Processing method and electronic device

Abstract

This disclosure provides a processing method, comprising: in response to receiving a first request, matching the first request with at least one pre-stored information group to obtain a matching result, wherein the information group is generated based on a second request received prior to receiving a request associated with the first request, the information group including request information corresponding to a third request and a processing result obtained by processing the third request based on a first model, the third request being a sub-request generated based on the second request; and in response to the matching result indicating that the first request matches a target information group in the at least one information group, feeding back the first request based on the processing result corresponding to the target information group. This disclosure also provides an electronic device capable of performing the above method.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence, and more specifically, to a processing method and an electronic device. Background Technology

[0002] Currently, human-computer interaction systems based on large language models generally adopt a "question-and-answer" processing mode. In this mode, for each user input request, the large language model must be invoked to perform a complete inference calculation and generate a response. When the user's initial request is broad or vague, multiple rounds of follow-up questions are needed to gradually clarify their true intention. Each new interaction means a completely new and complete model inference process, resulting in a serious waste of computing resources and slow feedback speed. Summary of the Invention

[0003] In view of this, the present disclosure provides a processing method and an electronic device.

[0004] One aspect of this disclosure provides a processing method, comprising: in response to receiving a first request, matching the first request with at least one pre-stored information group to obtain a matching result, wherein the information group is generated based on a second request received prior to receiving a request associated with the first request, the information group including request information corresponding to a third request and a processing result obtained by processing the third request based on a first model, the third request being a sub-request generated based on the second request; and in response to the matching result indicating that the first request matches a target information group in the at least one information group, feeding back the first request based on the processing result corresponding to the target information group.

[0005] According to an embodiment of this disclosure, the process of generating an information group includes: in response to receiving a second request, generating at least one third request associated with the second request; processing each third request in parallel based on a first model to obtain a processing result corresponding to each third request; and storing the request information corresponding to the third request and the processing result corresponding to the third request as an information group.

[0006] According to embodiments of this disclosure, generating at least one third request associated with a second request includes: determining at least one sub-request associated with the second request based on at least one of a preset rule template, a preset generation tool, and a second model, wherein the size of the second model is smaller than that of the first model, and the size represents at least one of the number of parameters of the model, the size of the storage space occupied by the model, and the number of layers of the model.

[0007] According to embodiments of this disclosure, in response to receiving a second request, generating at least one third request associated with the second request includes: determining at least one intent associated with the second request based on the second request, wherein the information entropy corresponding to the intent is less than the information entropy corresponding to the second request; and generating at least one third request corresponding to each intent.

[0008] According to embodiments of this disclosure, determining at least one intent associated with a second request includes: in response to an information entropy corresponding to the second request being higher than an information entropy threshold, determining at least one intent associated with the second request.

[0009] According to embodiments of this disclosure, the process of generating an information group further includes: generating a first semantic vector corresponding to each third request; storing the request information corresponding to the third request and the processing result corresponding to the third request as a set of information groups, including: storing the first semantic vector as the request information corresponding to the third request and the processing result corresponding to the third request as a set of information groups.

[0010] According to an embodiment of this disclosure, in response to receiving a first request, matching the first request with at least one pre-stored information group to obtain a matching result includes: generating a second semantic vector corresponding to the first request based on the first request; and performing semantic similarity matching between the second semantic vector and the first semantic vector in each information group to obtain a matching result.

[0011] According to embodiments of this disclosure, the processing method further includes: in response to a matching result indicating that the first request does not match any of the information groups, processing the first request according to a first model to obtain a processing result corresponding to the first request.

[0012] According to embodiments of this disclosure, the processing method further includes: generating a semantic vector corresponding to the first request; and storing the processing result corresponding to the first request and the semantic vector corresponding to the first request as a set of information groups.

[0013] According to embodiments of this disclosure, the third request is determined based on at least one of a preset rule template, a preset generation tool, and a second model. The processing method further includes: adjusting the rules of the preset rule template according to the processing result corresponding to the first request, and / or adjusting the working parameters of the generation tool, and / or adjusting the model parameters of the second model.

[0014] According to embodiments of this disclosure, the processing method further includes: in response to receiving a fourth request, deleting the information group corresponding to the fourth request, the fourth request including indication information corresponding to at least one information group; and / or, in response to the information group being stored for a time greater than or equal to a time threshold, deleting the information group.

[0015] According to embodiments of this disclosure, the first request is feedback information for recommendation information, which is generated based on request information corresponding to at least a portion of the information group and / or the corresponding processing results.

[0016] Another aspect of this disclosure provides a processing apparatus, comprising: a matching module, configured to, in response to receiving a first request, match the first request with at least one pre-stored information group to obtain a matching result, wherein the information group is generated based on a second request received prior to receiving a request associated with the first request, the information group including request information corresponding to a third request and a processing result obtained by processing the third request based on a first model, the third request being a sub-request generated based on the second request; and a feedback module, configured to, in response to the matching result indicating that the first request matches a target information group in the at least one information group, provide feedback on the first request based on the processing result corresponding to the target information group.

[0017] Another aspect of this disclosure provides an electronic device, including: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the following operations: in response to receiving a first request, matching the first request with at least one pre-stored information group to obtain a matching result, wherein the information group is obtained based on a second request received prior to receiving the first request, the information group including request information corresponding to a third request and a processing result obtained by processing the third request based on a first model, the third request being a sub-request generated based on the second request; and in response to the matching result indicating that the first request matches a target information group in the at least one information group, feeding back the first request based on the processing result corresponding to the target information group.

[0018] Another aspect of this disclosure provides a computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform a processing method according to any of the foregoing embodiments.

