Question and answer interaction method and system

Abstract

The invention provides a question and answer interaction method and system.The method comprises the steps that when a query instruction sent by a user is received in real time, the query instruction is analyzed so as to detect a corresponding query target in real time; performing pre-retrieval processing on the query target to retrieve a corresponding initial statement in real time; converting the initial statement into a corresponding target vector in real time through a preset conversion model, and correspondingly storing the target vector into a preset vector library; and performing retrieval post-processing on the target vector in the preset vector library to generate a corresponding target statement in real time, and generating a corresponding answer in real time according to the target statement. According to the invention, whether the motor is overloaded or not can be accurately judged, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a question-answering interaction method and system. Background Art

[0002] With the rise of big models, AI technology research and application expansion has ushered in a "wave". With the amazing understanding, generation and reasoning capabilities of big models, it has gradually penetrated into various industries, including finance, medical care, games, advertising, manufacturing and other industries. Regardless of the industry, there is a general demand for unified knowledge base sedimentation and knowledge question and answer within the enterprise. Combining AI's general big model capabilities and RAG retrieval enhancement capabilities, building an enterprise-level question and answer retrieval enhancement agent can effectively solve the problems faced by traditional knowledge question and answer, such as "high time cost of manual maintenance of knowledge base" and "limited scope of question and answer", and realize automatic analysis and classification of structured and unstructured data, and provide concise, fast and accurate answers.

[0003] Among them, based on the large model natural language processing technology and primary RAG technology, it is first necessary to obtain structured or unstructured data from the external knowledge database. The primary RAG technology divides the text into multiple blocks, and uses the Embedding model to embed these blocks into vectors, and store these vectors in the vector database and establish an index; secondly, if the user's query falls into the domain that requires a specified knowledge base to answer, the system will retrieve the relevant context in the vector database; finally, the context and retrieval results will be packaged together to generate prompts for the large model, and guide the large model to generate answers.

[0004] Furthermore, the existing technology has problems with retrieval quality and response generation quality, among which the main problem of retrieval quality is low precision, that is, the retrieved document blocks are not necessarily relevant to the query content, which may lead to incorrect or incoherent information. Secondly, low recall rate is also a problem, that is, failure to retrieve all relevant document blocks, resulting in the large model being unable to obtain sufficient background information to generate accurate answers for the enterprise. In addition, the most prominent problem is the large model hallucination, that is, when the large model lacks sufficient knowledge, the generated answers may be false. Another problem is the lack of relevance of the answers, that is, the answers generated by the model fail to effectively respond to the query questions. Summary of the invention

[0005] Based on this, the purpose of the present invention is to provide a question-answering interaction method and system to solve the problems of low retrieval quality and low retrieval accuracy in the prior art.

[0006] The first aspect of the embodiment of the present invention proposes:

[0007] A question-answering interactive method, wherein the method comprises:

[0008] When receiving a query instruction sent by a user in real time, the query instruction is parsed and processed to detect the corresponding query target in real time;

[0009] Performing pre-retrieval processing on the query target to retrieve the corresponding initial statement in real time;

[0010] The initial sentence is converted into a corresponding target vector in real time through a preset conversion model, and the target vector is correspondingly stored in a preset vector library;

[0011] The target vector in the preset vector library is retrieved and processed to generate a corresponding target sentence in real time, and a corresponding answer is generated according to the target sentence in real time.

[0012] The beneficial effects of the present invention are: large models are combined with advanced RAGs to enhance large models by integrating external knowledge retrieval, thereby improving the accuracy, relevance and information coverage of generation. Advances in the retriever improve the performance of long documents and keyword queries in vocabulary, semantics and contextual relevance matching. The reranking architecture focuses on accurately predicting relevance, using a strategy that combines pre-training, model size and network architecture. The generator seamlessly integrates external evidence through truncation, extraction, weighting and memorization. The hybrid RAG system combines large models together to maximize quality while optimizing the scalability and throughput of large models.

[0013] Furthermore, the step of performing pre-retrieval processing on the query target to retrieve the corresponding initial statement in real time includes:

[0014] When the query target is acquired in real time, a target character string corresponding to the query target is detected in real time;

[0015] Performing a full scan on the target character string to extract corresponding keywords in real time, and generating corresponding index data in real time according to the keywords;

[0016] The corresponding metadata is added to the index data in real time to generate the initial statement accordingly.

