Conversation simplification method and device, electronic equipment and storage medium
By using a dialogue simplification method to divide historical dialogue content into question-answer pairs, and filtering related context units based on semantic similarity and rounds, the problems of computational resource consumption and response latency in long-cycle dialogues are solved, achieving high efficiency and coherence in the dialogue system.
Patent Information
- Application Number
- CN202511802006.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-02
- Publication Date
- 2026-02-24
AI Technical Summary
In long-term, multi-round dialogues, the continuous accumulation of historical context leads to problems such as increased consumption of computing resources and response latency, which are difficult to manage and optimize effectively with existing technologies.
By dividing historical dialogue content into question-answer pairs, relevant contextual units are selected based on semantic similarity and dialogue rounds. The simplified historical dialogue content is retained, and a lightweight semantic encoder is used for vectorization processing and dynamic threshold filtering to remove irrelevant or weakly relevant information.
It reduces the input length and computational overhead of the dialogue system, improves the density of contextual information and semantic focus, and ensures the coherence and response efficiency of the dialogue.
Smart Images
Figure CN121562633A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a method, apparatus, electronic device, and storage medium for simplifying dialogue. Background Technology
[0002] With the widespread application of LLM (Large Language Model) in intelligent dialogue systems, LLM-based intelligent agents have been deeply implemented in scenarios such as customer service, virtual assistants, and data analysis.
[0003] Such dialogue systems typically support coherent interactive experiences and complex task reasoning by maintaining multi-turn historical dialogue context. However, during long-term, multi-turn dialogues, the historical context accumulates continuously, causing the length of the input sequence to grow rapidly, resulting in significant computational resource consumption and response latency issues. Summary of the Invention
[0004] In view of this, the object of the present invention is to provide a simplified dialogue method, apparatus, electronic device and storage medium.
[0005] To achieve the above objectives, the technical solutions adopted in the embodiments of the present invention are as follows: In a first aspect, the present invention provides a method for simplifying dialogue, the method comprising: Retrieve new questions and historical dialogues from users in the current round; The historical dialogue content is divided into multiple question-and-answer pairs. Based on the user's new question and the dialogue round of each question-and-answer pair, the associated context unit is determined from all the question-and-answer pairs. The associated context unit is retained as the simplified historical dialogue content.
[0006] Optionally, the step of dividing the historical dialogue content into multiple question-answer pair units includes: Each question and response from each round of the historical dialogue is treated as a question-and-answer pair unit, resulting in multiple question-and-answer pair units.
[0007] Optionally, the step of determining the associated context unit from all the question-and-answer pairs based on the content of the user's new question and the dialogue round of each question-and-answer pair unit includes: The new question content of the user and each question-answer pair unit are vectorized to obtain a new question vector and multiple historical question-answer pair vectors; Based on each of the historical question-answer pair vectors and the new question vector, the semantic similarity between each question-answer pair unit and the content of the user's new question is obtained; Based on the semantic similarity between each question-answer pair and the user's new question, as well as the dialogue round of each question-answer pair, the associated context unit is determined from all the question-answer pairs.
[0008] Optionally, the step of determining the associated context unit from all question-answer pairs based on the semantic similarity between each question-answer pair unit and the user's new question content, and the dialogue round of each question-answer pair unit, includes: For each question-answer pair unit, a similarity threshold is determined based on the dialogue round corresponding to the question-answer pair unit. If the semantic similarity between the question-and-answer pair and the user's new question is not less than the similarity threshold corresponding to the question-and-answer pair, then the question-and-answer pair is determined as the associated context unit.
[0009] Optionally, the step of determining the similarity threshold corresponding to the question-answer pair unit based on the dialogue round corresponding to the question-answer pair unit includes: Obtain the turn distance between the dialogue turn corresponding to the question-answer pair unit and the current turn; Based on the round distance, the similarity threshold corresponding to the question-answer pair unit is obtained.
