Search

CN-121997924-A - Text processing system based on heuristic interaction

CN121997924ACN 121997924 ACN121997924 ACN 121997924ACN-121997924-A

Abstract

The application provides a text processing system based on heuristic interaction, which comprises the steps of obtaining information to be processed and corresponding characteristic information, and processing the information to be processed and the characteristic information by utilizing a target model to obtain target processing information.

Inventors

  • MAO JUNFENG

Assignees

  • 深圳市TCL高新技术开发有限公司

Dates

Publication Date
20260508
Application Date
20241107

Claims (13)

  1. 1. A method, the method comprising: acquiring information to be processed and corresponding characteristic information; And processing the information to be processed and the characteristic information by using a target model to obtain target processing information.
  2. 2. The method according to claim 1, wherein the feature information includes multi-modal information, the processing the information to be processed and the feature information with a target model to obtain target processing information includes: Generating multi-mode input information according to the information to be processed and the multi-mode information; And processing the multi-mode input information based on the target model to obtain target processing information.
  3. 3. The method of claim 2, wherein the object model comprises a first attention layer and a second attention layer, the first attention layer comprising a first attention unit and a first adapter unit, the second attention layer comprising a second attention unit and a second adapter unit, wherein a first input of the first attention unit is configured as an input of the object model, a first output of the first attention unit is connected with an input of the first adapter unit, a second output of the first attention unit and an output of the first adapter unit is connected with an input of the second attention unit, an input of the second adapter unit is connected with a first output of the second attention unit, and an output of the second adapter unit is configured as an output of the object model.
  4. 4. The method of claim 3, wherein processing the multimodal input information based on the target model to obtain target processing information comprises: Outputting first processing information corresponding to the multi-mode input information by utilizing the target model; and performing feature processing on the first processing information based on the target word set to obtain target processing information.
  5. 5. The method of claim 4, wherein outputting the first processing information corresponding to the multimodal input information using the object model comprises: Processing the multi-mode input information based on the first attention layer to obtain first characteristic information; and processing the first characteristic information based on the second attention layer to obtain the first processing information.
  6. 6. The method of claim 5, wherein processing the multimodal input information based on the first attention layer to obtain first feature information comprises: Inputting the multi-mode input information into the first attention unit to obtain first output information and second output information; And inputting the first output information into the first adapter unit to obtain first adapter output information, and fusing the first adapter output information and the second output information to obtain first characteristic information.
  7. 7. The method of claim 4, wherein the performing feature processing on the first processing information based on the target word set to obtain target processing information includes: Calculating a target word similarity table of the information to be processed by using a target word set; Calculating bias information of the information to be processed based on the word set number of the target word set, the dictionary number and the target word similarity table, wherein the bias information is used for representing the similarity degree of target words in the target word set and the information to be processed; and obtaining target processing information based on the bias information and the first processing information.
  8. 8. A method, the method comprising: Acquiring multi-mode training information; and carrying out parameter adjustment on the initial model based on the multi-mode training information and the adapter parameters to obtain a target model.
  9. 9. The method of claim 8, wherein the initial model comprises a first attention layer and a second attention layer, the first attention layer comprising a first attention unit and a first adapter unit, the second attention layer comprising a second attention unit and a second adapter unit; The parameter adjustment is performed on the initial model based on the multi-mode training information and the adapter parameter to obtain a target model, which comprises the following steps: extracting the characteristics of the multi-modal training information to obtain multi-modal characteristic information; Acquiring first model parameters in the first attention layer and/or the second attention layer and second model parameters of the second attention layer; performing parameter adjustment on the first model parameters based on the multi-mode characteristic information to obtain target model parameters; and carrying out parameter adjustment on the initial model based on the adapter parameters of the target adapter, the target model parameters and the second model parameters to generate a target model.
  10. 10. The method of claim 9, wherein the parameter adjusting the initial model based on the adapter parameters of the target adapter, the target model parameters, and the second model parameters to generate the target model comprises: Fusing the adapter parameters and the target model parameters to obtain first initial adjustment parameters, and fusing the adapter parameters and the second model parameters to obtain second initial adjustment parameters; updating the adapter parameters based on the first initial adjustment parameters and the second initial adjustment parameters to obtain target adapter parameters; Combining the target adapter parameters with the target model parameters to obtain first target adjustment parameters, and combining the target adapter parameters with the second model parameters to obtain second target adjustment parameters; and carrying out parameter adjustment on the initial model according to the first target adjustment parameter and the second target adjustment parameter to obtain a target model.
  11. 11. A system, the system comprising: the acquisition module is used for acquiring the information to be processed and the corresponding characteristic information; The interaction module is used for processing the information to be processed and the characteristic information by utilizing a target model to obtain target processing information; Further, the feature information includes multi-modal information, and the processing the information to be processed and the feature information by using a target model to obtain target processing information includes: Generating multi-mode input information according to the information to be processed and the multi-mode information; Processing the multi-mode input information based on the target model to obtain target processing information; Further, the object model comprises a first attention layer and a second attention layer, the first attention layer comprises a first attention unit and a first adapter unit, the second attention layer comprises a second attention unit and a second adapter unit, wherein a first input end of the first attention unit is configured as an input end of the object model, a first output end of the first attention unit is connected with an input end of the first adapter unit, a second output end of the first attention unit and an output end of the first adapter unit are connected with an input end of the second attention unit, an input end of the second adapter unit is connected with a first output end of the second attention unit, and a second output end of the second attention unit and an output end of the second adapter unit are configured as output ends of the object model; further, the processing the multimodal input information based on the target model to obtain target processing information includes: Outputting first processing information corresponding to the multi-mode input information by utilizing the target model; Performing feature processing on the first processing information based on a target word set to obtain target processing information; further, the outputting, by using the target model, the first processing information corresponding to the multimodal input information includes: Processing the multi-mode input information based on the first attention layer to obtain first characteristic information; processing the first characteristic information based on the second attention layer to obtain first processing information; further, the processing the multimodal input information based on the first attention layer to obtain first feature information includes: Inputting the multi-mode input information into the first attention unit to obtain first output information and second output information; Inputting the first output information into the first adapter unit to obtain first adapter output information, and fusing the first adapter output information and the second output information to obtain first characteristic information; Further, the feature processing is performed on the first processing information based on the target word set to obtain target processing information, including: Calculating a target word similarity table of the information to be processed by using a target word set; Calculating bias information of the information to be processed based on the word set number of the target word set, the dictionary number and the target word similarity table, wherein the bias information is used for representing the similarity degree of target words in the target word set and the information to be processed; Obtaining target processing information based on the bias information and the first processing information; the interaction module is further configured to perform parameter adjustment on an initial model based on the multi-modal training information and adapter parameters to obtain a target model; Further, the initial model includes a first attention layer including a first attention unit and a first adapter unit and a second attention layer including a second attention unit and a second adapter unit; The parameter adjustment is performed on the initial model based on the multi-mode training information and the adapter parameter to obtain a target model, which comprises the following steps: extracting the characteristics of the multi-modal training information to obtain multi-modal characteristic information; Acquiring first model parameters in the first attention layer and/or the second attention layer and second model parameters of the second attention layer; performing parameter adjustment on the first model parameters based on the multi-mode characteristic information to obtain target model parameters; Performing parameter adjustment on the initial model based on the adapter parameters, the target model parameters and the second model parameters of the target adapter to generate a target model; Further, the performing parameter adjustment on the initial model based on the adapter parameter of the target adapter, the target model parameter and the second model parameter to generate a target model includes: Fusing the adapter parameters and the target model parameters to obtain first initial adjustment parameters, and fusing the adapter parameters and the second model parameters to obtain second initial adjustment parameters; updating the adapter parameters based on the first initial adjustment parameters and the second initial adjustment parameters to obtain target adapter parameters; Combining the target adapter parameters with the target model parameters to obtain first target adjustment parameters, and combining the target adapter parameters with the second model parameters to obtain second target adjustment parameters; and carrying out parameter adjustment on the initial model according to the first target adjustment parameter and the second target adjustment parameter to obtain a target model.
  12. 12. An apparatus, the apparatus comprising: One or more processors; Memory, and One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the steps of the method of any of claims 1 to 7 or 8 to 10.
  13. 13. A computer readable storage medium, having stored thereon a computer program, the computer program being loaded by a processor to perform the steps of the method of any of claims 1 to 7 or 8 to 10.