[0017] Furthermore, the step of adding corresponding metadata to the index data in real time to generate the initial statement accordingly includes:

[0018] When the index data is acquired in real time, the index data is fully scanned to detect the corresponding index chain in real time;

[0019] Detecting in real time a target format corresponding to the index chain, wherein the target format is unique;

[0020] The initial statement is generated according to the index chain and the target format.

[0021] Furthermore, the step of generating the initial statement according to the index chain and the target format includes:

[0022] When the target format is acquired in real time, the corresponding algorithm database is called out in real time;

[0023] A target conversion algorithm corresponding to the target format is matched in real time in the algorithm database, and the index chain is converted into the initial statement in real time through the target conversion algorithm.

[0024] Furthermore, the step of performing post-processing on the target vector in the preset vector library to generate the corresponding target sentence in real time includes:

[0025] Extracting the corresponding vector set contained in the preset vector library in real time, and performing modular processing on the vector set to generate a number of text blocks accordingly;

[0026] The plurality of text blocks are parsed in real time to generate the target sentences accordingly.

[0027] Furthermore, the step of performing real-time parsing processing on the plurality of text blocks to generate the target sentence accordingly includes:

[0028] When a plurality of text blocks are acquired in real time, summarizing, paraphrasing and extracting key points of each text block in turn;

[0029] The processed condensed information is transmitted to the preset LLM model in real time to output the target sentence accordingly.

[0030] Furthermore, the step of transmitting the processed condensed information to the preset LLM model in real time to output the target sentence accordingly includes:

[0031] When the concentrated information is acquired in real time, the DTW algorithm contained in the preset LLM model is called out in real time;

[0032] The condensed information is converted by the DTW algorithm to generate the target sentence accordingly;

[0033] The condensed information and the target sentence are both unique.

[0034] The second aspect of the embodiment of the present invention proposes:

[0035] A question-answering interactive system, wherein the system comprises:

[0036] The parsing module is used to parse and process the query instruction sent by the user when receiving the query instruction in real time, so as to detect the corresponding query target in real time;

[0037] A first retrieval module, used to perform pre-retrieval processing on the query target to retrieve the corresponding initial sentence in real time;

[0038] A conversion module, used to convert the initial sentence into a corresponding target vector in real time through a preset conversion model, and store the target vector in a preset vector library;

[0039] The second retrieval module is used to perform post-retrieval processing on the target vector in the preset vector library to generate a corresponding target sentence in real time, and generate a corresponding answer according to the target sentence in real time.

[0040] Furthermore, the first retrieval module is specifically used for:

[0041] When the query target is acquired in real time, a target character string corresponding to the query target is detected in real time;

[0042] Performing a full scan on the target character string to extract corresponding keywords in real time, and generating corresponding index data in real time according to the keywords;

[0043] The corresponding metadata is added to the index data in real time to generate the initial statement accordingly.

[0044] Furthermore, the first retrieval module is specifically used for:

[0045] When the index data is acquired in real time, the index data is fully scanned to detect the corresponding index chain in real time;

[0046] Detecting in real time a target format corresponding to the index chain, wherein the target format is unique;

[0047] The initial statement is generated according to the index chain and the target format.

[0048] Furthermore, the first retrieval module is specifically used for:

[0049] When the target format is acquired in real time, the corresponding algorithm database is called out in real time;

[0050] A target conversion algorithm corresponding to the target format is matched in real time in the algorithm database, and the index chain is converted into the initial statement in real time through the target conversion algorithm.

[0051] Furthermore, the second retrieval module is specifically used for:

[0052] Extracting the corresponding vector set contained in the preset vector library in real time, and performing modular processing on the vector set to generate a number of text blocks accordingly;

[0053] The plurality of text blocks are parsed in real time to generate the target sentences accordingly.

[0054] Furthermore, the second retrieval module is specifically used for:

[0055] When a plurality of text blocks are acquired in real time, summarizing, paraphrasing and extracting key points of each text block in turn;

[0056] The processed condensed information is transmitted to the preset LLM model in real time to output the target sentence accordingly.

[0057] Furthermore, the second retrieval module is specifically used for:

[0058] When the concentrated information is acquired in real time, the DTW algorithm contained in the preset LLM model is called out in real time;

[0059] The condensed information is converted by the DTW algorithm to generate the target sentence accordingly;

[0060] The condensed information and the target sentence are both unique.