[0010] Optionally, the question-and-answer pair unit includes ancillary elements, which are thought chain information or tool call results associated with the question-and-answer pair unit. After determining the question-and-answer pair unit as the associated context unit, the method further includes: Based on the similarity threshold corresponding to the associated context unit, determine the similarity threshold of the subordinate elements of the associated context unit; Obtain the semantic similarity between the associated context unit's subordinate elements and the user's new question content; If the semantic similarity between the associated context unit's subordinate element and the user's new question is less than the subordinate element similarity threshold, then the subordinate element of the associated context unit will be deleted.
[0011] Optionally, the method further includes: Using the simplified historical dialogue content, the user's new question content, and the initial dialogue instructions, input content is constructed to respond to the user's new question content based on the input content, thus completing the question and answer for the current round.
[0012] Secondly, the present invention provides a dialogue simplification device, the device comprising: The acquisition module is used to acquire new questions and historical dialogue content from users in the current round. The determination module is used to divide the historical dialogue content into multiple question-and-answer pair units, and determine the associated context unit from all the question-and-answer pair units based on the new question asked by the user and the dialogue round of each question-and-answer pair unit. The simplification module is used to treat the associated context unit as simplified historical dialogue content.
[0013] Thirdly, the present invention provides an electronic device including a processor and a memory, the memory storing machine-executable instructions executable by the processor, the processor executing the machine-executable instructions to implement the dialogue simplification method described in the first aspect above.
[0014] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the dialogue simplification method as described in the first aspect above.
[0015] The present invention provides a dialogue simplification method, apparatus, electronic device, and storage medium that: acquires the content of the user's new question in the current round and the content of historical dialogue; divides the historical dialogue content into multiple question-answer pair units; determines the associated context units from all question-answer pair units based on the semantic similarity of the user's new question content and the dialogue round of each question-answer pair unit; and uses the associated context units as the simplified historical dialogue content. Because the present invention acquires the content of the user's new question in the current round and the content of historical dialogue, divides the historical dialogue content into multiple question-answer pair units, and filters out associated context units that are relevant to the current question and have a reasonable temporal order based on the current question and the dialogue round of each question-answer pair unit, retains these units and deletes the remaining irrelevant or weakly related units, thus obtaining the simplified historical dialogue content, it reduces the input length of the dialogue system, reduces computational overhead and response latency, while improving the density of contextual information and semantic focus, ensuring dialogue coherence.
[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This figure shows a schematic block diagram of an electronic device provided by an embodiment of the present invention; Figure 2 This diagram illustrates a flowchart of a dialogue simplification method provided by an embodiment of the present invention. Figure 1 ; Figure 3 This diagram illustrates a flowchart of a dialogue simplification method provided by an embodiment of the present invention. Figure 2 ; Figure 4 The diagram shows a functional block diagram of a dialogue simplification device provided in an embodiment of the present invention.
[0019] Icons: 100 - Electronic device; 110 - Memory; 120 - Processor; 130 - Communication module; 200 - Dialogue simplification device; 201 - Acquisition module; 202 - Determination module; 203 - Simplification module. Detailed Implementation
[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0022] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0023] Please refer to Figure 1This is a block diagram of an electronic device 100. The electronic device 100 includes a memory 110, a processor 120, and a communication module 130. The memory 110, processor 120, and communication module 130 are electrically connected to each other directly or indirectly to realize data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.
[0024] The memory 110 is used to store programs or data. The memory 110 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.
[0025] The processor 120 is used to read / write data or programs stored in the memory 110 and to perform corresponding functions.
[0026] The communication module 130 is used to establish a communication connection between the electronic device 100 and other communication terminals through the network, and to send and receive data through the network.
[0027] It should be understood that, Figure 1 The structure shown is only a schematic diagram of the electronic device 100. The electronic device 100 may also include components that are larger than... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown. Figure 1 The components shown can be implemented using hardware, software, or a combination thereof.
