Generation object-based customer source clue recommendation method, system and device, and medium

Abstract

The invention provides a generation object-based customer source clue recommendation method, system and device and a medium, and the method comprises the steps: preprocessing a main business of a user side, extracting downstream features of the main business to generate a basic directory of a recommended object, and updating a target list of a recommended object side according to the basic directory; performing multi-path recall on the recommended objects in the target list according to the user portrait, filtering and sorting recall results, and outputting a recall set and a candidate recall set; and screening list generation tasks of the candidate recall sets according to a recall set setting recommendation strategy, and returning the list generation tasks to a recommendation object side to update the target list.

Description

Technical Field

[0001] The present invention relates to the field of intelligent sales technology, and in particular to a method, system, device and medium for recommending customer source leads based on a generated object. Background Art

[0002] With the rapid development of information technology and artificial intelligence, recommendation systems have become a key tool to help users discover customer leads. In actual applications, facing segmented users across the industry, different industries focus on different customer leads based on their business attributes. The basic data of industry customer leads requires industry experts to sort out the business attributes. Too much redundant basic data requires a lot of manpower and time costs. Therefore, obtaining customer leads that are suitable for users is the first problem that the recommendation system needs to solve.

[0003] Secondly, the existing recommendation system mainly solves the adaptation relationship between the user side and the recommendation object side. When facing different industry scenarios, although there is a huge amount of basic source data, the user side has not established an effective connection relationship with the recommendation object side. For example, the recommendation object is defined in the scenario as a small customer group with common characteristics, while the basic source data is a huge database of enterprise collections and / or store collections, and there is no concept of customer groups. Therefore, the data that can be effectively used by the recommendation object side in the initialization stage is almost an empty set. Furthermore, since the industries covered are quite broad, it is also necessary to determine the specific type of recommendation object side that the user side needs to adapt. Therefore, the current recommendation system has the above-mentioned defects and cannot meet the user side's needs to obtain satisfactory recommendation objects. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention proposes a method, system, device and medium for recommending customer source clues based on a generated object.

[0005] The first aspect of the present invention discloses a customer source lead recommendation method based on a generated object, comprising:

[0006] Preprocess the main business of the user side, extract the downstream features of the main business to generate a basic list of recommended objects, and update the target list of recommended objects based on the basic list;

[0007] Perform multi-way recall of recommended objects in the target list based on user portraits, filter and sort the recall results, and then output the recall set and candidate recall set;

[0008] According to the recall set setting recommendation strategy, the candidate recall set list generation task is screened, and the list generation task is returned to the recommendation object side to update the target list.

[0009] In an optional embodiment, the pre-processing of the main business of the user side includes:

[0010] According to the main business information filled in by the user, the system performs cleaning processing such as deduplication, processing missing values, processing abnormal values, unifying the format and data verification, as well as data integration, data analysis and mining, and purification processing to extract key information to obtain the main business terms.

[0011] In an optional embodiment, the step of extracting downstream features of the main business to generate a basic list of recommended objects includes:

[0012] A large language model is used to build a downstream feature judger, and the downstream feature values ​​generated based on the main business words are set to Boolean type to judge the association between the main business and downstream features;

[0013] The association relationship of downstream features determines the branch path of the basic directory to be generated, wherein the first branch path sets type rules and word expansion rules for downstream features of known recommended object types to generate a basic directory, and the second branch path uses a large language model to generate titles and query conditions for downstream features of unknown recommended object types to generate a basic directory.

[0014] In an optional embodiment, the updating of the target list of the recommended object side according to the basic directory includes:

[0015] Obtain the user's regional preferences and add them to the basic directory. Obtain the entity content of the recommended object according to the branch path of the basic directory to generate a target list.

[0016] In an optional embodiment, the multi-channel recall of recommended objects in the target list according to the user portrait includes:

[0017] Recall the recommended objects in the target list according to the user preferences in the basic directory;

[0018] Set matching rules based on the user's historical input words and recall recommended objects that match the input words;

[0019] Recall similar recommended objects based on the historical recommended objects clicked by the user;

[0020] Recall similar recommended objects based on the historical recommended objects clicked by users with similar portraits.

[0021] In an optional embodiment, the recall results are filtered and sorted to output the recall set and the candidate recall set.

[0022] Obtain the user's historical preferences and historical consumption, and filter the recall results based on historical data;

[0023] The ranking model is constructed to calculate the ranking score of each recommended object by weighted average of the filtered recall results according to the user's main business association value, regional preference value, object type matching value, title similarity value and query condition similarity value, and output the recall set and candidate recall set of the topk recommended objects.