[0061] The third aspect of the embodiment of the present invention proposes:

[0062] A computer comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the question-answer interaction method as described above when executing the computer program.

[0063] The fourth aspect of the embodiments of the present invention proposes:

[0064] A readable storage medium stores a computer program, wherein the program, when executed by a processor, implements the question-answering interaction method as described above.

[0065] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A flowchart of a question-answering interactive method provided by a first embodiment of the present invention;

[0067] Figure 2 This is a structural block diagram of a question-answering interactive system provided in the third embodiment of the present invention.

[0068] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0069] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0070] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0071] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0072] See also Figure 1 , which shows the question-answer interaction method provided in the first embodiment of the present invention. The question-answer interaction method provided in this embodiment can accurately retrieve answers that meet user needs, thereby correspondingly improving the user experience.

[0073] Specifically, this embodiment provides:

[0074] A question-answering interactive method specifically comprises the following steps:

[0075] Step S10, when a query instruction sent by a user is received in real time, the query instruction is parsed and processed to detect a corresponding query target in real time;

[0076] Step S20, performing pre-retrieval processing on the query target to retrieve the corresponding initial statement in real time;

[0077] Step S30, converting the initial sentence into a corresponding target vector in real time through a preset conversion model, and storing the target vector in a preset vector library;

[0078] Step S40, performing post-retrieval processing on the target vector in the preset vector library to generate a corresponding target sentence in real time, and generating a corresponding answer according to the target sentence in real time.

[0079] Second embodiment

[0080] Furthermore, the step of performing pre-retrieval processing on the query target to retrieve the corresponding initial statement in real time includes:

[0081] When the query target is acquired in real time, a target character string corresponding to the query target is detected in real time;

[0082] Performing a full scan on the target character string to extract corresponding keywords in real time, and generating corresponding index data in real time according to the keywords;

[0083] The corresponding metadata is added to the index data in real time to generate the initial statement accordingly.

[0084] Furthermore, the step of adding corresponding metadata to the index data in real time to generate the initial statement accordingly includes:

[0085] When the index data is acquired in real time, the index data is fully scanned to detect the corresponding index chain in real time;

[0086] Detecting in real time a target format corresponding to the index chain, wherein the target format is unique;

[0087] The initial statement is generated according to the index chain and the target format.

[0088] Furthermore, the step of generating the initial statement according to the index chain and the target format includes:

[0089] When the target format is acquired in real time, the corresponding algorithm database is called out in real time;

[0090] A target conversion algorithm corresponding to the target format is matched in real time in the algorithm database, and the index chain is converted into the initial statement in real time through the target conversion algorithm.

[0091] Furthermore, the step of performing post-processing on the target vector in the preset vector library to generate the corresponding target sentence in real time includes:

[0092] Extracting the corresponding vector set contained in the preset vector library in real time, and performing modular processing on the vector set to generate a number of text blocks accordingly;

[0093] The plurality of text blocks are parsed in real time to generate the target sentences accordingly.

[0094] Furthermore, the step of performing real-time parsing processing on the plurality of text blocks to generate the target sentence accordingly includes:

[0095] When a plurality of text blocks are acquired in real time, summarizing, paraphrasing and extracting key points of each text block in turn;

[0096] The processed condensed information is transmitted to the preset LLM model in real time to output the target sentence accordingly.

[0097] Furthermore, the step of transmitting the processed condensed information to the preset LLM model in real time to output the target sentence accordingly includes:

[0098] When the concentrated information is acquired in real time, the DTW algorithm contained in the preset LLM model is called out in real time;

[0099] The condensed information is converted by the DTW algorithm to generate the target sentence accordingly;

[0100] The condensed information and the target sentence are both unique.

[0101] The pre-search technology provided in this embodiment specifically includes the following contents:

[0102] The first technique is to improve the quality of the index data. In reality, we cannot expect to put every document (regardless of its relevance) into the vector database and move directly to optimization. To improve the quality of the index data, the following measures should be taken:

[0103] 1) Delete text or documents that are not relevant to the specific task;

[0104] 2) reformat the index data into a form similar to that which the end user might use;

[0105] 3) Add metadata to documents to enable efficient and targeted retrieval.

[0106] The second technique is block optimization. Depending on the requirements of the downstream task, the optimal length of each block and its degree of overlap need to be determined. If the block length is too small, it may not contain all the information required by the large model to answer the user's query; if the block is too large, it may contain too much irrelevant information, causing the large model to be confused or exceed the context limit. It is not necessary to stick to a single block optimization method in all steps of the pipeline.

[0107] The third technique is to rewrite the user query before matching it in the vector database. The key to this step is to convert the user's query into a form that is similar in format and content to the query in the vector database. The Query2Doc technique generates pseudo documents and uses these documents to expand the query. Similarly, the HyDE (hypothetical document embedding) technique generates hypothetical documents related to the query.

[0108] The post-retrieval technology provided in this embodiment specifically includes the following contents:

[0109] After retrieving the relevant chunks from the database, there are multiple techniques to further improve the quality of the generation. Depending on the nature of the task and the format of the text chunks, one or more of the following techniques can be employed.

[0110] If the task is more relevant to a particular chunk, a common technique is to re-rank or re-score the chunks. As mentioned earlier, a high score in the vector similarity search does not necessarily mean the highest relevance. Therefore, a second round of re-ranking or scoring should be performed to find out the text chunks that are really useful for generating the answer. During the re-ranking or scoring process, the big model can be asked to rank the documents for relevance, or other methods such as keyword frequency analysis or metadata matching can be used to further refine the selection before passing these documents to the big model to generate the final answer.

[0111] If the task involves multiple blocks, such as summarization or comparison, information compression can be performed as post-processing before passing it to LLM to reduce noise or context length. For example, each block can be summarized, paraphrased, or key points can be extracted first, and then the aggregated condensed information can be passed to LLM for generation.

[0112] This embodiment also provides an optimization technology, which specifically includes the following contents:

[0113] In advanced RAG, in addition to optimizing the organization of knowledge and recall strategies, optimization of the embedding model is also crucial. First, fine-tuning the model directly affects the effectiveness of RAG. Training the embedding model with domain-specific data is essential to improve the relevance between user queries and knowledge. Second, dynamic embedding results vary depending on the context in which the word appears. For example, in the Transformer model, the embedding results of the same word will vary depending on the surrounding vocabulary.

[0114] See also Figure 2 , the third embodiment of the present invention provides:

[0115] A question-answering interactive system, wherein the system comprises:

[0116] The parsing module is used to parse and process the query instruction sent by the user when receiving the query instruction in real time, so as to detect the corresponding query target in real time;

[0117] A first retrieval module, used to perform pre-retrieval processing on the query target to retrieve the corresponding initial sentence in real time;

[0118] A conversion module, used to convert the initial sentence into a corresponding target vector in real time through a preset conversion model, and store the target vector in a preset vector library;

[0119] The second retrieval module is used to perform post-retrieval processing on the target vector in the preset vector library to generate a corresponding target sentence in real time, and generate a corresponding answer according to the target sentence in real time.

[0120] Furthermore, the first retrieval module is specifically used for:

[0121] When the query target is acquired in real time, a target character string corresponding to the query target is detected in real time;

[0122] Performing a full scan on the target character string to extract corresponding keywords in real time, and generating corresponding index data in real time according to the keywords;

[0123] The corresponding metadata is added to the index data in real time to generate the initial statement accordingly.

[0124] Furthermore, the first retrieval module is specifically used for:

[0125] When the index data is acquired in real time, the index data is fully scanned to detect the corresponding index chain in real time;

[0126] Detecting in real time a target format corresponding to the index chain, wherein the target format is unique;

[0127] The initial statement is generated according to the index chain and the target format.

[0128] Furthermore, the first retrieval module is specifically used for:

[0129] When the target format is acquired in real time, the corresponding algorithm database is called out in real time;

[0130] A target conversion algorithm corresponding to the target format is matched in real time in the algorithm database, and the index chain is converted into the initial statement in real time through the target conversion algorithm.

[0131] Furthermore, the second retrieval module is specifically used for:

[0132] Extracting the corresponding vector set contained in the preset vector library in real time, and performing modular processing on the vector set to generate a number of text blocks accordingly;

[0133] The plurality of text blocks are parsed in real time to generate the target sentences accordingly.

[0134] Furthermore, the second retrieval module is specifically used for:

[0135] When a plurality of text blocks are acquired in real time, summarizing, paraphrasing and extracting key points of each text block in turn;

[0136] The processed condensed information is transmitted to the preset LLM model in real time to output the target sentence accordingly.