[0028] In intelligent dialogue systems based on Large Language Models (LLMs), as the number of dialogue rounds increases, contextual information accumulates continuously, including the user's past questions, model-generated answers, reasoning processes (thought chains), tool calls, and their returned results. If all historical information is indiscriminately input into the model, the context length will grow rapidly, significantly increasing the model's computational burden, inference latency, and token usage costs, and may even introduce irrelevant or outdated information that interferes with current decision-making.
[0029] To address the aforementioned issues, this invention provides a simplified dialogue method, which will be described in detail below.
[0030] Please refer to Figure 2 The simplified dialogue method includes steps S101 to S103.
[0031] S101: Retrieve new questions and historical dialogues from users in the current round.
[0032] At the start of each round of dialogue processing, the system first performs a structured acquisition step of contextual data. When the system receives the user's latest input, it determines that it is the "new question content of the user" for the current round and uses it as the trigger signal for this round of dialogue processing.
[0033] Simultaneously, the system reads the "historical dialogue content." This historical dialogue content includes the user's original question text and the model-generated answer text from each round. Optionally, the historical dialogue content also includes supplementary elements such as the chain-of-thought (CoT) information upon which the answers in each round depended, instructions for calling external tools, and the returned results.
[0034] In addition, the dialogue system retains an initial Prompt unit to define the agent's role, behavioral rules, and functional boundaries. This unit remains unchanged throughout the dialogue and is always incorporated into the context input.
[0035] Therefore, the data obtained in step S101 includes three components: 1. New user question content: The natural language text entered by the user in the current round, which serves as the direct driving signal for the generation of this response; 2. Historical dialogue content: Consists of multiple question-and-answer pairs arranged chronologically, reflecting the complete evolution of the dialogue; 3. Initial Prompt Unit: Preset system-level instructions used to guide the agent's behavior consistency.
[0036] S102, divide the historical dialogue content into multiple question-and-answer pairs, and determine the related context units from all question-and-answer pairs based on the user's new question content and the dialogue round of each question-and-answer pair.
[0037] The historical dialogue content is divided into multiple independent "question-answer pair units" according to the chronological order of the dialogue rounds. Each question-answer pair unit corresponds to a complete user-model interaction process.
[0038] This modular partitioning method enables granular control over complex contextual information, facilitating subsequent semantic relevance evaluation and selective retention by unit.
[0039] In a possible implementation, the process of "dividing historical dialogue content into multiple question-and-answer pairs" can be as follows: treat the user's question content and the dialogue system's response content in each dialogue round in the historical dialogue content as a question-and-answer pair unit, thus obtaining multiple question-and-answer pair units.
[0040] Understandably, the basic components of each question-and-answer pair unit include the user's original question text and the answer text generated by the dialogue system in that round. Furthermore, if the round involves thought chain reasoning or the invocation of external tools, the question-and-answer pair unit may also include corresponding supplementary elements, such as intermediate reasoning steps, tool invocation instructions, and execution results.
[0041] After dividing the question-answer pair units, the new question content and each question-answer pair unit are vectorized to obtain the new question vector and multiple historical question-answer pair vectors. Based on each historical question-answer pair vector and the new question vector, the semantic similarity between each question-answer pair unit and the new question content is obtained.
[0042] The embodiments of the present invention employ a lightweight bidirectional encoder model as a semantic encoder to vectorize the new user question content and each question-answer pair unit.
[0043] For new user questions, the text is directly input into the semantic encoder, which outputs a dense semantic vector of fixed dimensions (e.g., 128 dimensions) to represent its overall semantic meaning.
[0044] For each question-and-answer pair, its core content—the "user's question text"—is concatenated with the "dialogue system's response text" to form a unified input text. This input text is then fed into the same semantic encoder to generate a corresponding semantic vector. This concatenation operation effectively integrates the informational intent of both the questioner and the responder, enhancing the understanding of contextual functional roles.