Description

Text processing system based on heuristic interaction Technical Field The application relates to the technical field of computers, in particular to a text processing system based on heuristic interaction. Background Currently, with the advent of AIGC (AI GENERATED content, artificial intelligence generation) and LLM (Large Language Model, large-scale language model) technologies, a variety of AI tools have emerged that can assist users in authoring, for example, chatGPT (CHAT GENERATIVE PRE-trained Transformer, content generation type pre-training language model), which is a language model that can respond accordingly based on content (text, image, voice, etc.) input by users. And, with the development of the LMM technology and the application technology, the user can implement question-answer interaction through the large-scale language model and input various interaction instructions into the large-scale language model, so as to call each application program to perform various interaction operations by using the large-scale language model. Disclosure of Invention The embodiment of the application provides a text processing system based on heuristic interaction. In a first aspect, an embodiment of the text processing system of the present application provides a method comprising the steps of: acquiring information to be processed and corresponding characteristic information; And processing the information to be processed and the characteristic information by using the target model to obtain target processing information. In a second aspect, an embodiment of the text processing system of the present application provides another method comprising the steps of: Acquiring multi-mode training information; and carrying out parameter adjustment on the initial model based on the multi-mode training information and the adapter parameters to obtain a target model. In a third aspect, the present application provides a system comprising: the acquisition module is used for acquiring the information to be processed and the corresponding characteristic information; The interaction module is used for processing the information to be processed and the characteristic information by utilizing the target model to obtain target processing information In a fourth aspect, the present application also provides an apparatus comprising: One or more processors; Memory, and One or more applications, wherein the one or more applications are stored in memory and configured to be executed by a processor to perform the steps of the methods described above. In a fifth aspect, the present application also provides a computer readable storage medium having stored thereon a computer program to be loaded by a processor for performing the steps of the methods described above. Drawings In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. FIG. 1 is a schematic view of a text processing method according to an embodiment of the present application; FIG. 2 is a flow chart of an embodiment of a text processing method according to an embodiment of the present application; FIG. 3 is a flow chart of an embodiment of model training in a text processing method according to an embodiment of the present application; FIG. 4 is a flowchart illustrating an embodiment of performing part-of-speech bias processing on first processing information in a text processing method according to an embodiment of the present application; FIG. 5 is a flow diagram of one embodiment of part-of-speech biasing provided in embodiments of the present application; FIG. 6 is a schematic diagram illustrating the architecture of one embodiment of a text processing system according to an embodiment of the present application; Fig. 7 is a schematic structural diagram of an embodiment of a text processing device according to an embodiment of the present application. Detailed Description The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to fall within the scope of the application. In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal