Methods, apparatus, electronic devices, storage media, and computer programs for identifying interaction information

Abstract

To provide a method of determining interaction information, a device, an electronic apparatus, a storage medium and a computer program which can create questioning dimensions for query information by combining large model previous specialized knowledge and subsequent supplements in user historical query information by means of the method of the chain of thoughts.SOLUTION: A method is to determine a plurality of questioning dimensions according to query information of a subject and historical query information, where each questioning dimension includes a dimension name and a plurality of options. The method is also to determine a target questioning dimension from the plurality of questioning dimensions according to evaluation values of the plurality of questioning dimensions and whether semantic information of the plurality of questioning dimensions are consistent with semantic information of a query result associated with the query information; and determine the interaction information according to the dimension name and the plurality of options in the target questioning dimension.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, particularly to fields such as large-scale models, generative models, NLP, intelligent search, etc. More specifically, the present disclosure provides a method, apparatus, electronic device, storage medium, and computer program for identifying interaction information.

Background Art

[0002] In complex search scenarios, due to the diversity and ambiguity of user expressions, the search engine cannot accurately capture the actual search intent, and there is a difference between the search results and the actual needs of the user.

Summary of the Invention

[0003] The present disclosure provides a method, apparatus, electronic device, storage medium, and computer program for identifying interaction information.

[0004] According to one aspect of the present disclosure, a method for identifying interaction information is provided, including identifying a plurality of question dimensions based on target query information and historical query information, where each question dimension includes a dimension name and a plurality of options; determining whether the semantic information of the multiple question dimensions matches the semantic information of the query results related to the query information; and identifying a target question dimension from the multiple question dimensions based on the evaluation values of the multiple question dimensions; and identifying interaction information based on the dimension name and the plurality of options in the target question dimension.

[0005] According to another aspect of this disclosure, an interaction information identification device is provided, comprising a first identification module, a second identification module, and a third identification module. The first identification module identifies a plurality of question dimensions based on target query information and historical query information, each question dimension including a dimension name and a plurality of choices. The second identification module identifies a target question dimension from the plurality of question dimensions based on whether the semantic information of the plurality of question dimensions matches the semantic information of the query results related to the query information, and based on the evaluation values ​​of the plurality of question dimensions. The third identification module identifies interaction information based on the dimension name and a plurality of choices in the target question dimension.

[0006] According to another aspect of the present disclosure, an electronic device is provided, comprising at least one processor and memory communicated with 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 perform the methods provided to the present disclosure.

[0007] According to another aspect of this disclosure, a non-temporary computer-readable storage medium is provided in which computer instructions are stored, thereby causing a computer to perform a method provided to this disclosure.

[0008] According to another aspect of this disclosure, a computer program is provided, and when the computer program is executed by a processor, the method provided in this disclosure is implemented.

[0009] It should be understood that the content described in this section is not intended to represent key points or important features of the embodiments of this disclosure, nor does it limit the scope of this disclosure. Other features of this disclosure will be readily apparent from the following description. [Brief explanation of the drawing]

[0010] The drawings are provided to better understand this technical proposal and do not limit the scope of this disclosure.

[0011] [Figure 1] Figure 1 is a schematic diagram illustrating an application scenario of the method and apparatus for identifying interaction information according to an embodiment of this disclosure. [Figure 2] Figure 2 is a schematic flowchart of the method for identifying interaction information according to the embodiment of this disclosure. [Figure 3A] Figure 3A is a schematic diagram illustrating the principle of the method for identifying interaction information according to an embodiment of this disclosure. [Figure 3B] Figure 3B is a schematic diagram of an interaction page according to an embodiment of the present disclosure. [Figure 4] Figure 4 is a schematic block diagram of the interaction information identification device according to an embodiment of the present disclosure. [Figure 5] Figure 5 is a structural block diagram of an electronic device that implements the method for identifying interaction information according to an embodiment of this disclosure. [Modes for carrying out the invention]

[0012] Illustrative embodiments of the present disclosure will be described below with reference to the drawings. Various details of the embodiments of the present disclosure are included herefor the sake of clarity, and should be considered as illustrative. Therefore, those skilled in the art will understand that various changes and modifications can 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 configurations will be omitted in the following description.

[0013] In the proposed technology disclosed herein, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information are all in accordance with the provisions of relevant laws and regulations and do not violate public order and morals.