[0045] Subsequently, the cosine similarity between the semantic vector of the user's new question and the semantic vector of each question-answer pair is calculated to obtain an initial similarity score, with a value range of [-1, 1]. To further adapt to the context filtering logic, this score is normalized and mapped to the interval [0, 1], serving as the final "semantic similarity" metric, reflecting the degree of semantic relevance between the question-answer pair and the current question.
[0046] To improve computational efficiency, the semantic vectors corresponding to the newly generated question-answer pair units are persistently stored after each round of dialogue. When the next round of user questions arrives, only the vector of the current question needs to be recalculated, without having to repeatedly encode all historical units, thus significantly reducing the overhead of real-time inference.
[0047] By structurally dividing historical dialogue content into question-and-answer pairs and combining a lightweight semantic model to calculate the semantic similarity between the dialogue content and the user's new questions, this embodiment of the invention achieves accurate quantification of contextual information relevance, providing a reliable data foundation for subsequent context filtering based on dynamic thresholds, and balancing semantic understanding depth and operational efficiency.
[0048] Traditional context management methods typically use fixed-length window truncation or static similarity thresholds for filtering, which are difficult to adapt to the characteristics of information importance evolving over time in long-term dialogues: on the one hand, recent interactions may be close in time but not necessarily semantically related; on the other hand, key decision-making criteria may appear in earlier rounds, and simply discarding them will lead to context breakage.
[0049] Therefore, this embodiment of the invention comprehensively considers semantic relevance and temporal distance factors to identify and filter out “associated context units” that have potential contributions to the generation of the current response from all question-answer pair units.
[0050] In a possible implementation, the process of filtering related higher-level units could be as follows: For each question-and-answer pair, determine the similarity threshold corresponding to the question-and-answer pair based on the dialogue round it corresponds to. If the semantic similarity between the question-and-answer pair and the user's new question is not less than the similarity threshold corresponding to the question-and-answer pair, then the question-and-answer pair is identified as a related context unit.
[0051] The implementation process of "determining the similarity threshold corresponding to the question-answer pair unit based on the dialogue round corresponding to the question-answer pair unit" can be as follows: obtain the round distance between the dialogue round corresponding to the question-answer pair unit and the current round; obtain the similarity threshold corresponding to the question-answer pair unit based on the round distance.
[0052] In this embodiment of the invention, a dialogue round distance k is defined for each question-answer pair unit relative to the current round: the currently being processed new question is defined as round 0, the most recent historical question-answer pair (i.e., the previous round) corresponds to k=1, the previous two rounds correspond to k=2, and so on. This round distance k reflects the temporal position of the question-answer pair unit in the dialogue flow and serves as the basic parameter for dynamic threshold calculation.
[0053] A dynamic retention threshold T_k that varies with round distance is constructed based on the sigmoid-like function, and its mathematical expression is as follows: T_k=min([1 / (1+e^(alpha(k-k_0))], T_max) Where: alpha is the steepness coefficient, which controls the rate of increase of the threshold, and can be 0.1; k_0 is the inflection point round, which represents the center position of the threshold increase, and can be 5; T_max is the maximum retention threshold, which prevents indiscriminate filtering of long-term content, and can be 0.8. T_k increases smoothly as k increases, reflecting the "relaxed for near and strict for far" selection principle. For dialogue content that is relatively close to the present (e.g., k<5), even if the semantic relevance is slightly low, it tends to be retained to maintain the continuity of the dialogue. For historical content that is further away, a higher threshold is set, and it is only retained when its semantics are highly relevant, thereby effectively suppressing the accumulation of noise.
[0054] The semantic similarity of each question-answer pair is compared with the dynamic threshold T_k at the distance k of the corresponding round: if the semantic similarity of the question-answer pair is greater than or equal to T_k, it is determined to be an associated context unit and included in the subsequent context reconstruction process; otherwise, it is regarded as irrelevant or weakly related information.
