] Communication strategy information is information tailored to a user's characteristics and behavior, providing salespeople with the best ways and strategies to communicate with that user. For example, a message describing a communication strategy.

[0206] By leveraging advanced content generation models, user information can be transformed into user introductions and communication strategies that are useful to sales personnel, thereby improving the efficiency and effectiveness of sales communication.

[0207] The target user's information is used as input into the second content generation model. Based on the input information, the model generates corresponding introductory and communication strategy information.

[0208] For example, if you input user information such as "User ID: 123456, satisfied with the test drive, concerned about range", the model will generate the following information: "This user is satisfied with the test drive experience and is particularly concerned about the vehicle's range capability", and the following communication strategy information: "It is recommended to focus on introducing the vehicle's range advantages and ask the user about their specific range requirements".

[0209] The user details page is a page in the system used to display detailed information about a single user, allowing sales personnel to gain a comprehensive understanding of the user's situation.

[0210] The generated user introductions and communication strategy information are displayed to sales staff in the system interface, making it easy for them to view and use this information, thereby enabling them to communicate better with users.

[0211] By displaying introductory information and communication strategies for target users on their details pages, sales personnel can gain a comprehensive understanding of these users and the best communication strategies, thereby improving communication efficiency and success rates, and ultimately increasing sales conversion rates. This display feature also facilitates unified management and access to user information for sales personnel, reducing the time cost of information retrieval and further optimizing the sales workflow.

[0212] In an optional embodiment of the present invention, the method further includes: displaying user management statistics and intent behavior information of the target user on the user details page.

[0213] User management statistics refer to statistical information related to user management, such as the number of communications, average call duration, and number of test drives. Intent behavior information refers to information reflecting user purchase intent and behavioral patterns, such as key moments, time periods, and changes in intent behavior.

[0214] By integrating user management statistics and intent behavior information of target users into the user details page, sales personnel can gain more comprehensive user insights and better formulate sales strategies.

[0215] For example, such as Figure 6 The diagram shown illustrates a user details page. The page can display user stories (i.e., introductory information), current user status (i.e., user management statistics), follow-up points (i.e., intent and behavior information), communication tips (i.e., communication strategy information), and other content.

[0216] By displaying target users' user management statistics and intent behavior information on their user detail pages, this method provides sales personnel with a more comprehensive user perspective. This allows them to gain a deeper understanding of users' historical interactions and behavioral preferences, thereby more accurately grasping changes in users' purchasing intentions and needs, and providing data support for developing personalized sales strategies. Furthermore, this centralized display of information also helps improve the overall collaboration efficiency of the sales team, enabling different members to communicate and adjust strategies based on the same information platform.

[0217] For example, such as Figure 7 The diagram shows the framework of a sales lead request processing system. The system comprises four parts. The first part is Natural Language Processing (NLP), used to prepare for determining intent tags. This includes: ASR (Speech-to-Text), text correction, named entity recognition, key information extraction, large-scale model intent summarization, automotive sales scenario locking (test drive, in-store visits, etc.), and extraction of automotive business scenario tags such as model and series. The second part is data processing, used for data fusion, i.e., determining intent tags. This includes: One-ID (mapped to the same identifier) ​​data fusion, tag word frequency statistics, tag text vectorization, semantic similarity calculation (embedding), and other modules or algorithms to determine seed tags and extended semantic tags. The third part is a knowledge graph, used for knowledge construction, i.e., determining relational information. This includes: a One-ID tag knowledge graph, a tag semantic similarity knowledge graph, a tag text vector database, a multi-level intent knowledge base, and a business data knowledge base. The fourth part is intent parsing, used for intent recognition, i.e., determining the target intent label. Specifically, this includes: multi-level intents in automotive business scenarios, training of multi-intent recognition models, query (sales lead request) rewriting, intent recognition and ranking under known intents, and query and label vector similarity retrieval under unknown intents. The sales lead request processing system can execute the above sales lead processing methods.

[0218] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0219] Reference Figure 8 The diagram illustrates a structural block diagram of a sales lead processing device according to another embodiment of the present invention, which may specifically include the following modules:

[0220] Request Module 201 is used to retrieve sales lead requests;

[0221] The tag determination module 202 is used to determine the target intent tag corresponding to the sales lead request based on the pre-determined intent tags and the correlation information between the intent tags;

[0222] The user filtering module 203 is used to match the target intent tag with the user tags in the user information and filter out the target users that match the target intent tag.