[0019] Another aspect of this disclosure provides a computer program product, including a computer program / instructions, characterized in that the computer program / instructions, when executed by a processor, implement the operation of the processing method of any of the foregoing embodiments. Attached Figure Description

[0020] The above and other objects, features and advantages of this disclosure will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0021] Figure 1 A flowchart illustrating a processing method according to an embodiment of the present disclosure is shown schematically.

[0022] Figure 2 A flowchart illustrating the generation of information groups in a processing method according to an embodiment of the present disclosure is shown schematically.

[0023] Figure 3A flowchart illustrating the generation of a third request in a processing method according to an embodiment of the present disclosure is shown schematically.

[0024] Figure 4 This schematically illustrates another flowchart of the process for generating information groups according to an embodiment of the present disclosure;

[0025] Figure 5 A flowchart illustrating the process of obtaining a matching result according to an embodiment of the present disclosure is shown schematically.

[0026] Figure 6 Another flowchart illustrating a processing method according to an embodiment of the present disclosure is shown schematically;

[0027] Figure 7 Another flowchart illustrating a processing method according to an embodiment of the present disclosure is shown schematically;

[0028] Figure 8 Another flowchart illustrating a processing method according to an embodiment of the present disclosure is shown schematically;

[0029] Figure 9A The illustration shows a flowchart of the information group pre-generation stage according to an embodiment of the present disclosure;

[0030] Figure 9B The illustration schematically shows a flowchart of processing a first request according to an embodiment of the present disclosure;

[0031] Figure 10 A block diagram schematically illustrates a processing apparatus according to embodiments of the present disclosure; and

[0032] Figure 11 A block diagram of an electronic device suitable for implementing the methods described above, according to embodiments of the present disclosure, is illustrated schematically. Detailed Implementation

[0033] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0034] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0035] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0036] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0037] In the embodiments disclosed herein, the collection, updating, analysis, processing, use, transmission, provision, disclosure, and storage of data (e.g., including but not limited to user personal information) comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures have been taken to prevent unauthorized access to user personal information data and to safeguard user personal information security, network security, and national security.

[0038] Embodiments of this disclosure provide a processing method, including: in response to receiving a first request, matching the first request with at least one pre-stored information group to obtain a matching result, wherein the information group is generated based on a second request received prior to receiving a request associated with the first request, the information group including request information corresponding to a third request and a processing result obtained by processing the third request based on a first model, the third request being a sub-request generated based on the second request; in response to the matching result indicating that the first request matches a target information group in the at least one information group, feeding back the first request based on the processing result corresponding to the target information group.

[0039] Figure 1 A flowchart illustrating a processing method according to an embodiment of the present disclosure is shown schematically.

[0040] like Figure 1 As shown, the processing method may include at least operations S110 to S120.

[0041] In operation S110, in response to receiving a first request, a matching result is obtained by matching the first request with at least one pre-stored information group. The information group is generated based on a second request received prior to receiving the request associated with the first request. The information group includes request information corresponding to a third request and processing results obtained by processing the third request based on the first model. The third request is a sub-request generated based on the second request.

[0042] The first request typically represents a question or interactive text entered by the user in the current interaction round. The second request is a prior request received in the same interaction session or historical interactions, preceding the first request in terms of time. The first model is a large language model with natural language processing and logical reasoning capabilities. The information group acts as a pre-generated and stored set of information, containing request information for more specific sub-requests (i.e., the third request) derived from the second request, as well as the processing results obtained by the first model after processing the third request.

[0043] In this operation, upon receiving the current question instruction, real-time inference calculations of a large-capacity model are not immediately triggered. Instead, it is used as a query condition to search for matching historical derived data in a pre-built data cache pool. The user's current follow-up question is often a specific elaboration on a prior broad topic, and these possible directions of elaboration have already been predicted and processed in advance when the prior topic is received.

[0044] For example, if a user entered "a coastal city" (the second request) in the previous round or several rounds of the current interaction, upon receiving this input, the system has automatically broken down this broad question into several specific sub-questions such as "tourist attractions in a coastal city" and "local delicacies in a coastal city" (the third request). A large language model has been pre-called to generate answers to these sub-questions, and the questions and answers are packaged and stored as multiple information groups. When the user subsequently asks "What's good to eat there?" (the first request) in the current round, upon receiving this instruction, it is matched against the stored information groups to verify whether it matches the pre-prepared responses related to "local delicacies."

[0045] In operation S120, in response to the matching result indicating that the first request matches with the target information group in at least one information group, the first request is fed back based on the processing result corresponding to the target information group.

[0046] A target information group refers to a specific information group among numerous pre-stored information groups whose request information matches the first request in terms of expression or core meaning. Feedback on the first request based on the processing result corresponding to the target information group involves outputting the final interactive response based on the extracted processing result, completing the current dialogue round. When the matching result indicates that a matching target information group has been found, the processing flow will directly bypass the real-time call to the first model, directly extract the processing result contained within the target information group and provide feedback, thereby achieving a rapid response to the first request.

[0047] For example, continuing from the previous example, if the matching result indicates that the first request "What are some good places to eat?" matches the information group (i.e., the target information group) containing the request information "specialty snacks of a coastal city", then the description text about seafood, specialty pastries and other delicacies that has been generated by the first model for the request information (i.e., the processing result) is directly extracted and fed back as the answer to the first request.

[0048] When there are multiple target information groups, operation S120 may include: jointly feeding back a first request based on the processing results corresponding to each of the multiple target information groups; and / or, the matching result also characterizes the degree of matching between the first request and the information group; determining the target processing result based on the degree of matching corresponding to each target information group; and feeding back the first request based on the target processing result.