[0055] In this embodiment of the invention, to further improve the efficiency of context simplification, after determining the question-answer pair unit as the associated context unit, the similarity threshold of the associated context unit's subordinate elements can be determined according to the similarity threshold corresponding to the associated context unit; the semantic similarity between the subordinate elements of the associated context unit and the user's new question content can be obtained; if the semantic similarity between the subordinate elements of the associated context unit and the user's new question content is less than the subordinate element similarity threshold, the subordinate elements of the associated context unit are deleted.
[0056] Understandably, given that the question-and-answer pair is marked as an associated unit, its subordinate elements (such as thought chains, tool call results, etc.) can be evaluated a second time. If the semantic similarity between the subordinate element and the user's new question is less than min(T_k*1.2, 1), then only the main question-and-answer text is retained, and the subordinate elements are discarded, thereby achieving fine-grained context compression.
[0057] S103, retain the associated context unit as the simplified historical dialogue content.
[0058] For each question-answer pair unit, if its semantic similarity is greater than the dynamic retention threshold T_k under the corresponding round distance, that is, it has been marked as an "associated context unit", then it is completely retained in the subsequent context input; otherwise, if it does not meet the retention condition, it is determined to be an unassociated or weakly related unit, and is deleted or masked, and will no longer participate in the current round of reasoning.
[0059] This filtering process employs a selective elimination mechanism, removing only historical fragments that are quantitatively assessed as semantically irrelevant, thus avoiding the loss of key information that may occur with traditional fixed-window truncation methods. For example, in a data analysis-related dialogue scenario, a user might ask in round 3, "Please generate a trend chart based on yesterday's sales data," and then in round 8, "Could you please provide more information on the regional distribution?"—Although the two questions are far apart, because they both point to the same data source, the semantic model can still detect a high relevance, thereby ensuring that the original query and tool call results are preserved and maintaining decision-making consistency.
[0060] It should be noted that, in this embodiment of the invention, question-answer pairs not identified as associated higher-level units are excluded from the model input context in the current inference round. All original dialogue records can still be fully archived in the background for auditing, debugging, or subsequent backtracking analysis, ensuring system interpretability and data integrity.
[0061] By employing a context simplification mechanism based on both semantic and turn-based judgment, this invention effectively filters noise information in long-cycle dialogues, improving the efficiency of LLM inference while enhancing the accuracy of agent decision-making and context awareness.
[0062] Further, please refer to Figure 3 The dialogue simplification method provided in this embodiment of the invention further includes step S104.
[0063] S104: Using the simplified historical dialogue content, the user's new question content, and the initial dialogue instructions, construct input content, and respond to the user's new question content based on the input content to complete the question and answer for the current round.
[0064] After simplifying the historical dialogue content, all relevant context units that meet the conditions are summarized, and combined with the initial Prompt unit that is always retained and the current user's new question content, a complete and semantically focused input context sequence is constructed and passed to the large language model for this round of response generation.
[0065] In summary, the embodiments of the present invention realize the transformation of the context filtering strategy from "static and extensive" to "dynamic and precise", which not only improves the information density of the context, but also significantly reduces the transmission and calculation overhead of invalid tokens, thereby optimizing the system resource utilization while ensuring the quality of the dialogue.
[0066] To perform the corresponding steps in the above embodiments and various possible methods, an implementation of the dialogue simplification device 200 is given below. Further, please refer to... Figure 4 , Figure 4This is a functional block diagram of a dialogue simplification device 200 provided in an embodiment of the present invention. It should be noted that the basic principle and technical effects of the dialogue simplification device 200 provided in this embodiment are the same as those in the above embodiments. For the sake of brevity, any parts not mentioned in this embodiment can be referred to the corresponding content in the above embodiments. The dialogue simplification device 200 includes: The acquisition module 201 is used to acquire the content of new questions asked by users in the current round and the content of historical dialogues.
[0067] The determination module 202 is used to divide the historical dialogue content into multiple question-and-answer pair units, and determine the associated context units from all question-and-answer pair units based on the user's new question content and the dialogue round of each question-and-answer pair unit.
[0068] The simplification module 203 is used to use the associated context unit as the simplified historical dialogue content.