[0014] In the proposed technologies described herein, the user's permission or consent was obtained before acquiring or collecting any user's personal information.

[0015] In complex search scenarios, the diversity and ambiguity of user expressions prevent search engines from accurately capturing the actual search intent, resulting in a gap between search results and the user's actual needs.

[0016] For example, the query "adjusting the body with herbal medicine" represents a relatively complex and ambiguous need. While this query relates to the user's preference for personalization, search engines return a wide range of query results related to keywords such as herbal medicine and body adjustment, often containing content that does not match the user's needs and failing to adequately satisfy their need for personalization. Some search engines recommend other possible query information to the user through related search methods, but such methods still make it difficult to discover the user's true needs, and irrelevant content is likely to be included.

[0017] Embodiments of this disclosure provide a method for identifying interaction information, which, through a thought chain method, can generate a question dimension for query information by combining large-scale model prior expertise with post-search supplementation in user history query information. By questioning the user, the method guides the user to provide more detailed information about the query objective, thereby mining the user's query needs. In this way, the search engine can provide query results that match the user's needs, improve the accuracy and relevance of the query results, and provide the user with a more personalized and satisfying search experience. Furthermore, since the user can actively participate in the query process without passively receiving recommendations from the search engine, the user's proactivity and participation can be improved, and the user experience can be enhanced.

[0018] The technical proposal related to this disclosure will be described in detail below with reference to the drawings and specific embodiments.

[0019] Figure 1 is a schematic diagram illustrating an application scenario of a method and apparatus for identifying interaction information according to an embodiment of the present disclosure.

[0020] Note that FIG. 1 is an illustration of a system architecture to which an embodiment of the present disclosure is applicable for those skilled in the art to understand the technical content of the present disclosure, and it does not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios.

[0021] As shown in FIG. 1, the system architecture 100 according to the embodiment can include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is a medium that provides a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types such as, for example, wired and / or wireless communication links.

[0022] The user can use the terminal devices 101, 102, 103 to interact with the server 105 via the network 104 to send and receive messages and the like. The terminal devices 101, 102, 103 may be various electronic devices having a display and supporting web page browsing, including but not limited to smartphones, tablet computers, laptop portable computers, desktop computers, and the like.

[0023] The server 105 may be a server that provides various services, for example, a background management server (merely an example) that provides support for a website browsed by the user using the terminal devices 101, 102, 103. The background management server can perform processing such as analysis on data such as received user requests, and feedback the processing results (for example, interaction information generated based on query information input by the user) to the terminal device.

[0024] Note that the method for identifying interaction information according to the embodiments of the present disclosure may generally be executed by the server 105. Accordingly, the apparatus for identifying interaction information according to the embodiments of the present disclosure may generally be installed in the server 105. The method for identifying interaction information according to the embodiments of the present disclosure may be executed by a server or a server cluster that is different from the server 105 and is communicable with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the apparatus for identifying interaction information according to the embodiments of the present disclosure may be installed in a server or a server cluster that is different from the server 105 and is communicable with the terminal devices 101, 102, 103 and / or the server 105.

[0025] It should be understood that the numbers of terminal devices, networks, and servers in FIG. 1 are merely illustrative. Any number of terminal devices, networks, and servers may be provided as necessary.

[0026] FIG. 2 is a schematic flowchart of a method for identifying interaction information according to an embodiment of the present disclosure.

[0027] As shown in FIG. 2, the method 200 for identifying interaction information may include operations S210 to S230.

[0028] In operation S210, based on the target query information and historical query information, a plurality of question dimensions are identified, and each question dimension includes a dimension name and a plurality of options.

[0029] For example, the target includes a user who uses a search engine.

[0030] For example, the query information may be content that requires a query input by the user, and the data type of the query information may be text. For example, the query information is "recommend interesting novels".

[0031] For example, historical query information may include query information from a predetermined period in the past, and this predetermined period may be half a month. In some embodiments, the content supplemented and expressed by the user may be included in the historical query information. For example, after a user enters one query into the search engine, they may not close the search engine but continue to enter several other pieces of content, or they may select query content provided by several pages in the search engine, and these pieces of content may be included in the historical query information.