[0049] The degree of matching reflects the closeness of the connection or content overlap between the first request and the request information in the information group, and can be measured by text matching score, keyword overlap ratio, or other relevance evaluation indicators. The target processing result is the final output content selected or merged from multiple processing results corresponding to multiple target information groups.

[0050] The first request may match multiple information groups simultaneously. In this case, the processing results corresponding to each of these target information groups can be extracted, concatenated, or summarized to provide feedback on the first request. Alternatively, when multiple target information groups exist, the matching degree score can be used as a weight or filtering criterion to select the one or several processing results with the highest matching degree as the target processing result for feedback, thus ensuring the accuracy of the feedback.

[0051] For example, if the first request is "Introduce local food, drinks, and entertainment," the matching result might indicate that the first request matches two target information groups simultaneously: one containing "local snacks" and the other containing "tourist attractions." In this case, the processing results for food and attractions corresponding to these two target information groups can be combined and fed back together. Alternatively, if the first request is "What are some traditional street foods?", and it is found that the matching score with the target information group containing "local snacks" is 0.9, while the matching score with the target information group containing "traditional culture" is 0.6, then the processing result corresponding to the target information group containing "local snacks" is determined based on the higher matching score and fed back as the target processing result.

[0052] According to embodiments of this disclosure, after receiving a second request, a related third request is pre-generated and processed by the first model in advance, constructing an information group containing request information and processing results. This pre-processing mechanism directly leads to a significant reduction in response latency. When a first request matching the information group is received, only a matching operation is needed to directly utilize the processing results corresponding to the target information group for feedback, completely avoiding the time-consuming process of repeatedly starting the first model for each first request. This, in turn, enables near-instantaneous smooth responses in multi-round interaction scenarios. Simultaneously, since existing processing results are used instead of real-time calls to the first model, the computational overhead for continuous or related requests is greatly reduced, alleviating the severe waste of computational resources in the serial processing mode.

[0053] Figure 2 A flowchart illustrating the generation of information groups in a processing method according to an embodiment of the present disclosure is shown.

[0054] like Figure 2 As shown, based on the aforementioned embodiments, the process of generating an information group may include operations S210 to S230.

[0055] In operation S210, in response to receiving the second request, at least one third request associated with the second request is generated. The third request is a sub-request derived from the second request, and its relevance lies in the continuation or refinement of the content theme or intent. There are usually multiple third requests to cover the various dimensions that the second request may imply. In some cases, there may be only one. After receiving the second request, by analyzing its potential expansion directions, at least one third request associated with it is generated, breaking down the broad or vague second request into more specific and independently processable sub-units. For example, after receiving the second request "a coastal city," by expanding its potential dimensions, at least one third request associated with it, such as "tourist attractions in a coastal city" or "local delicacies in a coastal city," is generated. These third requests focus on different aspects of the second request.

[0056] In operation S220, each third request is processed in parallel based on the first model to obtain the processing result corresponding to each third request. Parallel processing refers to submitting multiple third requests to the first model simultaneously or nearly simultaneously for inference calculation, avoiding the inefficient mode of sequential processing. This can be achieved by utilizing the batch processing capability of the first model itself, concatenating multiple third requests into a batch and inputting it into the model. The model can generate all corresponding processing results in one forward propagation, greatly improving processing efficiency. For example, continuing the previous example, third requests such as "tourist attractions in a coastal city" and "local delicacies in a coastal city" are packaged and submitted to the first model. Through parallel inference calculation, the processing result of each third request is obtained. For example, the result for tourist attractions includes a list and description of the attractions, and the result for local delicacies includes recommended dishes and special introductions.

[0057] In operation S230, the request information corresponding to the third request and the processing result corresponding to the third request are stored as an information group. The request information is a data form that represents the core content or characteristics of the third request and is used for subsequent matching and identification. The request information of the third request is paired with its corresponding processing result to form a complete information group and stored. This storage operation ensures the retrieval and reusability of the information group by establishing an association between the request information and the processing result.

[0058] For example, continuing from the previous example, the request information of the third request, "specialty snacks of a coastal city", is paired with its corresponding processing result (such as introductory text about seafood and pastries) to form an information group and stored in the cache space for easy retrieval later.

[0059] According to embodiments of this disclosure, at least one third request is generated in response to a second request, and a processing result is obtained based on parallel processing of the first model. Finally, the request information and the processing result are stored as an information group. This proactively prepares for potential subsequent requests, thus avoiding the time-consuming computation of starting the first model from scratch for each request in the face of continuous interaction, significantly reducing overall response latency and optimizing the smoothness of multi-turn dialogues. Simultaneously, because the parallel processing and storage mechanism reuses the computational output of the first model, it reduces repetitive reasoning for similar topics, saves computational resources, and solves the problem of resource waste in serial mode.

[0060] According to an embodiment of this disclosure, operation S210 may include: determining at least one sub-request associated with the second request based on at least one of a preset rule template, a preset generation tool, and a second model, wherein the size of the second model is smaller than that of the first model, and the size represents at least one of the number of parameters of the model, the size of the storage space occupied by the model, and the number of layers of the model.

[0061] The preset rule template can be a pre-configured set of logical rules. For example, for a request containing a place name, the template can automatically expand to include "the history of {place name}", "the food of {place name}", "the transportation of {place name}", etc. The preset generation tool can be an auxiliary mechanism specifically for request expansion, such as calling external knowledge graphs or search engine APIs to obtain entities and concepts related to the second request. The second model can be a model smaller than the first model, such as a lightweight sequence-to-sequence model or a small language model specifically for intent understanding and expansion. Its number of parameters, storage space, or number of layers are all smaller, resulting in lower computational complexity. The second model can be trained based on the user's historical habits and can accurately predict the user's possible subsequent questions. Specifically, by analyzing the user's historical interaction data (such as past query topics, preference sequences, and feedback patterns), the second model can be personalized by training or fine-tuning. For example, transfer learning methods can be used to integrate a user-specific dataset into a general small language model, thereby improving the accuracy of expanding the intent of individual users. This allows the second model to not only rely on general knowledge but also consider the user context when processing the second request, achieving a more suitable third request generation. Operation S210 can determine sub-requests solely based on a preset rule template, or solely through a preset generation tool, or solely using a second model, or a combination of these methods. This process transforms the extension of the second request into a lightweight or preset operation to quickly generate multiple independently processable related sub-requests.

