Search

CN-121980077-A - Car selection recommendation method and related equipment

CN121980077ACN 121980077 ACN121980077 ACN 121980077ACN-121980077-A

Abstract

The embodiment of the application provides an automobile selection recommending method and related equipment, and belongs to the technical field of artificial intelligence. The method comprises the steps of obtaining vehicle selection requirement information input by a user, inputting the vehicle selection requirement information into an automobile industry language understanding pipeline finely tuned in the automobile field to be processed to obtain a structured user requirement vector and user intention, wherein the automobile industry language understanding pipeline integrates a large language model finely tuned in the automobile field, searching from a preset automobile knowledge base based on the user requirement vector to obtain an initial candidate vehicle type set, calling a reordering model corresponding to the user intention to order the initial candidate vehicle type set according to the user intention to obtain a target recommended vehicle type list, and generating and outputting a vehicle selection recommendation result containing the target recommended vehicle type list. The embodiment of the application can realize the conversion from fuzzy requirements to accurate recommendations, thereby improving the intelligent level of the vehicle selecting service and the satisfaction degree of users.

Inventors

  • YANG FASHENG
  • FU LUSHAN

Assignees

  • 广州太平洋电脑信息咨询有限公司

Dates

Publication Date
20260505
Application Date
20251231

Claims (10)

  1. 1. An automobile selection recommendation method is characterized by comprising the following steps: Acquiring vehicle selecting requirement information input by a user; Inputting the vehicle selection demand information into a vehicle industry language understanding pipeline which is finely tuned in the vehicle field for processing to obtain a structured user demand vector and user intention, wherein the vehicle industry language understanding pipeline integrates a large language model which is finely tuned in the vehicle field; Searching from a preset automobile knowledge base based on the user demand vector to obtain an initial candidate automobile model set; And calling a reordering model corresponding to the user intention to order the initial candidate vehicle type set according to the user intention to obtain a target recommended vehicle type list, and generating and outputting a vehicle selection recommendation result containing the target recommended vehicle type list.
  2. 2. The method of claim 1, wherein the inputting the selection requirement information into the automotive industry language understanding pipeline of automotive field fine tuning for processing to obtain a structured user requirement vector and a user intention comprises: Performing text cleaning and standardization processing on the vehicle selecting requirement information to obtain a processed text; the word segmentation tool combined with the automobile special name dictionary is utilized to segment the processed text, and the special name label sequence is obtained by identifying the word segmentation result through the special noun identification model; Splicing the processed text and the special name tag sequence into a prompt template, and inputting the prompt template into the large language model subjected to corpus fine adjustment in the automotive field to perform generated slot filling to obtain structured data containing a plurality of business slot values; Calculating the user demand vector based on each slot position value and the explicit degree coefficient thereof in the structured data; And splicing the processed text and the structured data, and inputting the spliced text and the structured data into an intention classification model to obtain the user intention.
  3. 3. The method of claim 2, wherein prior to said inputting said selection demand information into the automotive industry language understanding pipeline of automotive domain fine tuning for processing to obtain a structured user demand vector and user intent, said method further comprises: carrying out weighted fusion on sentence vectors input by a user in the current round and the memory tensor in the previous round, and updating to obtain the memory tensor in the current round; And splicing the memory tensor of the current round with the sentence vector input by the user of the current round, and inputting the spliced memory tensor and the sentence vector into the language understanding assembly line of the automobile industry.
  4. 4. The method according to claim 2, wherein the calculating the user demand vector based on each slot value in the structured data and the explicit degree coefficient thereof comprises: For a first service slot sub-item, if a strong modal word is identified from user input, a first weight coefficient is allocated to the first service slot sub-item; if the weak modal word is identified, a second weight coefficient is distributed to the first service slot sub-item; if the modal word is not recognized, a basic weight coefficient is allocated to the first business slot sub-item, wherein the first weight coefficient is larger than the second weight coefficient, and the second weight coefficient is larger than the basic weight coefficient; carrying out weighted calculation on the slot position value of each service slot position sub-item, the corresponding weight coefficient and the inverse document frequency of the sub-item in the knowledge base to obtain an initial demand vector; And carrying out normalization processing on the initial demand vector to obtain the user demand vector.
  5. 5. The method of claim 1, wherein the retrieving, based on the user demand vector, from a preset automobile knowledge base to obtain an initial candidate vehicle model set includes: And mapping the user demand vector into a query vector, and searching K vehicle types which are most similar to the query vector in an automobile knowledge base constructed based on a vector database according to the query vector by using an approximate nearest neighbor search algorithm to form the initial candidate vehicle type set, wherein K is a positive integer.
  6. 6. The method according to any one of claims 1 to 5, wherein the calling a reordering model corresponding to the user intention to order the initial candidate vehicle model set according to the user intention to obtain a target recommended vehicle model list includes: If the user intention is the vehicle type comparison intention, calling a sequencing model based on comparison learning to reorder the initial candidate vehicle type set to obtain the target recommended vehicle type list; and if the user intention is the vehicle selection recommendation intention, invoking a depth factor decomposition machine model based on comparison learning to reorder the initial candidate vehicle type set so as to obtain the target recommended vehicle type list.
  7. 7. The method of claim 1, wherein the generating and outputting the pick-up recommendation comprising the list of target recommended vehicle types comprises: and generating a visual content component card according to the target recommended vehicle type list, wherein the visual content component card at least comprises vehicle type key parameters and user evaluation abstracts, and displaying the visual content component card on a target interface.
  8. 8. An automobile selection recommendation device, characterized in that the device comprises: The information acquisition module is used for acquiring the vehicle selection requirement information input by the user; the information processing module is used for inputting the vehicle selection demand information into an automobile industry language understanding pipeline which is finely tuned in the automobile field for processing to obtain a structured user demand vector and user intention, wherein the automobile industry language understanding pipeline integrates a large language model which is finely tuned in the automobile field; the demand retrieval module is used for retrieving from a preset automobile knowledge base based on the user demand vector to obtain an initial candidate vehicle type set; the vehicle type ordering module is used for calling a reordering model corresponding to the user intention to order the initial candidate vehicle type set according to the user intention to obtain a target recommended vehicle type list, and generating and outputting a vehicle selection recommendation result containing the target recommended vehicle type list.
  9. 9. A computer device, characterized in that it comprises a memory storing a computer program and a processor implementing the method according to any one of claims 1 to 7 when executing the computer program.
  10. 10. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 7.