[0223] Optionally, the target intent label includes a first target intent label and a second target intent label, and the label determination module includes:

[0224] The first tag determination submodule is used to determine the corresponding first target intent tag from the pre-determined intent tags based on the sales lead request;

[0225] The second tag determination submodule is used to determine the second target intent tag based on the correlation information between the first target intent tag and the intent tag.

[0226] Optionally, the device further includes:

[0227] The identification module is used to identify relevant data of the business scenario before the request to obtain sales leads, and to obtain intent tags under the business scenario;

[0228] The similarity calculation module is used to calculate the semantic similarity between intent tags in the business scenario.

[0229] The association determination module is used to determine the association information between the intent tags based on the intent tags in the business scenario and the semantic similarity.

[0230] Optionally, the intent label in the business scenario includes key information and related labels, and the recognition module includes:

[0231] An extraction submodule is used to extract key information from relevant data in the business scenario.

[0232] The tag generation submodule is used to generate corresponding relevant tags based on the relevant data of the business scenario using the first content generation model.

[0233] Optionally, the association information includes the association relationships and semantic similarity between the intent tags, and the association determination module includes:

[0234] The graph construction submodule is used to construct intent tag graph data based on the intent tags in the business scenario and the semantic similarity; wherein, the intent tag graph data includes the correlation and semantic similarity between the intent tags.

[0235] Optionally, the intent tag map data further includes: multi-level classification information of the intent tags, wherein the multi-level classification information represents the category affiliation relationship between the intent tags; the device further includes:

[0236] The training data construction module is used to construct training data based on the sales lead request samples and the multi-level classification information of the intent tags before determining the corresponding first target intent tag from the pre-determined intent tags according to the sales lead request; wherein, the training data includes the sales lead request samples and at least one level of intent tags corresponding to the tags;

[0237] The training module is used to train a multi-level intent recognition model using the training data;

[0238] The first determining submodule includes:

[0239] The identification unit is used to identify the sales lead request using a trained multi-level intent recognition model to obtain the first target intent label.

[0240] Optionally, the intent label graph data further includes: vector data corresponding to the intent label, and the first determining submodule includes:

[0241] The similarity calculation unit is used to calculate the similarity between the vector data corresponding to the sales lead request generated by the multi-level intent recognition model and the vector data corresponding to the intent tag in the intent tag graph data when the multi-level intent recognition model cannot recognize the sales lead request, and to obtain the similarity result.

[0242] The first determining unit is used to determine the first target intent tag based on the similarity result.

[0243] Optionally, the user filtering module includes:

[0244] The matching submodule is used to match the first target intent tag with the user tags in the user information and filter out target users that match the first target intent tag.

[0245] The filtering submodule is used to match the second target intent tag with the user tags in the user information when the number of target users matching the first target intent tag is less than the number required by the sales lead request, and to filter out the missing number of target users matching the second target intent tag.

[0246] Optionally, the filtering submodule includes:

[0247] The proportion determination unit is used to determine the proportion of each second target intent tag based on the semantic similarity between the second target intent tag and the first target intent tag;

[0248] A matching unit is used to match the second target intent tag with the user tags in the user information to obtain candidate users that match each of the second target intent tags;

[0249] The selection unit is used to select from the candidate users corresponding to each of the second target intent tags according to the quantity ratio, so as to obtain the target users with the shortage quantity.

[0250] Optionally, the device further includes:

[0251] The request receiving module is configured to receive a query request for any of the target users after the target intent tag is matched with the user tags in the user information and the target users that match the target intent tag are filtered out.

[0252] The information generation module is used to generate introduction information and communication strategy information for the target user based on the user information of the target user using a second content generation model;

[0253] The display module is used to display the target user's introduction information and the communication strategy information on the user details page.

[0254] According to an embodiment of the present invention, a sales lead request is first obtained. Then, based on pre-determined intent tags and the correlation information between the intent tags, a target intent tag corresponding to the sales lead request is determined. Finally, the target intent tag is matched with user tags in user information to filter out target users who match the target intent tag. This solution can automatically understand the intent of sales leads. Based on the pre-prepared correlation information of intent tags, it accurately transforms sales lead requests into target intent tags, deeply understands the complex semantics and deep needs in sales lead requests, and effectively makes up for the shortcomings of traditional methods in handling complex semantic relationships and understanding deep user needs. It not only quickly and accurately filters out the corresponding target users, but also greatly improves the efficiency and accuracy of sales lead mining, reduces manual processing costs and time, and improves the efficiency of sales lead processing.

[0255] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0256] Figure 9 This is a structural block diagram illustrating an electronic device 700 for sales lead processing according to an exemplary embodiment. For example, the electronic device 700 may be a vehicle-mounted computer, a computer, a digital broadcasting terminal, a messaging device, a game console, a medical device, a fitness device, etc.

[0257] Reference Figure 9 The electronic device 700 may include one or more of the following components: a processing component 702, a memory 704, a power supply component 706, a multimedia component 708, an audio component 710, an input / output (I / O) interface 712, a sensor component 714, and a communication component 716.

[0258] Processing component 702 typically controls the overall operation of electronic device 700, such as operations associated with display, telephone calls, data communication, camera operation, and recording operations. Processing component 702 may include one or more processors 720 to execute instructions to complete all or part of the steps of the methods described above. Furthermore, processing component 702 may include one or more modules to facilitate interaction between processing component 702 and other components. For example, processing component 702 may include a multimedia module to facilitate interaction between multimedia component 708 and processing component 702.

[0259] Memory 704 is configured to store various types of data to support the operation of device 700. Examples of this data include instructions for any application or method operating on electronic device 700, contact data, phonebook data, messages, pictures, videos, etc. Memory 704 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0260] Power supply component 706 provides power to various components of electronic device 700. Power supply component 706 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 700.

[0261] Multimedia component 708 includes a screen that provides an output interface between the electronic device 700 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 708 includes a front-facing camera and / or a rear-facing camera. When the electronic device 700 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0262] Audio component 710 is configured to output and / or input audio signals. For example, audio component 710 includes a microphone (MIC) configured to receive external audio signals when electronic device 700 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 704 or transmitted via communication component 716. In some embodiments, audio component 710 also includes a speaker for outputting audio signals.

[0263] I / O interface 712 provides an interface between processing component 702 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0264] Sensor assembly 714 includes one or more sensors for providing state assessments of various aspects of electronic device 700. For example, sensor assembly 714 may detect the on / off state of device 700, the relative positioning of components such as the display and keypad of electronic device 700, changes in position of electronic device 700 or a component of electronic device 700, the presence or absence of user contact with electronic device 700, orientation or acceleration / deceleration of electronic device 700, and temperature changes of electronic device 700. Sensor assembly 714 may include a proximity sensor configured to detect the presence of nearby entities without any physical contact. Sensor assembly 714 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 714 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0265] Communication component 716 is configured to facilitate wired or wireless communication between electronic device 700 and other devices. Electronic device 700 can access wireless networks based on communication standards, such as WiFi, 2G, or 3G, or combinations thereof. In one exemplary embodiment, communication component 716 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 716 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0266] In an exemplary embodiment, the electronic device 700 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0267] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 704 including instructions, which can be executed by a processor 720 of an electronic device 700 to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0268] A non-transitory computer-readable storage medium, when instructions in the storage medium are executed by a terminal's processor, enables the terminal to perform a sales lead processing method, the method comprising:

[0269] Request sales leads;

[0270] Based on the pre-determined intent tags and the correlation information between the intent tags, the target intent tag corresponding to the sales lead request is determined;

[0271] The target intent tags are matched with user tags in the user information to filter out target users who match the target intent tags.

[0272] Optionally, the target intent tag includes a first target intent tag and a second target intent tag. Determining the target intent tag corresponding to the sales lead request based on pre-determined intent tags and the correlation information between the intent tags includes:

[0273] Based on the sales lead request, determine the corresponding first target intent tag from the pre-determined intent tags;

[0274] The second target intent tag is determined based on the correlation information between the first target intent tag and the intent tags.

[0275] Optionally, prior to the request for obtaining sales leads, the method further includes:

[0276] Identify relevant data from the business scenario to obtain intent tags for that business scenario;

[0277] Calculate the semantic similarity between intent tags in the aforementioned business scenario;

[0278] Based on the intent tags in the business scenario and the semantic similarity, the association information between the intent tags is determined.

[0279] Optionally, the intent tags in the business scenario include key information and related tags. The process of identifying relevant data in the business scenario to obtain the intent tags in the business scenario includes:

[0280] Extract key information from relevant data in the aforementioned business scenario;

[0281] Based on the relevant data of the business scenario, the corresponding relevant tags are generated using the first content generation model.

[0282] Optionally, the association information includes the association relationships and semantic similarity between the intent tags, and determining the association relationship information between the intent tags based on the intent tags in the business scenario and the semantic similarity includes:

[0283] Based on the intent tags in the business scenario and the semantic similarity, an intent tag graph data is constructed; wherein, the intent tag graph data includes the relationships and semantic similarity between the intent tags.

[0284] Optionally, the intent tag map data further includes: multi-level classification information of the intent tags, wherein the multi-level classification information represents the category affiliation relationship between the intent tags; before determining the corresponding first target intent tag from the predetermined intent tags based on the sales lead request, the method further includes:

[0285] Training data is constructed based on the multi-level classification information of the sales lead request samples and the intent tags; wherein, the training data includes the sales lead request samples and at least one level of intent tags corresponding to the tags;

[0286] Using the training data, a multi-level intent recognition model is trained;

[0287] The step of determining the corresponding first target intent tag from the pre-determined intent tags based on the sales lead request includes:

[0288] The sales lead request is identified using a trained multi-level intent recognition model to obtain the first target intent label.

[0289] Optionally, the intent tag graph data further includes: vector data corresponding to the intent tag, and the step of determining the corresponding first target intent tag from the predetermined intent tags based on the sales lead request includes:

[0290] When the multi-level intent recognition model fails to recognize the sales lead request, the vector data corresponding to the sales lead request generated by the multi-level intent recognition model is used to perform similarity calculation with the vector data corresponding to the intent tag in the intent tag graph data to obtain a similarity result.

[0291] Based on the similarity results, the first target intent tag is determined.

[0292] Optionally, the step of matching the target intent tag with user tags in user information to filter out target users who match the target intent tag includes:

[0293] The first target intent tag is matched with the user tags in the user information to filter out target users who match the first target intent tag;

[0294] If the number of target users matching the first target intent tag is less than the number required by the sales lead request, the second target intent tag is matched with the user tags in the user information to filter out the missing number of target users matching the second target intent tag.

[0295] Optionally, the step of matching the second target intent tag with the user tags in the user information to filter out a shortage of target users matching the second target intent tag includes:

[0296] The proportion of each second target intent tag is determined based on the semantic similarity between the second target intent tag and the first target intent tag;

[0297] The second target intent tag is matched with the user tags in the user information to obtain candidate users that match each of the second target intent tags;

[0298] Based on the stated proportion, select from the candidate users corresponding to each of the second target intent tags to obtain the number of target users in short supply.

[0299] Optionally, after matching the target intent tag with user tags in the user information to filter out target users who match the target intent tag, the method further includes:

[0300] Receive query requests for any of the target users;

[0301] Based on the user information of the target user, the second content generation model is used to generate the introduction information and communication strategy information of the target user;

[0302] The user details page displays the target user's introduction information and the communication strategy information.

[0303] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0304] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0305] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0306] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0307] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0308] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0309] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0310] The above provides a detailed description of a sales lead processing method, a sales lead processing device, an electronic device, and a readable storage medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A sales lead processing method, characterized in that, The method includes: Request sales leads; Based on the pre-determined intent tags and the correlation information between the intent tags, the target intent tag corresponding to the sales lead request is determined; The target intent tags are matched with user tags in the user information to filter out target users who match the target intent tags.

2. The method according to claim 1, characterized in that, The target intent tag includes a first target intent tag and a second target intent tag. Determining the target intent tag corresponding to the sales lead request based on pre-determined intent tags and the correlation information between the intent tags includes: Based on the sales lead request, determine the corresponding first target intent tag from the pre-determined intent tags; The second target intent tag is determined based on the correlation information between the first target intent tag and the intent tags.