[0024] In an optional embodiment, the list generation task of screening the candidate recall set according to the recall set recommendation strategy is returned to the recommendation object side to update the target list.

[0025] Compare the preference matching degree between the user portrait and the recommended object according to the recall set and the candidate recall set, determine the extensible target recommended object in the candidate recall according to the preference matching degree, set the target recommended object as the target downstream feature of the list to be generated, organize the target downstream features into the list generation task, return the basic list generation step to expand the target list library.

[0026] The second aspect of the present invention discloses a customer source clue recommendation system based on a generated object, the system comprising:

[0027] The recommended object generation module is used to pre-process the main business of the user side, extract the downstream features of the main business to generate a basic list of recommended objects, and update the target list of the recommended object side according to the basic list;

[0028] The recommended object recall module is used to perform multi-way recall of recommended objects in the target list according to user portraits, filter and sort the recall results, and then output the recall set and candidate recall set;

[0029] The recommendation object linkage module is used to set the recommendation strategy for screening the candidate recall set list generation task, and return the list generation task to the recommendation object side to update the target list.

[0030] In an optional embodiment, the recommended object generation module includes:

[0031] The data preprocessing module is used to remove duplicates, process missing values, process abnormal values, unify formats and perform data verification based on the main business information filled in by the user, as well as perform data integration, data analysis and mining on the main business information and purify the extraction of key information to obtain the main business terms.

[0032] In an optional embodiment, the recommended object generation module includes:

[0033] A downstream feature processing module is used to construct a downstream feature judgement device using a large language model, and set the downstream feature value generated according to the main business word to a Boolean type to judge the association between the main business and the downstream feature;

[0034] The basic directory processing module is used to determine the branch path of the basic directory to be generated based on the association relationship of downstream features, wherein the first branch path sets type rules and word expansion rules for downstream features of known recommended object types to generate a basic directory, and the second branch path uses a large language model to generate titles and query conditions for downstream features of unknown recommended object types to generate a basic directory.

[0035] In an optional embodiment, the recommended object generation module includes:

[0036] The entity data acquisition module is used to obtain the user's regional preferences and add them to the basic directory, obtain the entity content of the recommended object according to the branch path of the basic directory, and generate a target list.

[0037] In an optional embodiment, the recommended object recall module includes:

[0038] The recall module is used to recall the recommended objects in the target list according to the user preferences in the basic directory; to recall the recommended objects that match the input words according to the matching rules set according to the user's historical input words; to recall similar recommended objects based on the historical recommended objects clicked by the user; and to recall similar recommended objects based on the historical recommended objects clicked by users with similar portraits.

[0039] In an optional embodiment, the recommended object recall module includes:

[0040] The filtering module is used to obtain the user's historical preferences and historical consumption, and filter the recall results based on historical data;

[0041] The sorting module is used to build a sorting model to calculate the ranking score of each recommended object by weighted average of the filtered recall results according to the user's main business association value, regional preference value, object type matching value, title similarity value and query condition similarity value, and output the recall set and candidate recall set of the topk recommended objects.

[0042] In an optional embodiment, the recommended object linkage module includes:

[0043] The target list extension module is used to compare the preference matching between the user portrait and the recommended object according to the recall set and the candidate recall set, determine the extensible target recommendation object in the candidate recall according to the preference matching, set the target recommendation object as the target downstream feature of the list to be generated, organize the target downstream features into the list generation task and return the basic list generation step to expand the target list library.

[0044] The third aspect of the present invention discloses a customer source clue recommendation device based on a generated object, comprising:

[0045] at least one processor, and

[0046] a memory communicatively connected to the at least one processor; wherein,

[0047] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the customer source lead recommendation method based on generating objects as disclosed in any one of the first aspects of the present invention.

[0048] A fourth aspect of the present invention discloses a computer-readable storage medium storing computer-executable instructions for causing a computer to execute a customer source lead recommendation method based on a generated object as disclosed in any one of the first aspect of the present invention.

