Search

CN-122019569-A - Intelligent searching grouping method, system, equipment and medium for large language model

CN122019569ACN 122019569 ACN122019569 ACN 122019569ACN-122019569-A

Abstract

The invention relates to the technical field of Internet retrieval, and particularly discloses a large language model intelligent search grouping method, a system, equipment and a medium, wherein the method comprises the steps of obtaining a search request submitted by a user and extracting search state characteristics for representing the search state of the user; the method comprises the steps of determining a preset number of search objects related to a search request based on the search state characteristics, synchronously generating description information of each search object, constructing a search operation matrix, calculating significance parameters of each search object at the current moment based on the search operation matrix, updating grouping weights of each search object, determining search processing resource amounts of each search object according to the grouping weights, calling a large language model based on the search processing resource amounts, generating and feeding back search results, and outputting differentiated search contents aiming at different search objects.

Inventors

  • ZHANG YAN

Assignees

  • 即梦计算机(北京)有限公司

Dates

Publication Date
20260512
Application Date
20260129

Claims (10)

  1. 1. A large language model intelligent search grouping method, the method comprising: Acquiring a search request submitted by a user, acquiring historical search behavior data corresponding to the search request, and extracting features of the search request and the historical search behavior data to obtain search state features for representing a search state of the user; Determining a preset number of search objects related to the search request based on the search state characteristics, and synchronously generating description information of each search object, wherein the description information at least comprises semantic characteristics and behavior characteristics of the search object; Acquiring behavior data of each search object at different time points based on a preset time period, and constructing a search operation matrix by taking time as a row and behavior characteristics as columns, wherein the search operation matrix is used for representing the change condition of the behavior of the search object along with time; Calculating the significance parameters of each search object at the current moment based on the search operation matrix, and updating the grouping weight of each search object according to the significance parameters; And determining the search processing resource quantity of each search object according to the grouping weight, calling a large language model based on the search processing resource quantity, and generating and feeding back a search result, wherein the search processing resource quantity is at least used for adjusting search expansion depth and search priority.
  2. 2. The method for intelligent search grouping of large language models as set forth in claim 1, wherein the step of obtaining search requests submitted by users, obtaining historical search behavior data corresponding to the search requests, and extracting features of the search requests and the historical search behavior data to obtain search state features for characterizing a search state of the users comprises: acquiring search requests submitted by users and historical search behavior data; performing text pretreatment on the search request and the historical search behavior data to obtain word groups, wherein the text pretreatment at least comprises word segmentation, denoising and keyword extraction; Converting word groups corresponding to the search request to obtain semantic vectors, and converting word groups corresponding to the historical search behavior data to obtain behavior characteristics; Counting semantic vectors and behavior characteristics as search state characteristics; the semantic vector and the behavior feature are text converted into a numerical format.
  3. 3. The method of claim 1, wherein the step of determining a predetermined number of search objects related to the search request based on the search status features, and generating description information of each search object synchronously comprises: Recalling a preset number of candidate search objects in a preset knowledge base according to the search state characteristics; extracting the content semantic features of each candidate search object; extracting historical behavior characteristics of the search object from the search log; and counting semantic features and behavior features to obtain the description information of the search object.
  4. 4. The method for intelligent search grouping of large language models according to claim 1, wherein the step of collecting behavior data of each search object at different time points based on a preset time period, and constructing a search operation matrix by taking time as a row and taking behavior features as columns comprises the steps of: receiving a behavior feature set which is input by a management party and used for describing the behavior of a search object; Acquiring behavior data of each search object at different time points based on a preset time period, wherein the data type of the behavior data is contained in a behavior feature set; constructing a matrix template by taking time as a row mark and taking behavior characteristics as a column mark; And filling the behavior data at different time points into a matrix template to obtain a search operation matrix.
  5. 5. The method for intelligent search grouping of large language models as set forth in claim 1, wherein the step of calculating a saliency parameter of each search object at a current time based on the search operation matrix, and updating the grouping weight of each search object according to the saliency parameter comprises: intercepting a submatrix containing the current moment from the search operation matrix; Normalizing the behavior characteristics in the sub-matrix; Weighting calculation is carried out on each behavior characteristic according to preset weights, and comprehensive scores are obtained; Taking the comprehensive score as a significance parameter of the search object at the current moment; and updating the weight value of the search object in the grouping process according to the significance parameter.
  6. 6. The method of claim 5, wherein the determining the amount of search processing resources for each search object according to the grouping weight, invoking the large language model based on the amount of search processing resources, and generating and feeding back the search result comprises: Analyzing the behavior change of each search object in different time periods based on the search operation matrix, and calculating the behavior deviation degree; Taking the behavior deviation degree as a correction factor of the grouping weight, and dynamically correcting the grouping weight to obtain corrected grouping weight; Determining the search processing resource quantity of each search object according to the corrected grouping weight, calling a large language model based on the search processing resource quantity, and generating and feeding back a search result; wherein the behavior deviation is used for representing the degree of abnormality of the behavior change of the search object relative to the historical average behavior.
  7. 7. A large language model intelligent search grouping system, said system comprising: the search feature extraction module is used for acquiring a search request submitted by a user, acquiring historical search behavior data corresponding to the search request, and extracting features of the search request and the historical search behavior data to obtain search state features for representing the search state of the user; The search object analysis module is used for determining a preset number of search objects related to a search request based on the search state characteristics and synchronously generating description information of each search object, wherein the description information at least comprises semantic characteristics and behavior characteristics of the search objects; the system comprises a search behavior analysis module, a search operation matrix and a search module, wherein the search behavior analysis module is used for acquiring behavior data of each search object at different time points based on a preset time period, taking time as a row and taking behavior characteristics as columns; the grouping weight determining module is used for calculating the significance parameter of each search object at the current moment based on the search operation matrix and updating the grouping weight of each search object according to the significance parameter; And the search result feedback module is used for determining the search processing resource quantity of each search object according to the grouping weight, calling a large language model based on the search processing resource quantity, and generating and feeding back a search result, wherein the search processing resource quantity is at least used for adjusting the search expansion depth and the search priority.
  8. 8. The large language model intelligent search grouping system of claim 7, wherein said search feature extraction module comprises: the data acquisition unit is used for acquiring search requests submitted by users and historical search behavior data; the preprocessing unit is used for carrying out text preprocessing on the search request and the historical search behavior data to obtain word groups, wherein the text preprocessing at least comprises word segmentation, denoising and keyword extraction; The text conversion unit is used for converting word groups corresponding to the search request to obtain semantic vectors, and converting word groups corresponding to the historical search behavior data to obtain behavior characteristics; the first feature statistics unit is used for counting semantic vectors and behavior features and is used as search state features; the semantic vector and the behavior feature are text converted into a numerical format.
  9. 9. A computer device comprising one or more processors and one or more memories, the one or more memories having stored therein at least one program code that, when loaded and executed by the one or more processors, implements the large language model smart search grouping method of claims 1-6.
  10. 10. A storage medium having stored therein at least one program code which, when loaded and executed by a processor, implements the large language model smart search grouping method of claims 1-6.