3. The method according to claim 1 or 2, characterized in that, Prior to the request for obtaining sales leads, the method further includes: Identify relevant data from the business scenario to obtain intent tags for that business scenario; Calculate the semantic similarity between intent tags in the aforementioned business scenario; Based on the intent tags in the business scenario and the semantic similarity, the association information between the intent tags is determined.

4. The method according to claim 3, characterized in that, The intent tags in the business scenario include key information and related tags. The process of identifying relevant data in the business scenario to obtain the intent tags in the business scenario includes: Extract key information from relevant data in the aforementioned business scenario; Based on the relevant data of the business scenario, the corresponding relevant tags are generated using the first content generation model.

5. The method according to claim 3 or 4, characterized in that, The association information includes the association relationships and semantic similarity between the intent tags. Determining the association relationship information between the intent tags based on the intent tags in the business scenario and the semantic similarity includes: Based on the intent tags in the business scenario and the semantic similarity, an intent tag graph data is constructed; wherein, the intent tag graph data includes the relationships and semantic similarity between the intent tags.

6. The method according to claim 5, characterized in that, The intent tag map data further includes: multi-level classification information of the intent tags, wherein the multi-level classification information represents the category affiliation relationship between the intent tags; before determining the corresponding first target intent tag from the pre-determined intent tags based on the sales lead request, the method further includes: Training data is constructed based on the multi-level classification information of the sales lead request samples and the intent tags; wherein, the training data includes the sales lead request samples and at least one level of intent tags corresponding to the tags; Using the training data, a multi-level intent recognition model is trained; The step of determining the corresponding first target intent tag from the pre-determined intent tags based on the sales lead request includes: The sales lead request is identified using a trained multi-level intent recognition model to obtain the first target intent label.

7. The method according to claim 6, characterized in that, The intent tag graph data further includes: vector data corresponding to the intent tag; the step of determining the corresponding first target intent tag from the pre-determined intent tags based on the sales lead request includes: When the multi-level intent recognition model fails to recognize the sales lead request, the vector data corresponding to the sales lead request generated by the multi-level intent recognition model is used to perform similarity calculation with the vector data corresponding to the intent tag in the intent tag graph data to obtain a similarity result. Based on the similarity results, the first target intent tag is determined.

8. The method according to any one of claims 2-7, characterized in that, The step of matching the target intent tag with user tags in user information to filter out target users who match the target intent tag includes: The first target intent tag is matched with the user tags in the user information to filter out target users who match the first target intent tag; If the number of target users matching the first target intent tag is less than the number required by the sales lead request, the second target intent tag is matched with the user tags in the user information to filter out the missing number of target users matching the second target intent tag.

9. The method according to claim 8, characterized in that, The step of matching the second target intent tag with the user tags in the user information to filter out a shortage of target users who match the second target intent tag includes: The proportion of each second target intent tag is determined based on the semantic similarity between the second target intent tag and the first target intent tag; The second target intent tag is matched with the user tags in the user information to obtain candidate users that match each of the second target intent tags; Based on the stated proportion, select from the candidate users corresponding to each of the second target intent tags to obtain the number of target users in short supply.

10. The method according to any one of claims 1-9, characterized in that, After matching the target intent tag with user tags in the user information to filter out target users who match the target intent tag, the method further includes: Receive query requests for any of the target users; Based on the user information of the target user, the second content generation model is used to generate the introduction information and communication strategy information of the target user; The user details page displays the target user's introduction information and the communication strategy information.

11. A sales lead processing device, characterized in that, The device includes: The request retrieval module is used to retrieve sales lead requests. The tag determination module is used to determine the target intent tag corresponding to the sales lead request based on the pre-determined intent tags and the correlation information between the intent tags; The user filtering module is used to match the target intent tags with user tags in the user information to filter out target users who match the target intent tags.

12. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the method described in any one of claims 1-10.

13. A readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is able to perform the method as described in any one of claims 1-10.

14. A computer program product, characterized in that, When the computer program product is run on an electronic device, it causes the processor to execute the steps of the method according to any one of claims 1-10.

Images