[0137] Furthermore, the second retrieval module is specifically used for:

[0138] When the concentrated information is acquired in real time, the DTW algorithm contained in the preset LLM model is called out in real time;

[0139] The condensed information is converted by the DTW algorithm to generate the target sentence accordingly;

[0140] The condensed information and the target sentence are both unique.

[0141] A fourth embodiment of the present invention provides a computer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the question-and-answer interaction method described above when executing the computer program.

[0142] A fifth embodiment of the present invention provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the question-and-answer interaction method as described above.

[0143] In summary, the question-answer interaction method and system provided by the above embodiments of the present invention can accurately retrieve answers that meet user needs, thereby correspondingly improving the user experience.

[0144] It should be noted that the above modules can be functional modules or program modules, and can be implemented by software or hardware. For modules implemented by hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0145] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0146] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0147] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0148] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0149] The above-described embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A question-answering interactive method, characterized in that: The method comprises: When receiving a query instruction sent by a user in real time, the query instruction is parsed and processed to detect the corresponding query target in real time; Performing pre-retrieval processing on the query target to retrieve the corresponding initial statement in real time; The initial sentence is converted into a corresponding target vector in real time through a preset conversion model, and the target vector is correspondingly stored in a preset vector library; The target vector in the preset vector library is retrieved and processed to generate a corresponding target sentence in real time, and a corresponding answer is generated according to the target sentence in real time.

2. The question-answering interactive method according to claim 1, characterized in that: The step of performing pre-retrieval processing on the query target to retrieve the corresponding initial statement in real time includes: When the query target is acquired in real time, a target character string corresponding to the query target is detected in real time; Performing a full scan on the target character string to extract corresponding keywords in real time, and generating corresponding index data in real time according to the keywords; The corresponding metadata is added to the index data in real time to generate the initial statement accordingly.

3. The question-answering interactive method according to claim 2, characterized in that: The step of adding corresponding metadata to the index data in real time to generate the initial statement accordingly includes: When the index data is acquired in real time, the index data is fully scanned to detect the corresponding index chain in real time; Detecting in real time a target format corresponding to the index chain, wherein the target format is unique; The initial statement is generated according to the index chain and the target format.

4. The question-answering interactive method according to claim 3, characterized in that: The step of generating the initial statement according to the index chain and the target format includes: When the target format is acquired in real time, the corresponding algorithm database is called out in real time; A target conversion algorithm corresponding to the target format is matched in real time in the algorithm database, and the index chain is converted into the initial statement in real time through the target conversion algorithm.

5. The question-answering interactive method according to claim 1, characterized in that: The step of performing post-processing on the target vector in the preset vector library to generate the corresponding target sentence in real time includes: Extracting the corresponding vector set contained in the preset vector library in real time, and performing modular processing on the vector set to generate a number of text blocks accordingly; The plurality of text blocks are parsed in real time to generate the target sentences accordingly.

6. The question-answering interactive method according to claim 5, characterized in that: The step of performing real-time parsing processing on the plurality of text blocks to generate the target sentence accordingly comprises: When a plurality of text blocks are acquired in real time, summarizing, paraphrasing and extracting key points of each text block in turn; The processed condensed information is transmitted to the preset LLM model in real time to output the target sentence accordingly.

7. The question-answering interactive method according to claim 6, characterized in that: The step of transmitting the processed condensed information to the preset LLM model in real time to output the target sentence accordingly includes: When the concentrated information is acquired in real time, the DTW algorithm contained in the preset LLM model is called out in real time; The condensed information is converted by the DTW algorithm to generate the target sentence accordingly; The condensed information and the target sentence are both unique.

8. A question-answering interactive system, characterized in that: The system comprises: The parsing module is used to parse and process the query instruction sent by the user when receiving the query instruction in real time, so as to detect the corresponding query target in real time; A first retrieval module, used to perform pre-retrieval processing on the query target to retrieve the corresponding initial sentence in real time; A conversion module, used to convert the initial sentence into a corresponding target vector in real time through a preset conversion model, and store the target vector in a preset vector library; The second retrieval module is used to perform post-retrieval processing on the target vector in the preset vector library to generate a corresponding target sentence in real time, and generate a corresponding answer according to the target sentence in real time.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the question-answer interaction method according to any one of claims 1 to 7 is implemented.

10. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the question-answering interaction method as described in any one of claims 1 to 7 is implemented.

Images