Search result processing method, device, equipment and medium

Abstract

The present disclosure proposes a method, device, equipment and medium for processing search results, which relates to artificial intelligence fields such as natural language processing and deep learning, including: obtaining a search question and a reference question and answer library; obtaining a set of associated reference questions of the search question based on a reference question index, and obtaining respective associated reference answers of the associated reference question set in a reference answer library, and inputting each associated reference answer into a pre-acquired large model to obtain a target search result of the search question; inputting the target search result into the large model, and identifying whether the target search result meets a preset decomposition condition through the large model; in response to identifying that the target search result meets the preset decomposition condition, decomposing the target search result into steps through the large model to obtain a decomposed single step set; executing each single step in the single step set in sequence through the large model, obtaining a target execution result of the single step set, and returning the target execution result to a user end.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, and in particular to artificial intelligence fields such as natural language processing and deep learning. Background Art

[0002] With the development of technology, the operation of search engines and other related systems has become more complicated. When users use the system, they may need to perform related system operations themselves to achieve the purpose of using the business functions in the system. However, the system operations that users need to perform may be more complicated. In this scenario, it is difficult for users to perform system operations.

[0003] In the related technology, users can perform related system operations by relying on manual services. The degree of manual dependence is high, and problems may accumulate on the manual service side, resulting in low operating efficiency and poor user experience. Summary of the invention

[0004] The present disclosure proposes a method, device, equipment and medium for processing search results.

[0005] According to a first aspect of the present disclosure, a method for processing search results is proposed, the method comprising: obtaining a search question and a reference question and answer library, wherein the reference question and answer library comprises a reference question index and a reference answer library; obtaining a set of associated reference questions of the search question based on the reference question index, and obtaining respective associated reference answers of the associated reference question set in the reference answer library, and inputting each associated reference answer into a pre-acquired large model to obtain a target search result of the search question; inputting the target search result into the large model, and identifying through the large model whether the target search result satisfies a preset decomposition condition; in response to identifying that the target search result satisfies the decomposition condition, performing step decomposition on the target search result through the large model to obtain a decomposed single step set; executing each single step in the single step set in sequence through the large model, obtaining a target execution result of the single step set, and returning the target execution result to the user end.

[0006] According to a second aspect of the present disclosure, a search result processing device is proposed, the device comprising: a first acquisition module, used to acquire a search question and a reference question and answer library, wherein the reference question and answer library comprises a reference question index and a reference answer library; a second acquisition module, used to acquire a set of associated reference questions of the search question based on the reference question index, and acquire the associated reference answers of the associated reference question set in the reference answer library, and input each associated reference answer into a pre-acquired large model to obtain a target search result of the search question; an identification module, used to input the target search result into the large model, and identify whether the target search result meets a preset decomposition condition through the large model; a decomposition module, used to decompose the target search result into steps through the large model in response to identifying that the target search result meets the decomposition condition, and obtain a decomposed single step set; an execution module, used to execute each single step in the single step set in sequence through the large model, obtain a target execution result of the single step set, and return the target execution result to the user end.

[0007] According to a third aspect of the present disclosure, an electronic device is proposed, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the search result processing method proposed in the first aspect above.

[0008] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is proposed, wherein the computer instructions are used to enable the computer to execute the search result processing method proposed in the first aspect above.

[0009] According to a fifth aspect of the present disclosure, a computer program product is proposed, including a computer program, and when the computer program is executed by a processor, the computer program implements the search result processing method proposed in the first aspect.

[0010] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0012] Figure 1 A flowchart of a method for processing search results according to an embodiment of the present disclosure;

[0013] Figure 2A flowchart of a method for processing search results according to another embodiment of the present disclosure;

[0014] Figure 3 A flowchart of a method for processing search results according to another embodiment of the present disclosure;

[0015] Figure 4 A flowchart of a method for processing search results according to another embodiment of the present disclosure;

[0016] Figure 5 A flowchart of a method for processing search results according to another embodiment of the present disclosure;

[0017] Figure 6 A flowchart of a method for processing search results according to another embodiment of the present disclosure;

[0018] Figure 7 A schematic diagram of the structure of a search result processing device according to an embodiment of the present disclosure;

[0019] Figure 8 A schematic block diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0020] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0021] Data processing is a basic part of system engineering and automatic control. Data is a form of expression of facts, concepts or instructions, which can be processed by manual or automatic devices. After data is interpreted and given a certain meaning, it becomes information. Data processing is the collection, storage, retrieval, processing, transformation and transmission of data. The basic purpose of data processing is to extract and derive valuable and meaningful data for certain specific people from a large amount of data that may be disorganized and difficult to understand.

[0022] Deep Learning (DL) is a new research direction in the field of machine learning. Deep learning is to learn the inherent laws and representation levels of sample data. The information obtained in the learning process is very helpful for interpreting data such as text, images and sounds. Its ultimate goal is to enable machines to have analytical learning capabilities like humans and to recognize data such as text, images and sounds.

[0023] Natural Language Processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies various theories and methods that can achieve effective communication between people and computers using natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field will involve natural language, that is, the language people use in daily life, so it is closely related to the study of linguistics, but there are important differences. Natural language processing is not a general study of natural language, but the development of computer systems that can effectively realize natural language communication.

[0024] Artificial Intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. Artificial Intelligence is a branch of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. Since the birth of artificial intelligence, the theory and technology of artificial intelligence have become increasingly mature, and the application field has continued to expand. It can be imagined that the technological products brought by artificial intelligence in the future will be the "container" of human wisdom. Artificial Intelligence can simulate the information process of human consciousness and thinking.

[0025] Figure 1 FIG. 1 is a flow chart of a method for processing search results according to an embodiment of the present disclosure. Figure 1 As shown, the method includes:

[0026] S101, obtaining a search question and a reference question and answer library, wherein the reference question and answer library includes a reference question index and a reference answer library.

[0027] In the disclosed embodiment, a search question may be input through an interactive interface of a user terminal, wherein search information input by the user terminal may be obtained, and a question may be extracted from the search information to obtain a search question to be searched.

[0028] Optionally, a question-and-answer library is provided in the server, and a plurality of questions and reference answers to the plurality of questions are stored in the question-and-answer library. In this scenario, the question-and-answer library may be marked as a reference question-and-answer library.

[0029] The reference question and answer library is provided with a reference question index storing reference questions, and a reference answer library storing reference answers to each reference question.

[0030] S102, based on the reference question index, obtain a set of related reference questions of the search question, and obtain the related reference answers of each set of related reference questions in the reference answer library, and input each related reference answer into the pre-acquired large model to obtain the target search result of the search question.

[0031] In the embodiment of the present disclosure, there are multiple reference questions in the reference question index. After obtaining the search question, it is possible to filter and identify the reference question in the reference question index to obtain at least one reference question that is associated with the search question as the associated reference question of the search question, and mark the set of the at least one associated reference question as a set of associated reference questions for the search question.

[0032] Optionally, a reference answer to each associated reference question may be obtained from a reference answer library according to the associated reference question set as an associated reference answer to each associated reference question.

[0033] In this scenario, the associated reference answers to each associated reference question can be understood as reference answers that are associated with the search question. Based on this part of the associated reference answers, the search results of the search question in the reference question and answer library can be obtained, which can be marked as target search results.

[0034] Optionally, the associated reference answers of each associated reference question may be input into a pre-acquired large model, and the associated reference answers may be integrated by calling the text integration capability in the model capability of the large model, and then the target search result of the search question may be obtained based on the integrated result.

[0035] The large model may be a Wenxin large model, or other large models that can achieve accurate integration of various related reference answers, and no specific limitation is made here.

[0036] S103, inputting the target search result into the large model, and identifying through the large model whether the target search result meets the preset decomposition condition.

[0037] In the disclosed embodiment, after obtaining the target search result of the search question, there is a possibility of decomposing the target search result into steps. In this scenario, a preset decomposition condition can be obtained, and whether the target search result satisfies the decomposition condition can be identified.

[0038] Optionally, the search results may be input into a pre-acquired large model, and the model capability of the large model may be used to identify whether the target search results meet the decomposition conditions.

[0039] As an example, the big model is set as the Wenxin big model, and the target search result can be input into the Wenxin big model, and the target search result is analyzed by the Wenxin big model to determine whether the target search result meets the preset decomposition condition.

[0040] S104, in response to identifying that the target search result meets the preset decomposition condition, the target search result is decomposed into steps through the large model to obtain a decomposed single step set.

[0041] In the disclosed embodiment, the large model can identify whether the target search results include operable individual operation steps. If it is identified that the target search results include operable individual operation steps, it can be determined that the target search results meet the preset decomposition conditions.

[0042] In this scenario, the large model can decompose the target search results, decompose the individual operational steps included therein, and mark the decomposed steps as single steps, thereby obtaining a single step set in the target search results.

[0043] S105, executing each single step in the single step set in sequence through the large model, obtaining the target execution result of the single step set, and returning the target execution result to the user end.

[0044] In the embodiment of the present disclosure, there is a set execution order between the single steps included in the target search results. As an example, for the single step 1, single step 2 and single step 3 included in the single step set, the output result obtained by executing single step 1 can be set as the input of single step 2, and the output result obtained by executing single step 2 can be set as the input of single step 3. In this example, the execution order between single step 1, single step 2 and single step 3 is single step 1→single step 2→single step 3.

[0045] Optionally, after determining the execution order between the single steps, the large model can execute the single steps one by one based on the execution order, and determine the result obtained by the last single step executed as the target execution result obtained by executing the set of single steps in sequence.

[0046] In this scenario, the target execution result can be used as the return information of the user end.

[0047] It can be understood that after the user inputs the search question, the target search results of the search question can be obtained from the reference question and answer library, and the target search results can be input into the pre-acquired large model. The model capability of the large model can be used to identify whether the target search results meet the decomposition conditions. After identifying that the target search results meet the decomposition conditions, the target search results can be decomposed into steps through the large model to obtain a single step set, and then each single step in the single step set can be autonomously executed in sequence.

[0048] The method for processing search results proposed in the present disclosure obtains a search question and a reference question and answer library, obtains a set of related reference questions of the search question according to a reference question index in the reference question and answer library, obtains related reference answers of each related reference question from a reference answer library in the reference question and answer library, and inputs each related reference answer into a large model, thereby obtaining a target search result of the search question. Optionally, the target search result is input into the large model, and when the large model recognizes that the target search result meets a preset decomposition condition, the target search result is decomposed into steps based on the large model to obtain a single step set, and then the target execution result is obtained by sequentially executing each single step in the single step set through the large model, and then the target execution result is returned to the user end. In the present disclosure, each associated reference answer of the search question is obtained by referring to the question and answer library, and each associated reference answer is integrated through the big model to obtain the target search result of the search question, and the target search result is decomposed and automatically executed through the big model. By calling the text integration capability of the big model, the efficiency of obtaining the target search result is improved, and the integration quality of each associated reference answer is optimized. The big model is used to identify whether the target search result meets the decomposition condition, and the accuracy of the decomposition condition identification is improved. When it is identified that the target search result meets the decomposition condition, the model capability of the big model is called to decompose the target search result, which improves the decomposition efficiency and decomposition accuracy of the target search result. Then, a single step set is autonomously executed through the big model, without the need for users to operate the system themselves, which reduces the difficulty of users' operations and improves the execution efficiency of a single step set. There is no need to rely on manual customer service, which reduces the probability of problem backlog on the manual customer service end, improves the accuracy and efficiency of system operation, and optimizes the user experience.

[0049] In the above embodiment, the acquisition of target search results can be combined with Figure 2 Further understanding, Figure 2 FIG. 1 is a flow chart of a method for processing search results according to another embodiment of the present disclosure. Figure 2 As shown, the method includes:

[0050] S201, obtaining a search question and a reference question and answer library, wherein the reference question and answer library includes a reference question index and a reference answer library.

[0051] Optionally, historical question and answer text and a historical question and answer library may be obtained, and the historical question and answer text may be input into a large model to obtain an updated question and answer text based on the historical question and answer library, wherein a historical question index and a historical answer library in the historical question and answer library are obtained.

[0052] In the disclosed embodiment, historical question and answer texts may be captured based on a preset text acquisition path.

[0053] As an example, Figure 3 As shown, it can be Figure 3 The knowledge extraction module shown is from Figure 3 The historical question and answer text is captured from the user space historical text and the customer service log text, thereby obtaining the historical question and answer text.

[0054] like Figure 3 As shown, the knowledge extraction module can be Figure 3 The extraction capability of the large model shown enables the capture of user space history text and customer service log text, where: Figure 3 The large model shown may be a Wenxin large model.

[0055] In the disclosed embodiment, the reference question and answer library that needs to be updated may be marked as a historical question and answer library, wherein the historical question and answer library includes a historical question index and a historical answer library.

[0056] In this scenario, the question information in the historical question and answer text can be compared with the historical question index, and the answer information in the historical question and answer text can be compared with the historical answer library to obtain the part of the historical question and answer text that is different from the historical question and answer library, and the part of the question and answer text with the difference can be marked as the updated question and answer text of the historical question and answer text based on the historical question and answer library.

[0057] Optionally, the historical question index and the historical answer library are updated based on the updated question and answer text to obtain an updated reference question index and reference answer library.

[0058] Among them, updated questions and updated answers to the updated questions in the updated question and answer text can be extracted.

[0059] In the disclosed embodiment, the question part in the updated question and answer text can be extracted through the extraction capability of the large model, and the extracted question part can be marked as an updated question.

[0060] Furthermore, the answer part in the updated question and answer text is extracted according to the updated question, thereby obtaining the answer to the updated question and marking it as the updated answer.

[0061] As an example, Figure 3 As shown, through Figure 3 The extraction module shown is Figure 3 The extraction capability of the large model shown is called to extract the updated questions and answers to the updated questions in the updated question and answer text, thereby obtaining Figure 3 The extraction results are shown.

[0062] like Figure 3 As shown, the extraction results include updated question 1, updated question 2 and updated question 3, as well as updated answer 1 of updated question 1, updated answer 2 of updated question 2 and updated answer 3 of updated question 3.

[0063] Optionally, the historical question index is updated based on the updated question to obtain an updated reference question index.

[0064] In the disclosed embodiment, the historical question index can be modified by updating the question, wherein the updated question can be added to the historical question index, or the historical question already existing in the historical question index can be modified according to the updated question, thereby updating the historical question index and obtaining an updated reference question index.

[0065] The update problem can be vectorized to obtain an update problem vector of the update problem.

[0066] Optionally, the update problem may be algorithmically processed based on a vectorization algorithm in related technologies, and then a vector of the update problem may be obtained according to a result of the algorithm processing, and the vector may be marked as an update problem vector.

[0067] like Figure 3 As shown, it is possible to Figure 3 The update problem 1 in the extraction result shown is vectorized to obtain the update problem vector 1 of the update problem 1<d11,d12,d13> , and vectorize the update problem 2 in the extraction result to obtain the update problem vector 2 of the update problem 2<d21,d22,d23> , we can also vectorize the update problem 3 in the extraction result, and then obtain the update problem vector 3 of the update problem 3<d31,d32,d33> .

[0068] Furthermore, hash processing is performed on the update question vector to obtain a fourth identifier of the update question vector.

[0069] In the disclosed embodiment, the update question has an identifier associated with it, and the update answer to the update question has an identifier associated with the identifier of the update question. In this scenario, the update question vector can be hashed and the string obtained by the hashing process can be used as the fourth identifier of the update question vector.

[0070] Optionally, index update information for updating the question vector is constructed according to the fourth identifier, and the index update information is added to the historical question index to obtain an updated reference question index.

[0071] In the embodiment of the present disclosure, the fourth identifier may be linked to the update question vector, so as to obtain the update question vector linked with the fourth identifier as the index update information of the update question.

[0072] As an example, Figure 3 As shown, set Figure 3Problem 1 is shown as updated problem vector 1<d11,d12,d13> The fourth identifier, such as Figure 3 As shown, it can be<d11,d12,d13> Linked with the Question 1 ID, thus obtaining the index update information 1 for updating Question 1.<d11,d12,d13> Question 1 Identification".

[0073] Accordingly, the index update information 2 of update question 2 can be obtained by referring to the method for obtaining the index update information 1 of update question 1.<d21,d22,d23> Issue 2 Identification", and update issue 3 index update information 3"<d31,d32,d33> Question 3: Identification” will not be repeated here.

[0074] Furthermore, the index update information of the updated question can be added to the historical question index to implement the update of the historical question index by the updated question, and the updated historical question index can be marked as a reference question index.

[0075] As an example, index update information 1, index update information 2, and index update information 3 may be added to the historical question index, such as Figure 3 As shown, subspaces are set under the historical question index, and different index update information can be added to different subspaces.

[0076] like Figure 3 As shown, it is assumed that the correlation between update question 1 and update question 3 satisfies the condition of being placed in the same subspace, and the correlation between update question 2 and update question 1 and update question 3 does not satisfy the condition of being placed in the same subspace. In this example, the index update information 1 of update question 1 can be updated.<d11,d12,d13> Issue 1 Identifier" and update issue 3 index update information 3"<d31,d32,d33> Issue 3 logo" added to Figure 3 The subspace 1 shown will update the index update information 2 of question 2"<d21,d22,d23> Question 2 logo" added to Figure 3 The subspace 2 shown in FIG. 1 further implements the updating of the historical question index by updating question 1, updating question 2 and updating question 3 to obtain the updated reference question index.

[0077] Optionally, the historical answer library is updated based on the updated answers to obtain an updated reference answer library.

[0078] In the disclosed embodiments, part of the answers in the historical question and answer library are stored in the historical answer library. In some implementations, updated answers can be added to the historical answer library to update the historical answer library. In other implementations, corresponding answers in the historical answer library can be modified according to the updated answers to update the historical answer library.

[0079] Optionally, in response to identifying that the updated answer does not exist in the historical answer library, a fifth identifier matching the fourth identifier is obtained, and the fifth identifier is linked to the updated answer to obtain linked answer update information, and the answer update information is added to the historical answer library to obtain an updated reference answer library.

[0080] In the disclosed embodiment, the updated answer may be matched with each historical answer in the historical answer library to identify whether there is an answer associated with the updated answer in each historical answer.

[0081] Optionally, the similarity between each historical answer and the updated answer may be obtained. For any historical answer, when the similarity is greater than or equal to a preset similarity threshold, it can be determined that the historical answer is an associated answer of the updated answer.

[0082] Since the updated answer is obtained through the answer part in the historical question and answer text that is different from the historical answers in the historical answer library, in this scenario, the associated answers of the updated answer in the historical answer library can be marked as historical answers to be updated.

[0083] In the scenario where it is identified that the updated answer does not exist in the historical answer library, a fifth identifier matching the fourth identifier can be obtained, and the fifth identifier can be linked to the updated answer to obtain answer update information of the updated answer.

[0084] The fifth identifier may be the same as the fourth identifier, or may be an identifier that has a mapping relationship with the fourth identifier, which is not limited here.

[0085] In this scenario, the obtained answer update information can be added to the historical answer library to obtain an updated reference answer library.

[0086] As an example, setting Figure 3 The question 1 shown is identified as the fifth identifier of the updated answer 1, the question 2 is identified as the fifth identifier of the updated answer 2, and the question 3 is identified as the fifth identifier of the updated answer 3. Then the question 1 identifier can be linked to the updated answer 1 to obtain Figure 3 Shown are answer update information 1 "Question 1 Identifier - Answer 1" for updating answer 1, answer update information 2 "Question 2 Identifier - Answer 2" for updating answer 2, and answer update information 3 "Question 3 Identifier - Answer 3" for updating answer 3.

[0087] Furthermore, answer update information 1, answer update information 2 and answer update information 3 may be added to the historical answer library, thereby obtaining an updated reference answer library.

[0088] Optionally, in response to identifying that there are historical answers to be updated in the historical answer library, the historical answers to be updated are modified according to the updated answers to obtain an updated reference answer library.

[0089] In a scenario where it is identified that an updated answer has a historical answer to be updated in the historical answer library, the difference between the updated answer and the historical answer to be updated can be obtained, and the difference between the historical answer to be updated can be modified to obtain an updated reference answer library.

[0090] Furthermore, based on the reference question index and the reference answer library, a reference question and answer library is obtained.

[0091] Optionally, the reference question index and the reference answer library are stored in respective preset storage locations, and the reference question index and the reference answer library can be stored separately based on the preset storage locations to obtain a reference question and answer library consisting of the reference question index and the reference answer library.

[0092] S202, obtaining the question similarity between each reference question in the reference question index and the search question.

[0093] As an example, Figure 4 As shown, the search question input by the user can be parsed to obtain keywords in the search question, and the keywords are vectorized to obtain a vector of the search question.

[0094] In an embodiment of the present disclosure, a vector search can be performed in a reference question index in a reference question and answer library based on the vector of the search question, wherein the similarity between the vector of the search question and the vectors of each reference question in the reference question index can be obtained based on a vector similarity algorithm in the relevant technology, and marked as the question similarity between the search question and each reference question.

[0095] S203: For any reference question, in response to a question similarity of the reference question being greater than or equal to a preset similarity threshold, determining the reference question as an associated reference question of the search question.

[0096] In an embodiment of the present disclosure, for any reference question, when the question similarity between the search question and the reference question is greater than or equal to a preset similarity threshold, it can be determined that the search question and the reference question are similar questions. In this scenario, it can be determined that the reference answer to the reference question can be used as the answer to the search question.

[0097] The reference question may be marked as an associated reference question of the search question in the reference question index.

[0098] like Figure 4 As shown, it can be Figure 4The vector retrieval module shown obtains the similarity between the vector of the search question and the vectors of each reference question, and then obtains the related reference questions of the search question from the reference question index according to the obtained similarities of each question.

[0099] Optionally, there may be at least one associated reference question for the search question in the reference question index, and a set consisting of the at least one associated reference question may be marked as a set of associated reference questions for the search question in the reference question index.

[0100] S204, obtaining the first identifier of each associated reference question, and obtaining the associated reference answer of each associated reference question from the reference answer library according to each first identifier, and inputting each associated reference answer into the big model to obtain the target search result of the search question.

[0101] In the disclosed embodiment, each reference question stored in the reference question index has identification information. In this scenario, the identification information of the associated reference question can be obtained from the reference question index and marked as the first identification of the associated reference question.

[0102] As an example, setting Figure 4 Shows the reference issue index information"<d11,d12,d13> Question 1 Logo”, vector<d11,d12,d13> The corresponding question 1 is the associated reference question of the search question in the reference question index, so the "question 1 identifier" in the question index information is the vector<d11,d12,d13> The corresponding association reference is the first identifier of question 1.

[0103] Correspondingly, each answer in the reference answer library has identification information, and the answer can be obtained from the reference answer library based on the first identification of each related reference question in the related reference question set, thereby obtaining the answer to each related reference question and marking it as the related reference answer to each related reference question.

[0104] Optionally, the second identifier of each reference answer in the reference answer library may be obtained.

[0105] In the disclosed embodiment, the identification information of each reference answer in the reference answer library may be marked as the second identification of the reference answer.

[0106] As an example, Figure 4 As shown, Figure 4 The reference answer library shown includes reference answer 1, reference answer 2 and reference answer 3, wherein the identification information of reference answer 1 is Figure 4 The "Question 1 Identification" shown, the identification information of reference answer 2 is Figure 4 The "Question 2 Identification" shown, the identification information of reference answer 3 is Figure 4"Question 3 Logo" is shown.

[0107] In this example, "Question 1 Identifier" is the second identifier of reference answer 1, "Question 2 Identifier" is the second identifier of reference answer 2, and "Question 3 Identifier" is the second identifier of reference answer 3.

[0108] Optionally, for any first identifier, a third identifier matching the first identifier is obtained from each second identifier, and a reference answer corresponding to the third identifier is determined as an associated reference answer to the associated reference question corresponding to the first identifier.

[0109] In the disclosed embodiment, for any first identifier, identification information matching the first identifier can be obtained from each second identifier as a third identifier, and the reference answer corresponding to the third identifier can be used as an associated reference answer to the associated reference question corresponding to the first identifier.

[0110] like Figure 4 As shown, reference question 1 in the reference question index is set as the associated reference question of the search question, wherein the first identifier of the associated reference question is "question 1 identifier".

[0111] If it is set that the "Question 1 Identifier" in each second identifier in the reference answer library matches the first identifier, then the "Question 1 Identifier" in the reference answer library can be marked as the third identifier, and the reference answer 1 corresponding to the third identifier can be determined as the associated reference answer to the associated reference question corresponding to the first identifier.

[0112] Optionally, the associated reference answers to the associated search questions are input into the big model for integration to obtain integrated target search results.

[0113] In the disclosed embodiment, the associated reference answers of each associated reference question can be input into the big model for text integration, wherein the big model can read the text information of each associated reference answer, determine the integration order between the text information, and splice the associated reference answers based on the integration order, thereby obtaining and outputting the spliced ​​result, and marking the result as the target search result of the search question.

[0114] The method for processing search results proposed in the present disclosure obtains a search question and a reference question and answer library, and obtains the question similarity between the search question and each reference question in the reference question index in the reference question and answer library, thereby obtaining a set of associated reference questions of the search question in the reference question index, and obtaining associated reference answers of each associated reference question from the reference answer library according to the first identifier of each associated reference question. Optionally, each associated reference answer is integrated through a large model to obtain a target search result of the search question. In the present disclosure, the construction of a reference question and answer library is realized based on historical question and answer texts, and the target search result of the search question is obtained from the reference question and answer library, which improves the efficiency of obtaining the target search result, and integrates each associated reference answer through the integration capability of the large model to obtain the target search result, which improves the integration efficiency and accuracy of the target search result, improves the completeness and accuracy of the target search result in answering the search question, optimizes the quality of the target search result, and thus optimizes the user's search experience.

[0115] In the above embodiment, the decomposition and automatic execution of the target search results can be combined with Figure 5 Further understanding, Figure 5 FIG. 1 is a flow chart of a method for processing search results according to another embodiment of the present disclosure. Figure 5 As shown, the method includes:

[0116] S501, in response to identifying that a target search result satisfies a preset decomposition condition, the target search result is decomposed into steps using a large model to obtain a decomposed single step set.

[0117] Optionally, in response to the large model recognizing that there are decomposable operation steps in the target search result, it is determined that the target search result meets the decomposition condition.

[0118] In the disclosed embodiment, the target search result may include multiple operation steps, wherein the multiple operation steps may be marked as decomposable operation steps.

[0119] In this scenario, the target search results can be input into the big model, and the big model can be used to identify whether there are decomposable operation steps in the target search results. When the big model identifies that there are decomposable operation steps in the target search results, it can be determined that the target search results meet the decomposition conditions.

[0120] As an example, it is assumed that the target search results include statement 1, statement 2 and statement 3, wherein statement 1, statement 2 and statement 3 are respectively guiding statements for system operation, and statement 1, statement 2 and statement 3 are different independent operation steps. In this example, statement 1, statement 2 and statement 3 can be determined as decomposable operation steps in the target search results.

[0121] In this scenario, it can be determined that the target search result meets the preset decomposition condition.

[0122] Optionally, according to the decomposable operation steps, the target search results are decomposed into steps through a large model to obtain a decomposed single step set.

[0123] In the disclosed embodiment, the decomposable operation steps included in the target search results can be decomposed through the model capabilities of the large model, and each step in the decomposable operation steps can be decomposed into an independent single step, thereby obtaining a single step set after the target search results.

[0124] Based on the above example, the decomposable operation steps in the target search results are set to step 1 corresponding to statement 1, step 2 corresponding to statement 2, and step 3 corresponding to statement 3. In this example, the model capability of the large model can be used to decompose step 1 corresponding to statement 1 in the target search results into an independent single step, and based on the same decomposition method, step 2 corresponding to statement 2 and step 3 corresponding to statement 3 can be decomposed into independent single steps, thereby obtaining a set of single steps in the target search results.

[0125] It can be understood that the target search result "sentence 1 sentence 2 sentence 3" is decomposed into:

[0126] "Single Step 1: Statement 1

[0127] Single Step 2: Statement 2

[0128] Single Step 3: Statement 3"

[0129] Then a single step set of target search results is obtained.

[0130] It should be noted that the target search result may not satisfy the preset decomposition condition. In response to the large model identifying that the target search result does not satisfy the decomposition condition, the target search result is returned to the user end.

[0131] In the disclosed embodiment, when it is recognized through the model capability of the large model that there are no decomposable operation steps in the target search results, it can be determined that the target search results do not meet the preset decomposition conditions.

[0132] As an example, the target search result is set to be a text of the knowledge introduction type. This type of text may not contain content of the operation steps that can perform system operations. In this scenario, it can be determined that there are no decomposable operation steps in the target search result.

[0133] Optionally, the target search result may be used as the final result returned to the user end.

[0134] S502, obtaining the target execution plug-in of each single step from the pre-configured candidate plug-in library through the large model.

[0135] In the disclosed embodiment, before executing each single step, an execution plug-in can be selected from a candidate plug-in library for each single step through the model capability of the large model and marked as a target execution plug-in for each single step.

[0136] The candidate plug-in library may include multiple candidate plug-ins, such as Figure 4 As shown, candidate plugin libraries may include Figure 4 The summary plug-in shown as a candidate plug-in, the monitoring plug-in as a candidate plug-in, and the notification plug-in as a candidate plug-in may also include candidate plug-ins with other functions, which are not specifically limited here.

[0137] As an example, Figure 4 As shown, it can be Figure 4 The large model shown decomposes the target search results into steps, wherein the target search results can be input into Figure 4 The large model shown decomposes the target search results through the decomposition capability of the large model, thereby obtaining a single step set in the target search results.

[0138] Furthermore, the large model is used to match the plug-ins for each single step. Figure 4 The target execution plug-in of each single step is obtained from the candidate plug-in library shown.

[0139] S503, obtaining the execution order of each single step.

[0140] In the embodiment of the present disclosure, there is a set execution order between each single step.

[0141] As an example, it is assumed that the target search results include single step 1, single step 2 and single step 3, wherein the output of single step 1 is the input of single step 2, and the output of single step 2 is the input of single step 3.

[0142] In this example, it can be determined that the execution order among single step 1, single step 2 and single step 3 is that single step 2 is executed after single step 1 is executed, and single step 3 is executed after single step 2 is executed.

[0143] S504, based on the execution order, each target execution plug-in is controlled by the large model to execute each single step in sequence, obtain the target execution result of the single step set, and return the target execution result to the user end.

[0144] Optionally, for any single step, in response to the target execution plug-in of the single step having missing function parameters, the missing parameter item is obtained and returned to the user end.

[0145] In the embodiment of the present disclosure, for any single step, it is possible to identify whether the target execution plug-in of the single step has missing running function parameters based on the plug-in matching function in the related art.

[0146] When it is identified that the running function corresponding to the target execution plug-in is missing a running function parameter, the missing parameter item in the running function can be obtained, and the obtained missing parameter item can be returned to the user end through the large model.

[0147] Optionally, in response to the user inputting filling parameters for missing parameter items, the target execution plug-in is run through the large model to obtain the step execution result of a single step, so as to obtain the target execution result of a single step set.

[0148] In the embodiment of the present disclosure, after receiving the missing parameter item, the user terminal can display it on the interactive interface and guide the user to fill in the missing parameter item on the interactive interface, and mark the parameters filled in by the user under the missing parameter item as the filling parameters of the missing parameter item.

[0149] like Figure 4 As shown, Figure 4 The large model shown also has Figure 4 The function of the context cache shown can be understood as that the target execution plug-in can perform multiple rounds of function execution during the execution of a single step. During this execution process, there may be two or more rounds of function execution with missing function parameters.

[0150] In this scenario, among the missing parameter items of function 2 of the target execution plug-in, some missing parameter items can be obtained from the running function parameters of function 1. These missing parameter items can be marked as the first part of missing parameter items, and the remaining items of the missing parameter items of function 2 except the first part of missing parameter items can be marked as the second part of missing parameter items.

[0151] In this scenario, the large model can be Figure 4 The function of the context cache shown obtains filling parameters for the first part of missing parameter items from the running function parameters of function 1, and returns the second part of missing parameter items to the user end, obtains the filling parameters input by the user through the user end, and then obtains the filling parameters of all the missing parameter items of function 2 according to the filling parameters of the first part of missing parameter items and the filling parameters of the second part of missing parameter items, so as to realize the running of function 2 by the target execution plug-in.

[0152] Optionally, after the user inputs the filling parameters of the missing parameter items, they can be transmitted to the corresponding target execution plug-in. In this scenario, the large model can realize the execution of its corresponding single step by controlling the operation of the target execution plug-in, and mark the result of the execution as the step execution result of the single step.

[0153] It should be noted that the target execution plug-in may not have a situation where the running function parameters are missing. In this scenario, the corresponding single step can be executed by the target execution plug-in, and then the step execution result of the single step corresponding to the target execution plug-in can be obtained.

[0154] Optionally, a target execution result of a set of single steps is obtained according to the step execution results of each single step.

[0155] Among them, the target execution result of the single step set can be obtained based on the step execution result of the last single step executed in the single step set, or can be obtained by integrating the step execution results of multiple single steps, which is not specifically limited here.

[0156] Furthermore, the target execution result of the acquired single step set is returned to the user end.

[0157] The processing method of search results proposed in the present disclosure inputs the target search results into a large model, and when it is identified that the target search results meet the preset decomposition conditions, the target search results are decomposed into steps through the large model to obtain a single step set, and the target execution plug-in of each single step is obtained from the candidate plug-in library through the large model. Optionally, the large model controls each target execution plug-in to execute in sequence according to the execution order between each single step to obtain the target execution result of the single step set, and returns it to the user end. In the present disclosure, each single step is executed based on the model capability of the large model, which improves the execution efficiency of each single step, does not require the user to perform system operations by himself, and reduces the user's operation difficulty. When there is a situation where the running function parameters are missing during the execution of the target execution plug-in, the missing parameter items are returned to the user end to obtain the filling parameters, which improves the execution flexibility of the single step set and the accuracy of the target execution result, reduces the user's operation difficulty, and optimizes the user's experience.

[0158] To better understand the above embodiments, Figure 6 , Figure 6 FIG. 1 is a flow chart of a method for processing search results according to another embodiment of the present disclosure. Figure 6 As shown, the method includes:

[0159] like Figure 6As shown in the figure, the model capability of the large model can be used to extract updated question and answer texts from historical question and answer texts, and the historical question and answer library can be updated according to the extracted updated question and answer texts to obtain Figure 6 The reference question and answer library is shown.

[0160] like Figure 6 As shown, the input information is received through the user end, and Figure 6 The demand parsing module shown includes a request processing layer and a request parsing layer, obtains the search question from the input information, and obtains the related reference answers to the search question from the reference question and answer library, and inputs the related reference answers into the big model. The related reference answers are integrated through the model capabilities of the big model to obtain the target search results of the search question.

[0161] Optionally, the target search result is input into the large model, and the model capability of the large model is used to identify whether the target search result satisfies the decomposition submission. When it is identified that the target search result satisfies the decomposition conditions, the target search result is decomposed into steps to obtain a single step set of the target search result.

[0162] Furthermore, large models can be obtained from Figure 6 In the candidate plug-in library shown, which includes screenshot plug-ins, monitoring collection plug-ins, and intervention plug-ins, the target execution plug-in of each single step is obtained, and each target execution plug-in is controlled to execute each single step in sequence, thereby obtaining the target execution result of the single step set, and returning the target execution result to the user end.

[0163] The search result processing method proposed in the present invention obtains various related reference answers to the search questions by referring to the question and answer library, and integrates the various related reference answers through a large model to obtain a target search result, and then decomposes and automatically executes the target search result through the large model, thereby improving the integration efficiency of the related reference answers and the quality of the target search results obtained after integration, improving the decomposition efficiency of the target search results and the automatic execution efficiency of each single step, without the need for users to operate the system themselves, reducing the difficulty of user operation, without relying on manual customer service, reducing the probability of problem backlog on the manual customer service end, improving the accuracy and efficiency of system operation, and optimizing the user experience.

[0164] Corresponding to the search result processing methods proposed in the above-mentioned embodiments, an embodiment of the present disclosure further proposes a search result processing device. Since the search result processing device proposed in the embodiment of the present disclosure corresponds to the intention degree prediction + model training method proposed in the above-mentioned embodiments, the implementation method of the above-mentioned search result processing method is also applicable to the search result processing device proposed in the embodiment of the present disclosure, and will not be described in detail in the following embodiments.

[0165] Figure 7 FIG. 1 is a schematic diagram of a structure of a search result processing device according to an embodiment of the present disclosure. Figure 7 As shown, the search result processing device 700 includes a first acquisition module 71, a second acquisition module 72, an identification module 73, a decomposition module 74 and an execution module 75, wherein:

[0166] A first acquisition module 71 is used to acquire a search question and a reference question and answer library, wherein the reference question and answer library includes a reference question index and a reference answer library;

[0167] A second acquisition module 72 is used to acquire a set of related reference questions of the search question based on the reference question index, and to acquire the related reference answers of the related reference questions in the reference answer library, and to input the related reference answers into the pre-acquired macro model to obtain the target search result of the search question;

[0168] An identification module 73, used to input the target search result into the large model, and identify whether the target search result meets a preset decomposition condition through the large model;

[0169] A decomposition module 74, configured to, in response to identifying that the target search result satisfies a preset decomposition condition, decompose the target search result into steps to obtain a decomposed single step set;

[0170] The execution module 75 is used to execute each single step in the single step set in sequence, obtain the target execution result of the single step set, and return the target execution result to the user end.

[0171] In the disclosed embodiment, the decomposition module 74 is further configured to: in response to identifying that the target search result does not satisfy the decomposition condition, return the target search result to the user end.

[0172] In the disclosed embodiment, the second acquisition module 72 is further used to: obtain the question similarity between each reference question in the reference question index and the search question; for any reference question, in response to the question similarity of the reference question being greater than or equal to a preset similarity threshold, determine that the reference question is an associated reference question of the search question; obtain a first identifier of each associated reference question, and obtain an associated reference answer for each associated reference question from a reference answer library based on each first identifier, so as to obtain a target search result for the search question.

[0173] In the disclosed embodiment, the second acquisition module 72 is also used to: obtain a second identifier for each reference answer in the reference answer library; for any first identifier, obtain a third identifier that matches the first identifier from each second identifier, and determine the reference answer corresponding to the third identifier as the associated reference answer to the associated reference question corresponding to the first identifier; integrate the associated reference answers to each associated reference question to obtain an integrated target search result.

[0174] In the disclosed embodiment, the first acquisition module 71 is also used to: obtain historical question and answer text and a historical question and answer library, and obtain an updated question and answer text of the historical question and answer text based on the historical question and answer library, wherein a historical question index and a historical answer library in the historical question and answer library are obtained; based on the updated question and answer text, the historical question index and the historical answer library are updated to obtain an updated reference question index and a reference answer library; based on the reference question index and the reference answer library, a reference question and answer library is obtained.

[0175] In the disclosed embodiment, the first acquisition module 71 is also used to: extract updated questions and updated answers to the updated questions in the updated question and answer text; update the historical question index based on the updated questions to obtain an updated reference question index; and update the historical answer library based on the updated answers to obtain an updated reference answer library.

[0176] In the disclosed embodiment, the first acquisition module 71 is also used to: vectorize the update problem to obtain an update problem vector of the update problem; hash the update problem vector to obtain a fourth identifier of the update problem vector; construct index update information of the update problem vector based on the fourth identifier, and add the index update information to the historical problem index to obtain an updated reference problem index.

[0177] In the disclosed embodiment, the first acquisition module 71 is also used for: in response to identifying that the updated answer does not exist in the historical answer library, obtaining a fifth identifier that matches the fourth identifier, and linking the fifth identifier with the updated answer to obtain the linked answer update information; adding the answer update information to the historical answer library to obtain an updated reference answer library.

[0178] In the disclosed embodiment, the first acquisition module 71 is further used for: in response to identifying that there are historical answers to be updated in the historical answer library, modifying the historical answers to be updated according to the updated answers to obtain an updated reference answer library.

[0179] In the disclosed embodiment, the first acquisition module 71 is also used to: extract historical questions from the historical question and answer text; in response to the existence of related historical questions in the historical question index, obtain historical answers to the historical questions; obtain related historical answers to the related historical questions from the historical answer library; in response to the historical answers being different from the related historical answers, obtain an updated question and answer text based on the historical questions and the historical answers.

[0180] In the disclosed embodiment, the first acquisition module 71 is further used for: in response to the absence of an associated historical question in the historical question index, obtaining an updated question and answer text according to the historical question and the historical answer.

[0181] In the embodiment of the present disclosure, the execution module 75 is also used to: obtain the target execution plug-in of each single step from the preconfigured candidate plug-in library; obtain the execution order of each single step; based on the execution order, execute each single step in sequence through each target execution plug-in to obtain the target execution result of the single step set, and return the target execution result to the user end.

[0182] In the disclosed embodiment, the execution module 75 is also used to: for any single step, in response to the existence of missing running function parameters in the target execution plug-in of the single step, obtain the missing parameter items and return the missing parameter items to the user end; run the target execution plug-in based on the filling parameters of the missing parameter items input by the user end to obtain the step execution result of the single step, so as to obtain the target execution result of the single step set.

[0183] In the disclosed embodiment, the execution module 75 is further used to: in response to the presence of decomposable operation steps in the target search results, determine that the target search results meet the decomposition conditions; and decompose the target search results according to the decomposable operation steps to obtain a single step set after decomposition.

[0184] The search result processing device proposed in the present disclosure obtains a search question and a reference question and answer library, obtains a set of related reference questions of the search question according to the reference question index in the reference question and answer library, obtains related reference answers of each related reference question from the reference answer library in the reference question and answer library, and inputs each related reference answer into a large model, thereby obtaining a target search result of the search question. Optionally, the target search result is input into the large model, and when the large model recognizes that the target search result meets a preset decomposition condition, the target search result is decomposed into steps based on the large model to obtain a single step set, and then the target execution result is obtained by sequentially executing each single step in the single step set through the large model, and then the target execution result is returned to the user end. In the present disclosure, each associated reference answer of the search question is obtained by referring to the question and answer library, and each associated reference answer is integrated through the big model to obtain the target search result of the search question, and the target search result is decomposed and automatically executed through the big model. By calling the text integration capability of the big model, the efficiency of obtaining the target search result is improved, and the integration quality of each associated reference answer is optimized. The big model is used to identify whether the target search result meets the decomposition condition, and the accuracy of the decomposition condition identification is improved. When it is identified that the target search result meets the decomposition condition, the model capability of the big model is called to decompose the target search result, which improves the decomposition efficiency and decomposition accuracy of the target search result. Then, a single step set is autonomously executed through the big model, without the need for users to operate the system themselves, which reduces the difficulty of users' operations and improves the execution efficiency of a single step set. There is no need to rely on manual customer service, which reduces the probability of problem backlog on the manual customer service end, improves the accuracy and efficiency of system operation, and optimizes the user experience.

[0185] According to an embodiment of the present disclosure, the present disclosure also proposes an electronic device, a readable storage medium, and a computer program product.

[0186] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0187] like Figure 8As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 809 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0188] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 806, such as various types of displays, speakers, etc.; a storage unit 809, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0189] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as the processing method of search results. For example, in some embodiments, the processing method of search results may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 809. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the processing method of the search results described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the processing method of search results in any other appropriate manner (e.g., by means of firmware).

[0190] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0191] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be presented to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0192] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0193] To propose interactions with a user account, the systems and techniques described herein may be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user account; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user account can propose input to the computer. Other types of devices may also be used to propose interactions with the user account; for example, feedback proposed to the user account may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user account may be received in any form (including acoustic input, voice input, or tactile input).

[0194] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user account computer with a graphical user account interface or a web browser through which a user account can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0195] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0196] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0197] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for processing search results, wherein: The method comprises: Obtaining a search question and a reference question and answer library, wherein the reference question and answer library includes a reference question index and a reference answer library; Obtaining question similarity between each reference question in the reference question index and the search question; For any reference question, in response to the question similarity of the reference question being greater than or equal to a preset similarity threshold, determining the reference question as an associated reference question of the search question; Obtaining a second identifier for each reference answer in the reference answer library; For any first identifier, obtaining a third identifier matching the first identifier from each second identifier, and determining a reference answer corresponding to the third identifier as an associated reference answer to an associated reference question corresponding to the first identifier; Input the related reference answers of each related reference question into the big model for integration to obtain the integrated target search results; Inputting the target search result into the large model, and identifying whether the target search result meets a preset decomposition condition through the large model; In response to identifying that the target search result satisfies the decomposition condition, performing step decomposition on the target search result by using the large model to obtain a decomposed single step set; The large model is used to sequentially execute each single step in the single step set, obtain a target execution result of the single step set, and return the target execution result to the user end.

2. The method according to claim 1, wherein: The method further comprises: In response to identifying, through the large model, that the target search result does not satisfy the decomposition condition, the target search result is returned to the user terminal.

3. The method according to claim 1, wherein: The obtaining of the reference question and answer library, wherein the reference question and answer library includes a reference question index and a reference answer library, includes: Acquire a historical question and answer text and a historical question and answer library, and input the historical question and answer text into the large model to obtain an updated question and answer text of the historical question and answer text based on the historical question and answer library, wherein a historical question index and a historical answer library in the historical question and answer library are acquired; Based on the updated question and answer text, the historical question index and the historical answer library are updated to obtain the updated reference question index and the reference answer library; Based on the reference question index and the reference answer library, the reference question and answer library is obtained.

4. The method according to claim 3, wherein: The updating of the historical question index and the historical answer library based on the updated question and answer text to obtain the updated reference question index and the reference answer library includes: Extracting updated questions and updated answers to the updated questions from the updated question and answer text; The historical question index is updated based on the updated question to obtain the updated reference question index; The historical answer library is updated based on the updated answer to obtain the updated reference answer library.

5. The method according to claim 4, wherein: The updating of the historical question index based on the updated question to obtain the updated reference question index includes: Vectorizing the update problem to obtain an update problem vector for the update problem; Performing hash processing on the update question vector to obtain a fourth identifier of the update question vector; The index update information of the updated question vector is constructed according to the fourth identifier, and the index update information is added to the historical question index to obtain the updated reference question index.

6. The method according to claim 5, wherein: The updating of the historical answer library based on the updated answer to obtain the updated reference answer library includes: In response to identifying that the updated answer does not have a to-be-updated historical answer in the historical answer library, obtaining a fifth identifier matching the fourth identifier, and linking the fifth identifier with the updated answer to obtain linked answer update information; The answer update information is added to the historical answer library to obtain the updated reference answer library.

7. The method according to claim 6, wherein: The method further comprises: In response to identifying that the updated answer contains the to-be-updated historical answer in the historical answer library, the to-be-updated historical answer is modified according to the updated answer to obtain the updated reference answer library.

8. The method according to claim 3, wherein: The step of obtaining a historical question and answer text and a historical question and answer library, and inputting the historical question and answer text into the large model to obtain an updated question and answer text of the historical question and answer text based on the historical question and answer library, wherein obtaining a historical question index and a historical answer library in the historical question and answer library includes: Extracting historical questions from the historical question-and-answer text by using the large model; In response to the historical question having a related historical question in the historical question index, obtaining a historical answer to the historical question; Acquire the associated historical answers to the associated historical questions from the historical answer library; In response to the historical answer being different from the associated historical answer, the updated question and answer text is obtained based on the historical question and the historical answer.

9. The method according to claim 8, wherein: The method further comprises: In response to the historical question not having any associated historical question in the historical question index, the updated question-and-answer text is obtained according to the historical question and the historical answer.

10. The method according to claim 1, wherein: The step of sequentially executing each single step in the single step set by using the large model, obtaining a target execution result of the single step set, and returning the target execution result to the user end includes: Obtaining the target execution plug-in of each single step from a preconfigured candidate plug-in library through the large model; Get the execution order of each single step; Based on the execution order, each target execution plug-in is controlled by the large model to execute each single step in sequence, to obtain the target execution result of the single step set, and to return the target execution result to the user end.

11. The method according to claim 10, wherein: Based on the execution order, the large model is used to control each target execution plug-in to execute each single step in sequence to obtain the target execution result of the single step set, including: For any single step, in response to a missing running function parameter of a target execution plug-in of the single step, obtaining a missing parameter item and returning the missing parameter item to the user end; In response to the filling parameters of the missing parameter items input by the user end, the target execution plug-in is controlled to run through the large model to obtain the step execution result of the single step, so as to obtain the target execution result of the single step set.

12. The method according to claim 1, wherein: In response to identifying that the target search result satisfies the decomposition condition, the target search result is decomposed into steps by the large model to obtain a decomposed single step set, including: In response to the large model identifying that there are decomposable operation steps in the target search result, determining that the target search result satisfies the decomposition condition; According to the decomposable operation steps, the target search results are decomposed into steps using the large model to obtain the decomposed single step set.

13. A device for processing search results, wherein: The device comprises: A first acquisition module is used to acquire a search question and a reference question and answer library, wherein the reference question and answer library includes a reference question index and a reference answer library; A second acquisition module is used to obtain the question similarity between each reference question in the reference question index and the search question; for any reference question, in response to the question similarity of the reference question being greater than or equal to a preset similarity threshold, determine that the reference question is an associated reference question of the search question; obtain a second identifier of each reference answer in the reference answer library; for any first identifier, obtain a third identifier matching the first identifier from each second identifier, and determine the reference answer corresponding to the third identifier as an associated reference answer to the associated reference question corresponding to the first identifier; input the associated reference answers of each associated reference question into the large model for integration to obtain an integrated target search result; An identification module, used to input the target search result into the large model, and identify whether the target search result meets a preset decomposition condition through the large model; A decomposition module, configured to, in response to identifying that the target search result satisfies the decomposition condition, decompose the target search result into steps by using the large model to obtain a decomposed single step set; An execution module is used to execute each single step in the single step set in sequence through the large model, obtain a target execution result of the single step set, and return the target execution result to the user end.

14. The device according to claim 13, wherein: The decomposition module is further used for: In response to identifying, through the large model, that the target search result does not satisfy the decomposition condition, the target search result is returned to the user terminal.

15. The device according to claim 13, wherein: The first acquisition module is further used for: Acquire a historical question and answer text and a historical question and answer library, and input the historical question and answer text into the large model to obtain an updated question and answer text of the historical question and answer text based on the historical question and answer library, wherein a historical question index and a historical answer library in the historical question and answer library are acquired; Based on the updated question and answer text, the historical question index and the historical answer library are updated to obtain the updated reference question index and the reference answer library; Based on the reference question index and the reference answer library, the reference question and answer library is obtained.

16. The device according to claim 15, wherein: The first acquisition module is further used for: Extracting updated questions and updated answers to the updated questions from the updated question and answer text; The historical question index is updated based on the updated question to obtain the updated reference question index; The historical answer library is updated based on the updated answer to obtain the updated reference answer library.

17. The device according to claim 16, wherein: The first acquisition module is further used for: Vectorizing the update problem to obtain an update problem vector for the update problem; Performing hash processing on the update question vector to obtain a fourth identifier of the update question vector; The index update information of the updated question vector is constructed according to the fourth identifier, and the index update information is added to the historical question index to obtain the updated reference question index.

18. The device according to claim 17, wherein: The first acquisition module is further used for: In response to identifying that the updated answer does not have a to-be-updated historical answer in the historical answer library, obtaining a fifth identifier matching the fourth identifier, and linking the fifth identifier with the updated answer to obtain linked answer update information; The answer update information is added to the historical answer library to obtain the updated reference answer library.

19. The device according to claim 18, wherein: The first acquisition module is further used for: In response to identifying that the updated answer contains the to-be-updated historical answer in the historical answer library, the to-be-updated historical answer is modified according to the updated answer to obtain the updated reference answer library.

20. The device according to claim 15, wherein: The first acquisition module is further used for: Extracting historical questions from the historical question-and-answer text by using the large model; In response to the historical question having a related historical question in the historical question index, obtaining a historical answer to the historical question; Acquire the associated historical answers to the associated historical questions from the historical answer library; In response to the historical answer being different from the associated historical answer, the updated question and answer text is obtained based on the historical question and the historical answer.

21. The device according to claim 20, wherein: The first acquisition module is further used for: In response to the historical question not having any associated historical question in the historical question index, the updated question-and-answer text is obtained according to the historical question and the historical answer.

22. The device according to claim 13, wherein: The execution module is further used for: Obtaining the target execution plug-in of each single step from a preconfigured candidate plug-in library through the large model; Get the execution order of each single step; Based on the execution order, each target execution plug-in is controlled by the large model to execute each single step in sequence, to obtain the target execution result of the single step set, and to return the target execution result to the user end.

23. The device according to claim 22, wherein: The execution module is further used for: For any single step, in response to a missing running function parameter of a target execution plug-in of the single step, obtaining a missing parameter item and returning the missing parameter item to the user end; In response to the filling parameters of the missing parameter items input by the user end, the target execution plug-in is controlled to run through the large model to obtain the step execution result of the single step, so as to obtain the target execution result of the single step set.

24. The device according to claim 13, wherein: The execution module is further used for: In response to the large model identifying that there are decomposable operation steps in the target search result, determining that the target search result satisfies the decomposition condition; According to the decomposable operation steps, the target search results are decomposed into steps using the large model to obtain the decomposed single step set.

25. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 12.

26. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-12.

27. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 12.

Images