[0049] Compared with the prior art, the present invention has the following advantages:

[0050] The present invention proposes a method, system, device and medium for recommending customer source clues based on generated objects. The method includes: pre-processing the main business of the user side, extracting the downstream characteristics of the main business to generate a basic list of recommended objects, and updating the target list of the recommended object side according to the basic list; performing multi-way recall of the recommended objects in the target list according to the user portrait, filtering and sorting the recall results, and outputting the recall set and the candidate recall set; setting the recommendation strategy according to the recall set to screen the list generation task of the candidate recall set, and returning the list generation task to the recommended object side to update the target list. Compared with the prior art, it has the following effects: the basic list is generated by AI to accurately obtain the target list, thereby providing reliable data support for the recommendation side, and the recommendation side returns the candidate recall set of potential customer source clues to the generation side to increase and update the target list library. The continuous cycle of this process is dynamic and autonomous, which can help users obtain accurate and complete recommended objects, solve the problem that the item pool of customer acquisition clues is empty and the user-item cannot be adapted and linked in the recommendation scenario, and has great application value in the field of intelligent sales. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0052] Figure 1 A flow chart of a customer source lead recommendation method based on a generated object of the present invention;

[0053] Figure 2It is a schematic diagram of the customer source clue recommendation system based on the generated object of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0055] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the content disclosed in this application, some changes in design, manufacturing or production based on the technical content disclosed in this application are just conventional technical means, and should not be understood as insufficient content disclosed in this application.

[0056] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0057] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships, for example, "A and / or B" can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0058] Example 1

[0059] See also Figure 1 The embodiment of the present invention discloses a customer source lead recommendation method based on a generated object, comprising:

[0060] 101. Preprocess the main business of the user side, extract the downstream features of the main business to generate a basic list of recommended objects, and update the target list of the recommended objects according to the basic list;

[0061] It should be noted that for recommendation system products, they are mainly used in TOB and TOC business sections, providing customer source clues for many small and micro manufacturers, shops, studios, self-employed users, etc. Different industries have different customer development needs according to their respective business characteristics. They may pay more attention to bidding information, and may pay more attention to customers who are active on short video platforms. For example, if the user's main business is hardware and the region is Guangdong, the customer source clues adapted to the user may include but be limited to the list of employees of companies in the plastics industry in Guangdong Province that need hardware products, or the list of employees of logistics companies in Guangdong Province that need hardware products for transportation equipment assembly, etc. It can be understood that the customer source clues in the above examples are actually a small customer group with common characteristics. Compared with the specific recommendation words such as products, companies, and stores, the basic source data for obtaining recommendation words (such as crawling enterprise information through enterprise knowledge graphs) is only a database of a set of recommendation words, and there is no specific customer group concept. Therefore, how to establish an effective connection between the user side and the recommendation side is a problem that the current recommendation system needs to consider. In this application, the following will elaborate on how to solve the pain points of the current recommendation system.

[0062] In an optional embodiment, the pre-processing of the main business of the user side includes:

[0063] According to the main business information filled in by the user, the system performs cleaning processing such as deduplication, processing missing values, processing abnormal values, unifying the format and data verification, as well as data integration, data analysis and mining, and purification processing to extract key information to obtain the main business terms.

[0064] Specifically, in order to further establish connections between users, a user account needs to be created before the user uses the recommendation system. Therefore, the user needs to fill in the user's main business according to the actual situation. In addition to extracting the preferences of the user's target object through the user's main business words, other users can also establish connections with the user through the main business information. The main business words can help analyze user portraits, including but not limited to the customer sources that the user can provide and the customer that the user wants.

[0065] In an optional embodiment, the step of extracting downstream features of the main business to generate a basic list of recommended objects includes:

[0066] A large language model is used to build a downstream feature judger, and the downstream feature values ​​generated based on the main business words are set to Boolean type to judge the association between the main business and downstream features;

[0067] The association relationship of downstream features determines the branch path of the basic directory to be generated, wherein the first branch path sets type rules and word expansion rules for downstream features of known recommended object types to generate a basic directory, and the second branch path uses a large language model to generate titles and query conditions for downstream features of unknown recommended object types to generate a basic directory.

[0068] Specifically, a large language model is used for training using downstream corpora to learn downstream features of different main businesses, and the feature values ​​of the downstream features generated after inputting the main business words use a Boolean type to indicate whether an object meets a certain condition, such as whether the type of the downstream feature is individual / store / enterprise (helping users to classify the types of recommended objects), whether the downstream feature focuses on winning bids (helping users to classify the preferences for recommended objects), whether the downstream feature depends on distance (helping users to classify the regional preferences for recommended objects), whether the downstream feature focuses on company employees / bosses (helping users to screen the types of customer acquisition leads), and whether the downstream feature is financial business / liquor business (helping users to classify the business fields of recommended objects). This can generate more downstream customer acquisition leads that may be related to the main business based on the user's main business.

