CN-122019860-A - Vehicle-mounted chip selection method, device, equipment and storage medium

CN122019860ACN 122019860 ACN122019860 ACN 122019860ACN-122019860-A

Abstract

A vehicle-mounted chip model selection method, device, equipment and storage medium comprises the steps of continuously obtaining relevant information of a vehicle-mounted chip according to a network crawling tool and an internal data source to update a local multidimensional feature library for constructing the vehicle-mounted chip and construct a local vector multi-element database, receiving a natural language problem input by a user for vehicle-mounted chip model selection, carrying out natural language processing and feature extraction on the natural language problem to generate a feature vector of the natural language problem, carrying out similarity retrieval on the feature vector of the natural language problem and a vector in the local vector multi-element database, sorting candidate chips according to feature weights, inputting the sorted candidate chips and matched features thereof into a preset generation model to generate vehicle-mounted chip model selection suggestions, and solving the technical problems that the traditional vehicle-mounted chip model selection method mainly depends on manual experience and a simple retrieval mechanism and is difficult to meet the high-efficiency and accurate decision requirement of modern industry.

Inventors

LI XIAOBO
ZHU ZENG
MEN CHEN

Assignees

东风汽车集团股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260303

Claims (10)

1. The vehicle-mounted chip type selecting method is characterized by comprising the following steps of: Continuously acquiring relevant information of the vehicle-mounted chip according to the network crawling tool and the internal data source so as to update a local multidimensional feature library for constructing the vehicle-mounted chip and construct a local vector multi-metadata database; Receiving a natural language problem selected by a vehicle-mounted chip input by a user, performing natural language processing and feature extraction on the natural language problem, and generating a feature vector of the natural language problem; Performing similarity retrieval on the feature vector of the natural language problem and the vector in the local vector multivariate database, and sequencing candidate chips according to feature weights by combining KNN algorithm and semantic similarity calculation; inputting the sorted candidate chips and the matching characteristics thereof into a preset generation model, integrating multidimensional matching information through the preset generation model, and generating a vehicle-mounted chip model selection suggestion.
2. The method for selecting the vehicle-mounted chip according to claim 1, wherein the continuously acquiring the relevant information of the vehicle-mounted chip according to the network crawling tool and the internal data source to update the local multidimensional feature library for constructing the vehicle-mounted chip and construct the local vector multi-metadata database comprises: The method comprises the steps of organizing chip characteristics according to preset dimensions by acquiring basic parameters and technical characteristic data of a vehicle-mounted chip, and distributing weights to each dimension and characteristic items to construct a local multidimensional characteristic library of the vehicle-mounted chip, wherein the preset dimensions comprise a first layer of dimensions and a second layer of dimensions, the first layer of dimensions comprise business dimensions, technical dimensions and policy dimensions, and the second layer of dimensions comprise manufacturers, models, key parameters, historical usage of the unit and industry usage; And continuously acquiring related data of the vehicle-mounted chip through a network crawling tool and an internal data source, preprocessing, slicing and vectorizing the acquired related data, updating the local multidimensional feature library and constructing a local vector multi-metadata database.
3. The method for selecting the vehicle-mounted chip according to claim 2, wherein the continuously acquiring the relevant data of the vehicle-mounted chip through the network crawling tool and the internal data source, preprocessing, slicing and vectorizing the acquired relevant data, updating the local multidimensional feature library and constructing a local vector multi-metadata library comprises the following steps: acquiring data from enterprise internal historical data, industry standard files, chip manufacturer public parameters and public statistical reports; Periodically accessing a chip manufacturer official network, an automobile industry vertical website and an industry media report through a network crawling tool to acquire latest chip parameter change and design import information; removing stop words and punctuations by using a natural language word segmentation technology, simplifying the attention mechanism of a long text, and generating a standardized text; Dividing and slicing the standardized text according to a preset granularity; And vectorizing the sliced text by using an embedded model, generating vector representation, storing the vector representation, the original text, the structured feature value, the source identification, the acquisition time, the dimension label and the current weight value into the local multidimensional feature library to update the local vector multi-element database, and constructing the local vector multi-element database.