[0062] According to embodiments of this disclosure, by determining sub-requests associated with the second request based on at least one of a preset rule template, a preset generation tool, and a smaller second model, this lightweight determination mechanism directly reduces the computational complexity and time overhead of the expansion phase, avoiding the latency of directly calling the first model to process broad requests. This, in turn, makes the overall processing response of generating sub-requests faster. Simultaneously, introducing multiple expansion methods increases the flexibility of the solution, allowing the selection of the most suitable expansion strategy based on the type and context of the second request.

[0063] Figure 3 A flowchart illustrating the generation of a third request in a processing method according to an embodiment of this disclosure is shown schematically.

[0064] like Figure 3 As shown, based on the aforementioned embodiments, operation S210 may include operations S310 to S320.

[0065] In operation S310, based on the second request, at least one intent associated with the second request is determined, where the information entropy of the intent is less than the information entropy of the second request. An intent can be a potential concern or sub-topic extracted from the second request; information entropy is an indicator that quantifies the uncertainty or ambiguity of a request, with a higher value indicating more possible interpretations or directions covered by the request. The process of determining at least one intent based on the second request achieves a refined decomposition of the broad request by reducing information entropy, transforming the ambiguity of the second request into multiple intents with lower uncertainty.

[0066] For example, after receiving the second request "a certain coastal city", its information entropy is high, covering multiple potential directions. Through analysis, at least one intention such as "tourist attractions" or "local snacks" is determined. These intentions have low information entropy and focus on specific aspects.

[0067] In operation S320, at least one third request corresponding to each intent is generated. The third request is a sub-request that further specifies the intent to expand the depth or breadth of the intent. Generating at least one third request corresponding to each intent, by transforming the intent into an operable request form, enables the layer-by-layer expansion of the second request, ensuring that the third request inherits the low information entropy characteristic of the intent.

[0068] For example, following the previous example, for the intent "local specialties", the third request is generated as "What are the traditional street snacks in a certain coastal city", which further refines the specific content of the intent.

[0069] According to embodiments of this disclosure, by determining at least one low-information-entropy intent based on a second request and generating at least one third request corresponding to each intent, this layer-by-layer refinement mechanism from high uncertainty to low uncertainty improves the accuracy of expanding the second request. This makes the generated third requests more aligned with potential subsequent interaction directions, avoiding the generation of too many irrelevant or low-relevance sub-requests. This, in turn, improves the matching efficiency of subsequent information groups when facing the first request and reduces ineffective preprocessing computational paths. Furthermore, since the generation of intents and third requests reduces ambiguity, it also reduces the overall computational overhead of the preprocessing stage, alleviating the problems of high response latency and wasted computational resources caused by processing generalized requests.

[0070] According to an embodiment of this disclosure, operation S310 may include: in response to the information entropy corresponding to the second request being higher than an information entropy threshold, determining at least one intent associated with the second request, wherein the information entropy corresponding to the intent is lower than the information entropy corresponding to the second request. The operation of determining the associated intent is only performed in response to the information entropy of the second request being higher than the information entropy threshold. This conditional triggering ensures that refinement is only performed when there is high uncertainty in the second request, avoiding the inefficiency of performing expansion on all requests. The information entropy threshold can be set according to the actual application scenario and needs. For example, for a customer service robot, the information entropy threshold for a simple greeting can be set higher to avoid unnecessary expansion. For example, upon receiving a second request for "a coastal city," if its calculated information entropy is higher than the information entropy threshold, the operation of determining low information entropy intents such as "tourist attractions" or "local snacks" is triggered.

[0071] According to embodiments of this disclosure, at least one intent associated with the second request is determined only when the information entropy corresponding to the second request is higher than an information entropy threshold. This conditional judgment mechanism directly ensures that the expansion operation is initiated only for the highly ambiguous second request, avoiding unnecessary breakdown of specific requests. This improves the fit between the generated third request and potential subsequent interactions, making the construction of subsequent information groups more targeted and improving matching efficiency in subsequent interactions. Simultaneously, since threshold control reduces the number of unnecessary intent determinations, it also reduces the computational overhead of the expansion phase.

[0072] Figure 4 Another flowchart illustrating the generation of information groups in a processing method according to an embodiment of the present disclosure is shown schematically.

[0073] like Figure 4 As shown, based on the aforementioned embodiments, the process of generating information groups may include operation S410, and operation S230 may include operation S420.

[0074] In operation S410, a first semantic vector corresponding to each third request is generated. Generating the first semantic vector for each third request involves inputting the text content of the third request into a pre-trained semantic encoding model and mapping it to a vector space, thus achieving structured encoding of the request information. For example, continuing from the previous embodiment, for the third request "a specialty snack in a coastal city," its corresponding first semantic vector is generated to capture its core semantic elements.

[0075] In operation S420, the first semantic vector is used as the request information corresponding to the third request, and the processing result corresponding to the third request is stored together as an information group. The first semantic vector is here designated as the specific representation of the request information. By combining and storing the first semantic vector and the processing result as an information group, it is ensured that the semantic representation of the request is closely bound to the corresponding result, forming a complete and reusable data unit.