[0069] Furthermore, the basic directory can be understood as a query tag for querying the target recommended object in the basic source data. By adopting different title generation and query condition generation steps according to different downstream features, a query sentence related to the user's main business needs is obtained. For downstream objects such as winning bidders, bidding roles, short video active users, and account opening bank users, the rule-generated title + extended keywords are used as query conditions. The rule generation determines the title that more accurately expresses the user's customer development needs according to industry rules, and the extended keywords expand the input downstream features through the large language model. For example, for the downstream feature value indicating whether to pay attention to the winning bid, the basic directory generated by the rule-generated title and word expansion processing is the user's need to find the winning bid, or when the user's main business downstream is an individual, a store or an enterprise, the basic directory can be generated by rules and word expansion as short video customer development, Gaode map shop / enterprise search, etc., thereby constructing a user portrait of the user's customer source clue needs. For the case where the object type cannot be determined, the industry situation learned by the large language model can be used to generate the related basic directory as much as possible, thereby helping the user find more recommended objects.

[0070] In an optional embodiment, the updating of the target list of the recommended object side according to the basic directory includes:

[0071] Obtain the user's regional preferences and add them to the basic directory. Obtain the entity content of the recommended object according to the branch path of the basic directory to generate a target list.

[0072] Specifically, the above steps complete the user portrait through the basic directory, and the query dimension can be completed by further adding the user's regional preference. Since the basic directory itself is a fixed query condition, the target list can be obtained directly from the basic source data without consuming resources to sort out the business attributes within and between different industries. The generated target list, for example, if the user's main business is hardware, then the generated target list is a list of company employees in a certain area who need hardware products. Among them, it can be further determined based on the company's situation (such as the company's structure and the number of people in the company) whether the individual responsible for procurement is the department leader or the boss, thereby generating the contact information of the procurement department leader or the boss. The list information and contact information of multiple different customers constitute the target list as recommended objects. It can be understood that as the user provides more main business information, more target lists can be generated. Therefore, the process of generating the target list is a dynamic growth process. Compared with the existing recommendation system, it can reduce resource consumption and gradually expand the content content of the target list library and the accuracy of recommendations.

[0073] 102. Perform multi-way recall of recommended objects in the target list according to the user portrait, filter and sort the recall results, and then output the recall set and candidate recall set;

[0074] In an optional embodiment, the multi-channel recall of recommended objects in the target list according to the user portrait includes:

[0075] Recall the recommended objects in the target list according to the user preferences in the basic directory;

[0076] Set matching rules based on the user's historical input words and recall recommended objects that match the input words;

[0077] Recall similar recommended objects based on the historical recommended objects clicked by the user;

[0078] Recall similar recommended objects based on the historical recommended objects clicked by users with similar portraits.

[0079] In an optional embodiment, the recall results are filtered and sorted to output the recall set and the candidate recall set.

[0080] Obtain the user's historical preferences and historical consumption, and filter the recall results based on historical data;

[0081] The ranking model is constructed to calculate the ranking score of each recommended object by weighted average of the filtered recall results according to the user's main business association value, regional preference value, object type matching value, title similarity value and query condition similarity value, and output the recall set and candidate recall set of the topk recommended objects.

[0082] Specifically, the user can recall from the target list library generated in step 101, and the multi-way recall strategy can return the target list that is most similar to the user profile represented by the basic directory as much as possible, and at the same time use the user's behavior data to make the recommendation results as personalized as possible, thereby refining the recommendation content for each user and making the recommendation results more personalized and customized.

[0083] Furthermore, in order to display the recommendation results as much as possible, the scores of the recommendation results are calculated through the main business scorer, regional preference scorer, customer type scorer, etc., and the recommendation results close to the user portrait are sorted according to the scores and displayed on the user's display interface. The recommendation results that have some similarities but cannot completely match the user portrait are marked as candidate recall sets, which may contain potential customers of the user.

[0084] 103. Set a recommendation strategy for the recall set to filter the candidate recall set list generation task, and return the list generation task to the recommendation object side to update the target list.

[0085] In an optional embodiment, the list generation task of screening the candidate recall set according to the recall set recommendation strategy is returned to the recommendation object side to update the target list.

[0086] Compare the preference matching degree between the user portrait and the recommended object according to the recall set and the candidate recall set, determine the extensible target recommended object in the candidate recall according to the preference matching degree, set the target recommended object as the target downstream feature of the list to be generated, organize the target downstream features into the list generation task, return the basic list generation step to expand the target list library.

[0087] Specifically, through the potential customer source leads displayed by the candidate recall set, the candidate recall set is returned to the downstream feature processing step to regenerate a new basic directory, thereby expanding the target list library. For example, when the result of the recall based on user preferences in the multi-way recall is empty, it is necessary to trigger the list generation task based on the main business words and regions. If the basic part of the list other than the region is consistent with the user's business, it is necessary to generate a matching list that meets the user's target region. It can be understood that the target list generation process on the generation side is related to the target list recall process on the recommendation side. The basic directory is generated by AI to accurately obtain the target list, thereby providing reliable data support for the recommendation side. The recommendation side returns the candidate recall set of potential customer source leads to the generation side to increase and update the target list library. The continuous cycle of this process is dynamic and autonomous, which can help users obtain accurate and complete recommendation objects (customer acquisition leads), and solves the problem of empty item pool and inability to adapt user-item linkage in the recommendation scenario. It has great application value in the field of intelligent sales.

[0088] like Figure 2As shown, the second aspect of the present invention discloses a customer source clue recommendation system based on a generated object, the system comprising:

[0089] The recommended object generation module is used to pre-process the main business of the user side, extract the downstream features of the main business to generate a basic list of recommended objects, and update the target list of the recommended object side according to the basic list;

[0090] The recommended object recall module is used to perform multi-way recall of recommended objects in the target list according to user portraits, filter and sort the recall results, and then output the recall set and candidate recall set;

[0091] The recommendation object linkage module is used to set the recommendation strategy for the recall set to filter the candidate recall set list generation task, and return the list generation task to the recommendation object side to update the target list.

[0092] In an optional embodiment, the recommended object generation module includes:

[0093] The data preprocessing module is used to remove duplicates, process missing values, process abnormal values, unify formats and perform data verification based on the main business information filled in by the user, as well as perform data integration, data analysis and mining on the main business information and purify the extraction of key information to obtain the main business terms.

[0094] In an optional embodiment, the recommended object generation module includes:

[0095] A downstream feature processing module is used to construct a downstream feature judgement device using a large language model, and set the downstream feature value generated according to the main business word to a Boolean type to judge the association between the main business and the downstream feature;

[0096] The basic directory processing module is used to determine the branch path of the basic directory to be generated based on the association relationship of downstream features, wherein the first branch path sets type rules and word expansion rules for downstream features of known recommended object types to generate a basic directory, and the second branch path uses a large language model to generate titles and query conditions for downstream features of unknown recommended object types to generate a basic directory.

[0097] In an optional embodiment, the recommended object generation module includes:

[0098] The entity data acquisition module is used to obtain the user's regional preferences and add them to the basic directory, obtain the entity content of the recommended object according to the branch path of the basic directory, and generate a target list.

[0099] In an optional embodiment, the recommended object recall module includes:

[0100] The recall module is used to recall the recommended objects in the target list according to the user preferences in the basic directory; to recall the recommended objects that match the input words according to the matching rules set according to the user's historical input words; to recall similar recommended objects based on the historical recommended objects clicked by the user; and to recall similar recommended objects based on the historical recommended objects clicked by users with similar portraits.

[0101] In an optional embodiment, the recommended object recall module includes:

[0102] The filtering module is used to obtain the user's historical preferences and historical consumption, and filter the recall results based on historical data;

[0103] The sorting module is used to build a sorting model to calculate the ranking score of each recommended object by weighted average of the filtered recall results according to the user's main business association value, regional preference value, object type matching value, title similarity value and query condition similarity value, and output the recall set and candidate recall set of the topk recommended objects.

[0104] In an optional embodiment, the recommended object linkage module includes:

[0105] The target list extension module is used to compare the preference matching between the user portrait and the recommended object according to the recall set and the candidate recall set, determine the extensible target recommendation object in the candidate recall according to the preference matching, set the target recommendation object as the target downstream feature of the list to be generated, organize the target downstream features into the list generation task and return the basic list generation step to expand the target list library.

[0106] The third aspect of the present invention discloses a customer source clue recommendation device based on a generated object, comprising:

[0107] at least one processor, and

[0108] a memory communicatively connected to the at least one processor; wherein,

[0109] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the customer source lead recommendation method based on generating objects as disclosed in any one of the first aspects of the present invention.