Description

Car selection recommendation method and related equipment Technical Field The application relates to the technical field of artificial intelligence, in particular to an automobile selection recommending method and related equipment. Background With the vigorous development and the increasingly vigorous competition of the automobile market, consumers face challenges of information overload and decision complexity during vehicle selection, and the traditional vehicle selection modes relying on offline recommendation, media evaluation and personal experience have the limitations of insufficient efficiency, strong subjectivity and the like, so that comprehensive and objective vehicle purchase suggestions are difficult to provide. In recent years, the breakthrough of artificial intelligence technology, particularly large models in the fields of natural language, visual recognition and the like brings innovation opportunities for the automobile industry, and intelligent application based on the large models can deeply understand user demands, so that highly personalized and accurate vehicle selection service is provided, and diversified and intelligent demands of users in the vehicle selection process are effectively met. Disclosure of Invention The embodiment of the application mainly aims to provide an automobile selecting recommendation method and related equipment, which can improve the intelligent level of automobile selecting service and the satisfaction degree of users. In order to achieve the above object, an aspect of an embodiment of the present application provides an automobile selection recommendation method, which includes: Acquiring vehicle selecting requirement information input by a user; Inputting the vehicle selection demand information into a vehicle industry language understanding pipeline which is finely tuned in the vehicle field for processing to obtain a structured user demand vector and user intention, wherein the vehicle industry language understanding pipeline integrates a large language model which is finely tuned in the vehicle field; Searching from a preset automobile knowledge base based on the user demand vector to obtain an initial candidate automobile model set; And calling a reordering model corresponding to the user intention to order the initial candidate vehicle type set according to the user intention to obtain a target recommended vehicle type list, and generating and outputting a vehicle selection recommendation result containing the target recommended vehicle type list. In some embodiments, the inputting the vehicle selection requirement information into a vehicle industry language understanding pipeline of the fine tuning of the vehicle field to process to obtain a structured user requirement vector and a user intention includes: Performing text cleaning and standardization processing on the vehicle selecting requirement information to obtain a processed text; the word segmentation tool combined with the automobile special name dictionary is utilized to segment the processed text, and the special name label sequence is obtained by identifying the word segmentation result through the special noun identification model; Splicing the processed text and the special name tag sequence into a prompt template, and inputting the prompt template into the large language model subjected to corpus fine adjustment in the automotive field to perform generated slot filling to obtain structured data containing a plurality of business slot values; Calculating the user demand vector based on each slot position value and the explicit degree coefficient thereof in the structured data; And splicing the processed text and the structured data, and inputting the spliced text and the structured data into an intention classification model to obtain the user intention. In some embodiments, before the inputting the vehicle selection requirement information into the vehicle industry language understanding pipeline of the fine tuning of the vehicle field to process to obtain the structured user requirement vector and the user intention, the method further includes: carrying out weighted fusion on sentence vectors input by a user in the current round and the memory tensor in the previous round, and updating to obtain the memory tensor in the current round; And splicing the memory tensor of the current round with the sentence vector input by the user of the current round, and inputting the spliced memory tensor and the sentence vector into the language understanding assembly line of the automobile industry. In some embodiments, the calculating the user demand vector based on each slot value in the structured data and the explicit degree coefficient thereof includes: For a first service slot sub-item, if a strong modal word is identified from user input, a first weight coefficient is allocated to the first service slot sub-item; if the weak modal word is identified, a second weight coefficient is distrib