[0076] According to embodiments of this disclosure, by generating a first semantic vector for each third request and storing it together with the processing result as an information group, this vectorized request information representation directly improves the accuracy of the information group in capturing the essence of the request, enabling the pre-generated processing result to be efficiently organized in a structured form. Furthermore, since the first semantic vector replaces the original request text as request information, the auxiliary computational burden during storage is reduced, and the matching process is no longer limited to literal expressions, but can be generalized to different expressions with similar semantics.

[0077] Figure 5 A flowchart illustrating the process of obtaining a matching result according to an embodiment of the present disclosure is shown.

[0078] like Figure 5 As shown, based on the aforementioned embodiments, operation S110 may include operations S510 to S520.

[0079] In operation S510, based on the first request, a second semantic vector corresponding to the first request is generated. The second semantic vector is a vector representation of the semantic content of the first request, used for subsequent semantic matching with the first semantic vector in the information group. By mapping the text content of the first request to a vector space using the same semantic encoding model as that used to generate the first semantic vector, the second semantic vector is generated, achieving structured encoding of the request information and providing a computable representation for semantic similarity matching. Using the same encoding model ensures that the two vectors reside in the same semantic space, making similarity calculations comparable. For example, if the first request is "What's good to eat?", then its corresponding second semantic vector is generated, which captures the core semantic features of the request, such as "food" and "recommended locations".

[0080] In operation S520, semantic similarity matching is performed based on the second semantic vector and the first semantic vector in each information group to obtain the matching result. Semantic similarity matching refers to determining the degree of association between the first request and each information group by calculating the similarity between the second semantic vector and each first semantic vector in the information group. This matching process is implemented through distance or similarity measures (such as cosine similarity) in vector space to quantitatively determine the fit between the first request and the pre-stored information groups. For example, if the first request is "What are some good foods?", the cosine similarity of its corresponding second semantic vector and the first semantic vector of "a specialty snack in a coastal city" in the information group is calculated. If the similarity reaches a set threshold, the match is considered successful.

[0081] According to embodiments of this disclosure, by converting a first request into a second semantic vector and matching it with the first semantic vector in the information group based on semantic similarity, the target information group semantically related to the first request can be quickly identified, thereby directly using the corresponding processing results for feedback. This vectorized matching method directly improves the semantic accuracy and robustness of the matching process, avoiding the limitations of simple keyword or string matching. Even if a user's question "What are some good things to eat?" is not exactly the same as the pre-stored "local specialties," they can still be accurately matched due to semantic similarity. This makes the matching results more reliable and reduces the probability of misjudgments or low-relevance matches. Furthermore, since semantic vector matching relies on efficient vector computation (such as using a vector database for approximate nearest neighbor search), it also reduces the computational complexity and time overhead of the matching stage, alleviating the problems of high response latency and wasted computational resources caused by matching.

[0082] Figure 6 Another flowchart illustrating a processing method according to an embodiment of the present disclosure is shown schematically.

[0083] like Figure 6 As shown, based on the foregoing embodiments, the processing method may further include operation S610.

[0084] In operation S610, in response to the matching result indicating that the first request does not match any of the information groups, the first request is processed according to the first model to obtain the processing result corresponding to the first request. A mismatch indicates that the semantics or content of the first request cannot correspond to any pre-stored information group. For example, when the highest similarity between the second semantic vector and all first semantic vectors is lower than a preset matching threshold, it is determined to be a mismatch. In response to the matching failure, the process switches to directly using the first model to process the first request to obtain its corresponding processing result. This branching process ensures that even when the pre-stored information groups cannot cover the request, the first request can still be responded to using the complete reasoning capability of the first model.

[0085] For example, following the aforementioned embodiments, if the first request is "the economic development history of a certain coastal city", and all information groups (such as focusing on tourism and snacks) do not match, then the first model is directly invoked to process the first request and generate a processing result about the economic development history.

[0086] According to embodiments of this disclosure, when a first request does not match any of the information groups, the first model is used to process the first request. This fallback mechanism directly ensures the integrity and robustness of the processing method, avoiding response failures due to insufficient coverage of pre-stored information groups. This further enables the provision of effective processing results even when faced with unexpected or unpredictable first requests. Since the first model is only invoked when there is a mismatch, combined with the aforementioned matching mechanism, the overall invocation frequency of the first model is reduced, thereby alleviating the high response latency and wasted computing resources caused by requiring complete inference for every request.

[0087] Figure 7 Another flowchart illustrating a processing method according to an embodiment of the present disclosure is shown schematically.

[0088] like Figure 7 As shown, based on the aforementioned embodiments, the processing method may further include operations S710~S720.

[0089] In operation S710, a semantic vector corresponding to the first request is generated. This semantic vector is a vector representation of the semantic content of the first request, similar to the aforementioned first or second semantic vector, and is used for subsequent storage and matching. Generating the semantic vector corresponding to the first request involves mapping the text of the first request to a vector space, thus achieving a quantified representation of its semantics.

[0090] In operation S720, the processing result corresponding to the first request and the semantic vector corresponding to the first request are stored as a set of information. Storing the processing result and semantic vector of the first request as a set of information transforms the real-time processing results into reusable cached data, enabling dynamic expansion of the information set and giving the system self-learning capabilities to continuously enrich the cached content. For example, continuing the previous example, the processing result corresponding to "the economic development history of a coastal city" (e.g., descriptive text of economic development stages) is paired with the semantic vector of the first request to form a new set of information and stored in the storage space, facilitating matching and reuse for subsequent similar requests.

[0091] According to embodiments of this disclosure, by generating a semantic vector corresponding to the first request and storing it along with the processing result as a set of information, this storage mechanism enables the caching of newly generated processing results in mismatch scenarios, transforming the originally one-time real-time inference output into a reusable set of information. This further improves the response efficiency to subsequent similar first requests, avoids repeated complete inference calculations for the same or similar requests, and dynamically expands the set of information sets.

[0092] Figure 8 Another flowchart illustrating a processing method according to an embodiment of the present disclosure is shown schematically.

[0093] like Figure 8 As shown, based on the foregoing embodiments, the third request is determined based on at least one of the preset rule template, the preset generation tool, and the second model, and the processing method may further include operation S810.

[0094] In operating S810, based on the processing result corresponding to the first request, the rules of the preset rule template are adjusted, and / or the working parameters of the generation tool are adjusted, and / or the model parameters of the second model are adjusted. This adjustment process, based on the processing result corresponding to the first request, uses the actual response content as feedback information to iteratively optimize the generation mechanism of the third request. This is a closed-loop feedback optimization process. For example, if the real-time processing result generated based on a certain first request is frequently matched in subsequent requests, it indicates that the expansion of that type of intent is successful, and the weight of the relevant rules can be increased. Conversely, if the pre-generated information group is not matched for a long time, its generation priority can be reduced or the generation strategy adjusted. This adjustment can be made individually for rule templates, working parameters, or model parameters, or in combination of multiple methods, to achieve adaptive improvement of the generation logic.

[0095] For example, if the processing result of the first request shows that the user prefers "history of economic development" rather than the preset "tourist attractions", then the rules of the extended dimension in the preset rule template can be adjusted accordingly, or the working parameters of the generation tool can be modified to increase the weight of economics-related factors, or the model parameters of the second model can be fine-tuned to improve the sensitivity to similar topics.

[0096] According to embodiments of this disclosure, by adjusting the rules of the preset rule template, the working parameters of the generation tool, and / or the model parameters of the second model based on the processing result corresponding to the first request, this feedback adjustment mechanism directly achieves dynamic optimization of the third request generation process, improves the fit between the generated sub-requests and the actual interaction direction, thereby increasing the hit rate of the information group in subsequent matching and reducing invalid preprocessing calculations caused by generation deviations.

[0097] According to embodiments of this disclosure, the processing method further includes: in response to receiving a fourth request, deleting the information group corresponding to the fourth request, wherein the fourth request includes indication information corresponding to at least one information group; and / or, in response to the storage time of the information group being greater than or equal to a time threshold, deleting the information group. The fourth request is used to indicate a specific request type for the deletion operation and includes indication information pointing to a target information group; the indication information may be an identifier or reference used to specify at least one information group to be deleted; the time threshold is a preset duration standard used to determine the validity period of the information group.

[0098] In response to receiving a fourth request to delete the corresponding information group, the system removes the specified information group through user-triggered deletion, enabling manual cleanup of stored content; and / or, automatically deletes the information group when its storage time reaches or exceeds a time threshold, achieving passive expiration management of the information group through a time-sensitivity mechanism. This deletion operation can be triggered solely by user actions or determined by time-sensitivity, or a combination of both, ensuring dynamic maintenance of the information group set.

[0099] For example, if a user sends a fourth request, "Delete the information group about 'tourist attractions in a coastal city'", which includes instruction information (such as keywords or IDs corresponding to the request information), the system will respond by deleting the information group; or, if an information group has been stored for more than a preset time threshold (such as 7 days), the information group will be automatically deleted to free up storage space.

[0100] According to embodiments of this disclosure, the streamlined management of information group sets through a deletion mechanism avoids storage expansion and an increased matching search space caused by the infinite accumulation of information groups, reduces computational overhead and time consumption during the matching process, and improves response efficiency. At the same time, the timely removal of irrelevant or expired information groups also reduces the probability of invalid matches.

[0101] According to embodiments of this disclosure, the first request is feedback information for recommended information, which is generated based on the request information corresponding to at least a portion of the information groups and / or the corresponding processing results. The first request can be feedback information for recommended information, which is generated by selecting the request information, processing results, or a combination of both from at least a portion of the information groups. This setting transforms the preprocessed output into user-visible recommendations, realizing a shift from passive response to proactive guidance.

[0102] After processing the second request for "a coastal city", the system generates recommended information based on the request information from multiple information groups (such as "tourist attractions" and "local snacks") and the processing results (such as a list of attractions and descriptions of snacks): "Topics you may be interested in: 1. Tourist attractions (click to view details); 2. Local snacks (click to view recommendations)". If the user clicks option 2, the processing instruction triggered by this click becomes the first request.

[0103] According to embodiments of this disclosure, the recommendation information is generated based on the request information and / or processing results of at least a portion of the information group. This proactive recommendation and feedback mechanism directly improves the utilization rate of the preprocessing results, enabling the generated information group to be transformed into user-visible options, reducing the frequency of users initiating broad requests from scratch, and also improving the user experience.

[0104] Figure 9A The illustration shows a schematic diagram of the information group pre-generation stage according to an embodiment of the present disclosure.

[0105] like Figure 9A As shown, firstly, the system receives an initial request from the user (corresponding to the second request in the aforementioned embodiment), for example, the user inputs the broad topic "City A". Then, the initial request undergoes intent recognition and expansion to explore the user's potential segmented needs. Based on the expansion results, batch prompts are generated, producing N specific questions such as "The history of City A" and "The food of City A" (corresponding to the multiple third requests determined in the aforementioned embodiment). Next, batch inference and caching operations are performed on these N questions, i.e., a large language model (such as the first model) is called to process these N questions respectively, obtaining corresponding pre-generated results 1 (e.g., an answer about "The history of City A..."), pre-generated results 2 (e.g., an answer about "The food of City A..."), up to pre-generated results N (corresponding to the processing results in the aforementioned embodiment). Finally, these N questions and their corresponding pre-generated results are paired as information groups and uniformly stored in a cache pool. Through this preprocessing mechanism, high-quality answer caches are prepared in advance before the user actually initiates a specific query (i.e., the first request), thus achieving a low-latency, extremely fast response when a match is found.

[0106] Figure 9B The illustration shows a schematic flowchart of processing a first request according to an embodiment of the present disclosure.

[0107] like Figure 9B As shown, firstly, the system receives the actual query currently input by the user (corresponding to the first request in the aforementioned embodiment), such as the user asking "What are some good places to eat?". Then, it performs feature extraction and vectorization mapping on the actual query to generate its corresponding vector representation (corresponding to the second semantic vector in the aforementioned embodiment). Next, it performs rapid semantic search and matching in a pre-built cache pool (i.e., a storage space containing multiple pre-stored information groups). Specifically, it calculates the semantic similarity between the second semantic vector and the request information vectors (i.e., the first semantic vectors) corresponding to each information group in the cache pool.

[0108] Based on this, if a match is successful (a cache hit, i.e., the similarity reaches or exceeds the threshold), the pre-generated processing result (such as a "list of recommended local snacks") is directly extracted from the matched target information group and quickly returned to the user as a response. This branch does not require calling large models for complex inference calculations, thus achieving extremely low latency and ultra-fast response.

[0109] If a match fails (cache miss, i.e., no match is found in any of the information groups), it means that the user's current query exceeds the coverage of the pre-stored information groups. In this case, a real-time inference fallback branch is triggered, calling a large language model (corresponding to the first model in the aforementioned embodiments) to perform complete real-time processing on the first request, obtaining the corresponding processing result and returning it to the user. Furthermore, in some embodiments, to continuously improve cache coverage, this newly received first request (and its semantic vector) is paired with the just-generated real-time processing result and stored as a new information group in the cache pool, achieving dynamic expansion of cached knowledge.

[0110] Figure 10 A block diagram of a processing apparatus according to an embodiment of the present disclosure is shown schematically.

[0111] like Figure 10 As shown, the processing device 1000 may include a matching module 1010 and a feedback module 1020.

[0112] The matching module 1010 is configured to, in response to receiving a first request, match the first request with at least one pre-stored information group to obtain a matching result. The information group is generated based on a second request received prior to the first request, and includes request information corresponding to a third request and a processing result obtained from processing the third request based on a first model. The third request is a sub-request generated based on the second request. In some embodiments, the matching module 1010 may be used to perform operation S110 in the above processing method, which will not be elaborated upon here.

[0113] The feedback module 1020 is used to respond to the matching result indicating that the first request matches with a target information group in at least one information group, and to provide feedback on the first request based on the processing result corresponding to the target information group. In some embodiments, the feedback module 1020 may be used to perform operation S120 in the above processing method, which will not be described in detail here.

[0114] Any one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure, or at least part of the functions of any one or more of them, can be implemented in one module. Any one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure can be implemented by dividing them into multiple modules. Any one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure can be at least partially implemented as hardware circuitry, such as a Field-Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a System-on-Chip, a System-on-a-Substrate, a System-on-Package, an Application-Specific Integrated Circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure can be at least partially implemented as computer program modules, which, when run, can perform corresponding functions.

[0115] For example, any plurality of the matching module 1010 and feedback module 1020 can be combined into one module / unit / subunit, or any one of the modules / units / subunits can be split into multiple modules / units / subunits. Alternatively, at least part of the functionality of one or more of these modules / units / subunits can be combined with at least part of the functionality of other modules / units / subunits and implemented in one module / unit / subunit. According to embodiments of this disclosure, at least one of the matching module 1010 and feedback module 1020 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the matching module 1010 and feedback module 1020 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0116] It should be noted that the data processing system part in the embodiments of this disclosure corresponds to the data processing method part in the embodiments of this disclosure. The specific description of the data processing system part is referred to in the data processing method part, and will not be repeated here.

[0117] Figure 11 A block diagram of an electronic device suitable for implementing the methods described above, according to embodiments of the present disclosure, is illustrated schematically. Figure 11The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0118] like Figure 11 As shown, an electronic device 1100 according to an embodiment of the present disclosure includes a processor 1101, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1102 or a program loaded from a storage portion 1108 into a random access memory (RAM) 1103. The processor 1101 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 1101 may also include onboard memory for caching purposes. The processor 1101 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0119] RAM 1103 stores various programs and data required for the operation of electronic device 1100. Processor 1101, ROM 1102, and RAM 1103 are interconnected via bus 1104. Processor 1101 performs various operations of the method flow according to embodiments of the present disclosure by executing programs in ROM 1102 and / or RAM 1103. It should be noted that the programs may also be stored in one or more memories other than ROM 1102 and RAM 1103. Processor 1101 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in said one or more memories.

[0120] According to embodiments of this disclosure, the electronic device 1100 may further include an input / output (I / O) interface 1105, which is also connected to a bus 1104. The electronic device 1100 may also include one or more of the following components connected to the input / output (I / O) interface 1105: an input section 1106 including a keyboard, mouse, etc.; an output section 1107 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1108 including a hard disk, etc.; and a communication section 1109 including a network interface card such as a LAN card, modem, etc. The communication section 1109 performs communication processing via a network such as the Internet. A drive 1110 is also connected to the input / output (I / O) interface 1105 as needed. A removable medium 1111, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 1110 as needed so that computer programs read from it can be installed into the storage section 1108 as needed.