[0110] The computer device may be a terminal, and the computer device includes a processor, a memory, a network interface, a display screen, and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a customer source clue recommendation method based on a generated object is implemented. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball, or touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0111] A fourth aspect of the present invention discloses a computer-readable storage medium storing computer-executable instructions for causing a computer to execute a customer source lead recommendation method based on a generated object as disclosed in any one of the first aspect of the present invention.

[0112] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the customer source clue recommendation method based on the generated object as described above. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0113] Alternatively, if the above modules of the present invention are implemented in the form of software function modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the relevant technology. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, terminal, or network device, etc.) to execute all or part of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, RAM, ROM, magnetic disks or optical disks.

[0114] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A customer source lead recommendation method based on a generated object, characterized in that: The method comprises: Preprocess the main business of the user side, extract the downstream features of the main business to generate a basic list of recommended objects, and update the target list of recommended objects based on the basic list; Perform multi-way recall of recommended objects in the target list based on user portraits, filter and sort the recall results, and then output the recall set and candidate recall set; According to the recall set setting recommendation strategy, the candidate recall set list generation task is screened, and the list generation task is returned to the recommendation object side to update the target list.

2. The customer source lead recommendation method based on generated objects according to claim 1, characterized in that: The main business of the pre-processing user side includes: According to the main business information filled in by the user, the system performs cleaning processing such as deduplication, processing missing values, processing abnormal values, unifying the format and data verification, as well as data integration, data analysis and mining, and purification processing to extract key information to obtain the main business terms.

3. The customer source lead recommendation method based on generated objects according to claim 1, characterized in that: The step of extracting downstream features of the main business to generate a basic list of recommended objects includes: A large language model is used to build a downstream feature judger, and the downstream feature values ​​generated based on the main business words are set to Boolean type to judge the association between the main business and downstream features; The association relationship of downstream features determines the branch path of the basic directory to be generated, wherein the first branch path sets type rules and word expansion rules for downstream features of known recommended object types to generate a basic directory, and the second branch path uses a large language model to generate titles and query conditions for downstream features of unknown recommended object types to generate a basic directory.

4. The customer source lead recommendation method based on generated objects according to claim 3, characterized in that: The updating of the target list of the recommended object side according to the basic directory includes: Obtain the user's regional preferences and add them to the basic directory. Obtain the entity content of the recommended object according to the branch path of the basic directory to generate a target list.

5. The customer source lead recommendation method based on generated objects according to claim 1, characterized in that: The multi-channel recall of recommended objects in the target list based on the user portrait includes: Recall the recommended objects in the target list according to the user preferences in the basic directory; Set matching rules based on the user's historical input words and recall recommended objects that match the input words; Recall similar recommended objects based on the historical recommended objects clicked by the user; Recall similar recommended objects based on the historical recommended objects clicked by users with similar portraits.

6. The customer source lead recommendation method based on generated objects according to claim 1, characterized in that: The recall results are filtered and sorted to output the recall set and the candidate recall set. Obtain the user's historical preferences and historical consumption, and filter the recall results based on historical data; The ranking model is constructed to calculate the ranking score of each recommended object by weighted average of the filtered recall results according to the user's main business association value, regional preference value, object type matching value, title similarity value and query condition similarity value, and output the recall set and candidate recall set of the topk recommended objects.

7. The customer source lead recommendation method based on generated objects according to claim 1, characterized in that: The list generation task of screening the candidate recall set according to the recommendation strategy of the recall set is returned to the recommendation object side to update the target list. Compare the preference matching degree between the user portrait and the recommended object according to the recall set and the candidate recall set, determine the extensible target recommended object in the candidate recall according to the preference matching degree, set the target recommended object as the target downstream feature of the list to be generated, organize the target downstream features into the list generation task, return the basic list generation step to expand the target list library.

8. A customer source lead recommendation system based on generated objects, characterized in that: The system comprises: The recommended object generation module is used to pre-process the main business of the user side, extract the downstream features of the main business to generate a basic list of recommended objects, and update the target list of the recommended object side according to the basic list; The recommended object recall module is used to perform multi-way recall of recommended objects in the target list according to user portraits, filter and sort the recall results, and then output the recall set and candidate recall set; The recommendation object linkage module is used to set the recommendation strategy for screening the candidate recall set list generation task, and return the list generation task to the recommendation object side to update the target list.

9. A customer source lead recommendation device based on a generated object, characterized in that: include: at least one processor, and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the customer source lead recommendation method based on generating objects as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the customer source lead recommendation method based on the generated object according to any one of claims 1 to 7.

Images