[0069] Optionally, the above modules can be stored in the form of software or firmware. Figure 1 The memory 110 shown is either stored in or embedded in the operating system (OS) of the electronic device 100, and can be used by... Figure 1 The processor 120 in the memory executes the commands. Meanwhile, the data and program code required to execute the above modules can be stored in memory.
[0070] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0071] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0072] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, electronic device 100, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for simplifying dialogue, characterized in that, The method includes: Retrieve new questions and historical dialogues from users in the current round; The historical dialogue content is divided into multiple question-and-answer pairs. Based on the user's new question and the dialogue round of each question-and-answer pair, the associated context unit is determined from all the question-and-answer pairs. The associated context unit is used as a simplified version of the historical dialogue content.
2. The dialogue simplification method according to claim 1, characterized in that, The step of dividing the historical dialogue content into multiple question-answer pair units includes: Each question and response from each round of the historical dialogue is treated as a question-and-answer pair unit, resulting in multiple question-and-answer pair units.
3. The dialogue simplification method according to claim 1, characterized in that, The step of determining the associated context unit from all the question-and-answer pairs based on the content of the user's new question and the dialogue round of each question-and-answer pair includes: The new question content of the user and each question-answer pair unit are vectorized to obtain a new question vector and multiple historical question-answer pair vectors; Based on each of the historical question-answer pair vectors and the new question vector, the semantic similarity between each question-answer pair unit and the content of the user's new question is obtained; Based on the semantic similarity between each question-answer pair and the user's new question, as well as the dialogue round of each question-answer pair, the associated context unit is determined from all the question-answer pairs.
4. The dialogue simplification method as described in claim 3, characterized in that, The step of determining the associated context unit from all question-answer pairs based on the semantic similarity between each question-answer pair unit and the user's new question content, and the dialogue round of each question-answer pair unit, includes: For each question-answer pair unit, a similarity threshold is determined based on the dialogue round corresponding to the question-answer pair unit. If the semantic similarity between the question-and-answer pair and the user's new question is not less than the similarity threshold corresponding to the question-and-answer pair, then the question-and-answer pair is determined as the associated context unit.
5. The dialogue simplification method as described in claim 4, characterized in that, The step of determining the similarity threshold corresponding to the question-answer pair unit based on the dialogue round corresponding to the question-answer pair unit includes: Obtain the turn distance between the dialogue turn corresponding to the question-answer pair unit and the current turn; Based on the round distance, the similarity threshold corresponding to the question-answer pair unit is obtained.
6. The dialogue simplification method as described in claim 4, characterized in that, The question-and-answer pair unit includes ancillary elements, which are the thought chain information or tool call results associated with the question-and-answer pair unit. After determining the question-and-answer pair unit as the associated context unit, the following is also included: Based on the similarity threshold corresponding to the associated context unit, determine the similarity threshold of the subordinate elements of the associated context unit; Obtain the semantic similarity between the associated context unit's subordinate elements and the user's new question content; If the semantic similarity between the associated context unit's subordinate element and the user's new question is less than the subordinate element similarity threshold, then the subordinate element of the associated context unit will be deleted.
7. The dialogue simplification method as described in any one of claims 1-6, characterized in that, The method further includes: Using the simplified historical dialogue content, the user's new question content, and the initial dialogue instructions, input content is constructed to respond to the user's new question content based on the input content, thus completing the question and answer for the current round.
8. A simplified dialogue device, characterized in that, The device includes: The acquisition module is used to acquire new questions and historical dialogue content from users in the current round. The determination module is used to divide the historical dialogue content into multiple question-and-answer pair units, and determine the associated context unit from all the question-and-answer pair units based on the new question asked by the user and the dialogue round of each question-and-answer pair unit. The simplification module is used to treat the associated context unit as simplified historical dialogue content.
9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing machine-executable instructions that can be executed by the processor to implement the dialogue simplification method according to any one of claims 1-7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the dialogue simplification method as described in any one of claims 1-7.