Description

Intelligent searching grouping method, system, equipment and medium for large language model Technical Field The invention relates to the technical field of Internet retrieval, in particular to a large language model intelligent search grouping method, a system, equipment and a medium. Background Under the internet technical background of big data age, the search function is the basic function of most internet services, when the existing search system responds to the search request of a user, the search request is generally uniformly analyzed by taking a single search target as the center, and a linearly ordered search result list is returned, or simple topic classification is only carried out in the result display stage, the search mode is difficult to simultaneously describe the differences among a plurality of potential search objects in one search process, the user can gradually acquire search feedback of different directions or different emphasis points through multiple searches or repeatedly adjusting search conditions, and the search interaction efficiency is low. In a complex search scenario, the same search request often corresponds to multiple semantic directions or potential search objects, for example, different product types, different usage scenarios or different information levels, and how to expand the feedback results of the search requirements to meet the requirements of users on multi-dimensional and multi-view search feedback is a technical problem to be solved by the technical scheme of the invention. Disclosure of Invention The invention aims to provide a large language model intelligent search grouping method, a system, equipment and a medium, which are used for solving the problems in the background technology. In order to achieve the above purpose, the present invention provides the following technical solutions: a large language model intelligent search grouping method, the method comprising: Acquiring a search request submitted by a user, acquiring historical search behavior data corresponding to the search request, and extracting features of the search request and the historical search behavior data to obtain search state features for representing a search state of the user; Determining a preset number of search objects related to the search request based on the search state characteristics, and synchronously generating description information of each search object, wherein the description information at least comprises semantic characteristics and behavior characteristics of the search object; Acquiring behavior data of each search object at different time points based on a preset time period, and constructing a search operation matrix by taking time as a row and behavior characteristics as columns, wherein the search operation matrix is used for representing the change condition of the behavior of the search object along with time; Calculating the significance parameters of each search object at the current moment based on the search operation matrix, and updating the grouping weight of each search object according to the significance parameters; And determining the search processing resource quantity of each search object according to the grouping weight, calling a large language model based on the search processing resource quantity, and generating and feeding back a search result, wherein the search processing resource quantity is at least used for adjusting search expansion depth and search priority. The method comprises the steps of obtaining search requests submitted by users, obtaining historical search behavior data corresponding to the search requests, extracting features of the search requests and the historical search behavior data, and obtaining search state features used for representing search states of the users, wherein the steps comprise: acquiring search requests submitted by users and historical search behavior data; performing text pretreatment on the search request and the historical search behavior data to obtain word groups, wherein the text pretreatment at least comprises word segmentation, denoising and keyword extraction; Converting word groups corresponding to the search request to obtain semantic vectors, and converting word groups corresponding to the historical search behavior data to obtain behavior characteristics; Counting semantic vectors and behavior characteristics as search state characteristics; the semantic vector and the behavior feature are text converted into a numerical format. The invention further provides that the step of determining a preset number of search objects related to the search request based on the search state features and synchronously generating the description information of each search object comprises the following steps: Recalling a preset number of candidate search objects in a preset knowledge base according to the search state characteristics; extracting the content semantic features of each candidate search obj