Search

CN-121998734-A - Product recommendation method, device, equipment, medium and product

CN121998734ACN 121998734 ACN121998734 ACN 121998734ACN-121998734-A

Abstract

The embodiment of the application provides a product recommendation method, device, equipment, medium and product, and relates to the field of artificial intelligence or financial science and technology. Firstly, receiving a product recommendation request of a user, and acquiring user identification information. Then, based on the user identification information, a corresponding intention recognition result is determined. And then, generating a plurality of recommended products and corresponding recommendation reasons according to the intention recognition result through the collaborative recommendation agent group. Finally, these recommendation results are output to the user. The method breaks through the limitation of the traditional fixed rule through flexible collaboration and intention depth recognition of the collaborative recommendation intelligent agent group, realizes accurate recommendation fitting the personalized needs of the user, and meanwhile, the presentation mode with the recommendation reason improves the interpretability and the credibility of the recommendation result, enhances the user interaction experience and effectively improves the adaptation degree and the user acceptance degree of the product recommendation.

Inventors

  • Gao Peixuan

Assignees

  • 中国工商银行股份有限公司

Dates

Publication Date
20260508
Application Date
20260116

Claims (11)

  1. 1. A method of product recommendation, the method comprising: receiving a product recommendation request, wherein the product recommendation request comprises user identification information; determining an intention recognition result corresponding to the user identification information; Determining a product recommendation result corresponding to the intention recognition result through a collaborative recommendation agent group, wherein the product recommendation result comprises a plurality of recommended products and recommendation reasons corresponding to the recommended products; And outputting the product recommendation result.
  2. 2. The method of claim 1, wherein the set of co-recommended agents includes a first agent, a second agent, and a third agent, wherein determining, by the set of co-recommended agents, a product recommendation corresponding to the intent recognition result includes: Performing exception evaluation on the user portrait associated with the user identification information through the first intelligent agent to generate an exception grade; Determining, by the second agent, the plurality of recommended products based on the anomaly level and the intent recognition result; And generating recommendation reasons corresponding to the recommended products based on the plurality of recommended products through the third agent, and determining the product recommendation results based on the recommended products and the recommendation reasons corresponding to the recommended products.
  3. 3. The method of claim 2, wherein the set of co-recommended agents further comprises a fourth agent, the method further comprising: if the user identification information indicates that the user is a new user, carrying out question-answer interaction with the new user through the fourth agent to obtain a question-answer result, and determining a user portrait of the new user based on the question-answer result; and if the user identification information indicates that the user is an old user, acquiring historical transaction data based on the user identification information, and analyzing and processing the historical transaction data to obtain a user image of the old user.
  4. 4. The method of claim 2, wherein the determining, by the second agent, the plurality of recommended products based on the anomaly level and the intent recognition result comprises: Acquiring a product knowledge graph through the second agent, wherein the product knowledge graph comprises a plurality of candidate products and product attributes corresponding to the candidate products; Determining the product grade corresponding to each candidate product according to the product attribute corresponding to each candidate product; and determining the recommended products from the candidate products based on the intention recognition result, the abnormal grade and the product grade corresponding to each candidate product.
  5. 5. The method of claim 4, wherein the generating, by the third agent, a recommendation reason corresponding to each of the recommended products based on the plurality of recommended products, comprises: and inputting the product attribute corresponding to each recommended product into a language generation model through the third agent to obtain the recommended reason corresponding to each recommended product.
  6. 6. A method according to claim 3, wherein the performing, by the fourth agent, a question-answer interaction with the new user to obtain a question-answer result includes: In the question-answer interaction process, a preset filtering word list is obtained through the fourth agent; Receiving a first user problem, and filtering the first user problem based on the filtering word list to obtain a second user problem; and carrying out question-answer interaction with the user based on the second user question until the question-answer result is obtained.
  7. 7. The method according to claim 1, wherein the method further comprises: Acquiring multi-mode data; Inputting the multi-mode data into a behavior feature analysis model to obtain a behavior feature result output by the behavior feature analysis model; and according to the behavior characteristic result, adjusting the product recommendation result to obtain an adjusted product recommendation result, and outputting the adjusted product recommendation result.
  8. 8. A product recommendation device, comprising: the receiving module is used for receiving a product recommendation request, wherein the product recommendation request comprises user identification information; the determining module is used for determining an intention recognition result corresponding to the user identification information; The determining module is further configured to determine a product recommendation result corresponding to the intention recognition result through a collaborative recommendation agent group, where the product recommendation result includes a plurality of recommended products and recommendation reasons corresponding to the recommended products; And the output module is also used for outputting the product recommendation result.
  9. 9. A product recommendation device is characterized by comprising a memory and a processor; The memory stores computer-executable instructions; The processor executing computer-executable instructions stored in the memory, causing the processor to perform the method of any one of claims 1-7.
  10. 10. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor are adapted to carry out the method of any one of claims 1-7.
  11. 11. A computer program product comprising a computer program which, when executed by a processor, implements the method of any of claims 1-7.