4. The method for selecting a vehicle-mounted chip according to claim 1, wherein the receiving the natural language question of the vehicle-mounted chip selection input by the user, performing natural language processing and feature extraction on the natural language question, and generating a feature vector of the natural language question, comprises: word segmentation processing is carried out on the natural language problem selected by the vehicle-mounted chip which is input by the user; Extracting business, technology or policy characteristics related to the vehicle-mounted chip model selection in the natural language problem; Constructing semantic representation according to the extracted business, technology or policy features and their interrelationships; generating feature vectors of the natural language question from the semantic representation using an embedding model.
5. The method for selecting the vehicle-mounted chip according to claim 1, wherein the step of searching the similarity between the feature vector of the natural language problem and the vector in the local vector multivariate database, and combining KNN algorithm and semantic similarity calculation, and sorting the candidate chips according to feature weights, comprises: calculating the similarity between the feature vector of the natural language problem and each vector in the local vector multivariate database; determining a plurality of records most similar to the natural language problem by using a KNN algorithm; weighting the similarity according to the current weight of each dimension characteristic; and arranging the candidate chips in a descending order according to the weighted comprehensive similarity.
6. The method for selecting the vehicle-mounted chip according to claim 1, wherein the step of inputting the ordered candidate chips and the matching features thereof into a preset generation model, integrating multi-dimensional matching information through the preset generation model, and generating the vehicle-mounted chip selection suggestion comprises the following steps: Inputting a plurality of candidate chips which are ranked at the front and each dimension matching characteristic list as a context into a preset generation model; weighting and fusing the multisource matching characteristics through the attention mechanism of the preset generation model; And generating natural language model selection suggestions including recommended chip priority, key matching characteristics of each chip and prompts of unsatisfied requirements.
7. The method for selecting the vehicle-mounted chip according to claim 1, wherein the step of inputting the sorted candidate chips and the matching features thereof into a preset generation model, integrating multi-dimensional matching information through the preset generation model, and generating the vehicle-mounted chip selection suggestion further comprises the steps of: respectively displaying the matching conditions of the recommended chips according to the business dimension, the technical dimension and the policy dimension; Displaying the contribution weight and the influence ordering diagram of each dimension feature on the final ordering; providing an input interface of user's acceptance degree of the recommended result, acceptance degree of each feature matching degree and feature weight adjustment opinion; Receiving and recording feedback data submitted by a user; Analyzing weight adjustment opinion aiming at specific characteristics in user feedback; Modifying the weight coefficient of the corresponding feature in the local multidimensional feature library; Synchronously writing the adjusted weight value into a weight metadata field of a related record in the local vector multivariate database; and carrying out batch migration adjustment on the weight templates of the similar characteristics according to the adjustment direction with higher occurrence frequency in the multiple feedback.
8. The vehicle-mounted chip type selection device is characterized by comprising: The construction module is used for continuously acquiring relevant information of the vehicle-mounted chip according to the network crawling tool and the internal data source so as to update a local multidimensional feature library for constructing the vehicle-mounted chip and construct a local vector multi-metadata database; The first generation module is used for receiving a natural language problem selected by a vehicle-mounted chip input by a user, carrying out natural language processing and feature extraction on the natural language problem, and generating a feature vector of the natural language problem; The sorting module is used for carrying out similarity retrieval on the feature vector of the natural language problem and the vector in the local vector multivariate database, and sorting the candidate chips according to the feature weight by combining KNN algorithm and semantic similarity calculation; the second generation module is used for inputting the ordered candidate chips and the matching characteristics thereof into a preset generation model, integrating multidimensional matching information through the preset generation model, and generating a vehicle-mounted chip model selection suggestion.
9. An on-vehicle chip-type selection device, characterized in that it comprises a processor, a memory, and an on-vehicle chip-type selection program stored on the memory and executable by the processor, wherein the on-vehicle chip-type selection program, when executed by the processor, implements the steps of the on-vehicle chip-type selection method according to any one of claims 1 to 7.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a vehicle-mounted chip-selection program, wherein the vehicle-mounted chip-selection program, when executed by a processor, implements the steps of the vehicle-mounted chip-selection method according to any one of claims 1 to 7.