[0121] According to embodiments of this disclosure, the method flow according to embodiments of this disclosure can be implemented as a computer software program. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via communication section 1109, and / or installed from removable medium 1111. When the computer program is executed by processor 1101, it performs the functions defined in the system of embodiments of this disclosure. According to embodiments of this disclosure, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0122] This disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs that, when executed, implement the method according to the embodiments of this disclosure.

[0123] According to embodiments of this disclosure, the computer-readable storage medium can be a non-volatile computer-readable storage medium. Examples include, but are not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this disclosure, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0124] For example, according to embodiments of this disclosure, a computer-readable storage medium may include one or more memories other than the ROM 1102 and / or RAM 1103 described above and / or ROM 1102 and RAM 1103.

[0125] Embodiments of this disclosure also include a computer program product comprising a computer program containing program code for performing the methods provided in the embodiments of this disclosure. When the computer program product is run on an electronic device, the program code is used to enable the electronic device to implement the processing methods provided in the embodiments of this disclosure.

[0126] When the computer program is executed by the processor 1101, it performs the functions defined in the system / apparatus of this disclosure embodiments. According to embodiments of this disclosure, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0127] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices or magnetic storage devices. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and downloaded and installed via communication section 1109, and / or installed from removable medium 1111. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof. According to embodiments of this disclosure, program code for executing the computer programs provided in embodiments of this disclosure can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code may be executed entirely on a user computing device, partially on a user device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to user computing devices via any type of network, including local area networks (LANs) or wide area networks (WANs), or they can be connected to external computing devices (e.g., via the Internet using an Internet service provider).

[0128] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. 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 indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated 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 or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions. Those skilled in the art will understand that the features described in the various embodiments of the present disclosure can be combined and / or combined in various ways, even if such combinations are not explicitly described in the present disclosure. In particular, the features described in the various embodiments of this disclosure may be combined and / or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0129] The embodiments of this disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this disclosure, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. A processing method, comprising: In response to receiving a first request, the system matches the first request with at least one pre-stored information group to obtain a matching result. The information group is generated based on a second request received before the first request is received. The information group includes request information corresponding to a third request and a processing result obtained by processing the third request based on a first model. The third request is a sub-request generated based on the second request. In response to the matching result indicating that the first request matches a target information group in the at least one information group, the first request is fed back based on the processing result corresponding to the target information group.

2. The method according to claim 1, wherein the process of generating the information group includes: In response to receiving the second request, at least one third request associated with the second request is generated; Based on the first model, each of the third requests is processed in parallel to obtain the processing result corresponding to each of the third requests; The request information corresponding to the third request and the processing result corresponding to the third request are stored as a group of information.

3. The method of claim 2, wherein generating at least one third request associated with the second request comprises: Based on at least one of a preset rule template, a preset generation tool, and a second model, at least one sub-request associated with the second request is determined, wherein the size of the second model is smaller than that of the first model, and the size represents at least one of the number of parameters of the model, the amount of storage space occupied by the model, and the number of layers of the model.

4. The method of claim 2, wherein generating at least one third request associated with the second request in response to receiving the second request comprises: Based on the second request, at least one intent associated with the second request is determined, wherein the information entropy corresponding to the intent is less than the information entropy corresponding to the second request; Generate at least one third request corresponding to each of the stated intentions.

5. The method of claim 4, wherein determining at least one intent associated with the second request comprises: If the information entropy corresponding to the second request is higher than the information entropy threshold, at least one intent associated with the second request is determined.

6. The method according to claim 2, wherein the process of generating the information group further includes: Generate a first semantic vector corresponding to each of the third requests; The step of storing the request information corresponding to the third request and the processing result corresponding to the third request as a group of information includes: The first semantic vector is used as the request information corresponding to the third request, and the processing result corresponding to the third request is stored as a group of information.

7. The method according to claim 6, wherein the step of responding to receiving a first request and matching the first request with at least one pre-stored information group to obtain a matching result includes: Based on the first request, generate a second semantic vector corresponding to the first request; Based on the second semantic vector and the first semantic vector in each of the information groups, a semantic similarity match is performed to obtain the matching result.

8. The method according to claim 1, further comprising: In response to the matching result indicating that the first request does not match any of the information groups, the first request is processed according to the first model to obtain the processing result corresponding to the first request; Generate the semantic vector corresponding to the first request; The processing result corresponding to the first request and the semantic vector corresponding to the first request are stored as a set of information.

9. The method according to claim 8, wherein the third request is determined based on at least one of a preset rule template, a preset generation tool, and a second model, and the method further comprises: Based on the processing result corresponding to the first request, adjust the rules of the preset rule template, and / or adjust the working parameters of the generation tool, and / or adjust the model parameters of the second model.

10. The method according to claim 1, further comprising: In response to receiving a fourth request, delete the information group corresponding to the fourth request, the fourth request including indication information corresponding to at least one of the information groups; And / or, In response to the information group being stored for a time greater than or equal to a time threshold, the information group is deleted.

11. The method according to claim 1, wherein the first request is feedback information for recommendation information, and the recommendation information is generated based on the request information corresponding to at least a portion of the information group, and / or the corresponding processing results.

12. An electronic device, comprising: At least one processor; as well as The memory connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor. These instructions, when executed by the at least one processor, enable the at least one processor to perform the following operations: in response to receiving a first request, matching the first request with at least one pre-stored information group to obtain a matching result, wherein the information group is obtained based on a second request received prior to receiving the first request, the information group including request information corresponding to a third request and a processing result obtained by processing the third request based on a first model, the third request being a sub-request generated based on the second request; and in response to the matching result indicating that the first request matches a target information group in the at least one information group, feeding back the first request based on the processing result corresponding to the target information group.

Images