Description

Product recommendation method, device, equipment, medium and product Technical Field The application relates to the field of artificial intelligence or financial science and technology, in particular to a product recommendation method, device, equipment, medium and product. Background The intelligent customer service system is an important tool for improving the customer service efficiency in banking industry. Through intelligent customer service, a customer can conveniently interact with the system at a mobile banking APP or webpage end to acquire personalized recommended services. Existing product recommendation approaches rely primarily on rule engines or simple FAQ matching mechanisms. The method is essentially characterized in that through presetting fixed business rules and keyword matching templates, surface layer analysis is carried out on a demand text input by a user, and then products meeting basic conditions are called from a product library to be recommended. However, such recommendation methods relying on a rule engine or simple FAQ matching cannot fully consider the personalized needs, preferences and investment backgrounds of users mainly based on fixed rules or templates, so that recommended financial products often have difficulty in meeting the diversified needs of users. Disclosure of Invention The embodiment of the application provides a product recommending method, a device, equipment, a medium and a product, which are used for solving the problem that the conventional product recommending mode depends on fixed rules or templates and cannot meet the diversified demands of users. In a first aspect, an embodiment of the present application provides a product recommendation method, including: receiving a product recommendation request, wherein the product recommendation request comprises user identification information; determining an intention recognition result corresponding to the user identification information; Determining a product recommendation result corresponding to the intention recognition result through a collaborative recommendation agent group, wherein the product recommendation result comprises a plurality of recommended products and recommendation reasons corresponding to the recommended products; And outputting the product recommendation result. In one possible implementation manner, the collaborative recommendation agent group comprises a first agent, a second agent and a third agent, wherein the determining, by the collaborative recommendation agent group, a product recommendation result corresponding to the intention recognition result comprises: Performing exception evaluation on the user portrait associated with the user identification information through the first intelligent agent to generate an exception grade; Determining, by the second agent, the plurality of recommended products based on the anomaly level and the intent recognition result; And generating recommendation reasons corresponding to the recommended products based on the plurality of recommended products through the third agent, and determining the product recommendation results based on the recommended products and the recommendation reasons corresponding to the recommended products. In one possible embodiment, the collaborative recommendation agent group further comprises a fourth agent, the method further comprising: if the user identification information indicates that the user is a new user, carrying out question-answer interaction with the new user through the fourth agent to obtain a question-answer result, and determining a user portrait of the new user based on the question-answer result; and if the user identification information indicates that the user is an old user, acquiring historical transaction data based on the user identification information, and analyzing and processing the historical transaction data to obtain a user image of the old user. In one possible embodiment, the determining, by the second agent, the plurality of recommended products based on the anomaly level and the intention recognition result includes: Acquiring a product knowledge graph through the second agent, wherein the product knowledge graph comprises a plurality of candidate products and product attributes corresponding to the candidate products; Determining the product grade corresponding to each candidate product according to the product attribute corresponding to each candidate product; and determining the recommended products from the candidate products based on the intention recognition result, the abnormal grade and the product grade corresponding to each candidate product. In one possible implementation manner, the generating, by the third agent, a recommendation reason corresponding to each recommended product based on the plurality of recommended products includes: and inputting the product attribute corresponding to each recommended product into a language generation model through the third agent to