Description

Vehicle-mounted chip selection method, device, equipment and storage medium Technical Field The present application relates to the field of data processing, and in particular, to a method, apparatus, device, and computer readable storage medium for selecting a vehicle-mounted chip. Background The automobile industry faces the transformation requirements of new energy and networking, and in the process, the whole automobile is required to be greatly adjusted in the related requirements of a whole automobile three-electric controller and an automobile machine system. Compared with the traditional fuel oil vehicle type, the new energy vehicle type needs to be provided with a large number of vehicle body situation sensing controllers and energy controllers thereof, and the safety and reliability of the vehicle type are determined by the performance and stability of the controllers. The key component of the high-performance controller is a high-performance vehicle-mounted chip. The new energy vehicle-mounted chip has the characteristics of multiple parameter conditions, high value and quick change. The traditional vehicle-mounted chip model selection method mainly depends on manual experience and a simple retrieval mechanism, and is difficult to meet the high-efficiency and accurate decision requirement of the modern industry. Along with the rapid increase of data volume and the improvement of professional requirements, the traditional knowledge base retrieval method is difficult to deal with complex query and fuzzy sentences, and cannot meet the requirements of a precise and rapid question-answering system. Meanwhile, the professional field has great influence on whole car products, all whole car enterprises of the related knowledge base are relatively closed, and the data change is difficult to update along with technology update to realize instant update. Meanwhile, due to the fact that the enterprises of professional technicians are relatively isolated, related knowledge cannot be updated accordingly, latest product information is obtained, the selectable product is narrow and closed, knowledge updating is slow, the model selection efficiency is low, the cost is high, and the error rate is high. Disclosure of Invention The application provides a vehicle-mounted chip type selection method, a device, equipment and a computer readable storage medium, which can solve the technical problems that the traditional vehicle-mounted chip type selection method in the prior art mainly depends on manual experience and a simple retrieval mechanism, and is difficult to meet the high-efficiency and accurate decision requirement of the modern industry. In a first aspect, an embodiment of the present application provides a method for selecting a vehicle-mounted chip, which is characterized by including: Continuously acquiring relevant information of the vehicle-mounted chip according to the network crawling tool and the internal data source so as to update a local multidimensional feature library for constructing the vehicle-mounted chip and construct a local vector multi-metadata database; Receiving a natural language problem selected by a vehicle-mounted chip input by a user, performing natural language processing and feature extraction on the natural language problem, and generating a feature vector of the natural language problem; Performing similarity retrieval on the feature vector of the natural language problem and the vector in the local vector multivariate database, and sequencing candidate chips according to feature weights by combining KNN algorithm and semantic similarity calculation; inputting the sorted candidate chips and the matching characteristics thereof into a preset generation model, integrating multidimensional matching information through the preset generation model, and generating a vehicle-mounted chip model selection suggestion. With reference to the first aspect, in an implementation manner, the continuously acquiring, according to the network crawling tool and the internal data source, related information of the vehicle-mounted chip, so as to update a local multidimensional feature library for constructing the vehicle-mounted chip, and construct a local vector multi-metadata database, includes: The method comprises the steps of organizing chip characteristics according to preset dimensions by acquiring basic parameters and technical characteristic data of a vehicle-mounted chip, and distributing weights to each dimension and characteristic items to construct a local multidimensional characteristic library of the vehicle-mounted chip, wherein the preset dimensions comprise a first layer of dimensions and a second layer of dimensions, the first layer of dimensions comprise business dimensions, technical dimensions and policy dimensions, and the second layer of dimensions comprise manufacturers, models, key parameters, historical usage of the unit and industry usage; And continuously acquiring related