[0032] For example, the dimension name may include "novel type" or "novel length," the options related to "novel type" may include "suspense mystery" or "science fiction," and the options related to "novel length" may include "novel," "novella," or "short story."

[0033] For example, keyword extraction may be performed on query information and historical query information, or a mapping relationship between keywords and question dimensions may be set in advance, and then the question dimensions may be identified based on the mapping relationship.

[0034] In operation S220, the target question dimension is identified from the multiple question dimensions based on whether the semantic information of the multiple question dimensions matches the semantic information of the query result related to the query information, and based on the evaluation values ​​of the multiple question dimensions.

[0035] For example, a query result can represent the content obtained by searching for query information. The data type of the query result may include text, images, videos, etc. For example, the query result may contain the content "The first recommended novel is a full-length novel, the second is a novella, and the third is a short story," and this query content corresponds to the meaning of the term "novel length" in the question dimension.

[0036] For example, the evaluation value may represent the degree to which the question dimension influences the query content, or it may represent the degree to which the question dimension influences the understanding of user query needs. The evaluation value may be a specific numerical value, or it may be an evaluation parameter that indicates a high or low evaluation value. Evaluation values ​​corresponding to the question dimension may be set in advance, or they may be evaluated using a large-scale model.

[0037] For example, a question dimension that does not match the query result's meaning but has a high evaluation value may be identified as the target question dimension.

[0038] In operation S230, interaction information is identified based on the dimension name and multiple choices in the target question dimension.

[0039] For example, the dimension name and multiple choices in the target question dimension may be used as interaction information. Alternatively, the dimension name in the target question dimension can be expressed in natural language. For example, the dimension name "Novel Type" could be converted to "What type of novel would you like to read?" and then the converted dimension information and choices could be used as interaction information, making the interaction information easier for the user to understand.

[0040] The embodiments of this disclosure mine the user's query needs by prompting the user with questions to encourage them to provide more detailed information about their query targets. In this way, the search engine can provide query results that match the user's needs, improve the accuracy and relevance of the query results, and provide the user with a more personalized and satisfying search experience. Furthermore, since the user can actively participate in the query process rather than passively receiving recommendations from the search engine, the user's proactivity and participation can be improved, and the user experience can be enhanced.

[0041] In practical applications, interaction information can be acquired and then applied in multiple ways. In one application scenario, interaction information can be displayed via a front-end page, allowing the user to make selections based on the question dimensions and supplement the query content. In another application scenario, training samples can be built using the interaction information, and other models can be trained using these training samples. In this way, the user can interact with the online application using other models that have been trained. The resources required for other models may be less than those required for large-scale models, and these other models may be SFT (Sparse Fine Tuning) models.

[0042] According to another embodiment of the present disclosure, historical query information may be selected first, and the selection process may include identifying candidate historical query information that satisfies predetermined selection conditions from among the candidate historical query information as historical query information, wherein the predetermined selection conditions may include at least one of the following: the intent of the candidate historical query information matches the intent of the query information; the search frequency of the candidate historical query information satisfies predetermined frequency conditions; and the entities included in the candidate historical query information match the entities included in the query information.

[0043] For example, one could first obtain historical query information for a predetermined period in the past, and then use this historical query information as candidate historical query information.

[0044] Next, it is possible to determine whether the intent of the candidate history query information and the query information matches. For example, a binary classification method can be used to determine whether the intent matches. For instance, the candidate history query information and the query information are input into a pre-trained classification model, and the classification model determines whether the intent matches. Alternatively, intent recognition may be performed on the candidate history query information and the query information separately, and then it may be determined whether the intents of the two match.

[0045] Next, the frequency of candidate history query information can be statistically analyzed, and then candidate history query information can be selected based on predetermined frequency conditions, and it is understood that the frequency can reflect the user's preferences. The predetermined frequency conditions may include the frequency of candidate history query information being greater than a threshold, and the threshold may be a numerical value such as 5. The predetermined frequency conditions may also include the frequency ranking of candidate history query information being in a high predetermined rank, for example, being in the top 10 in terms of frequency.

[0046] Next, it is determined whether the entities included in the selected candidate history query information match the entities included in the query information. The method for identifying entities may include named entity identification, and this embodiment is not limited to this.

[0047] For example, the query information and candidate history query information are "How to repair a first-brand mobile phone if it malfunctions" and "How to repair a second-brand mobile phone if it malfunctions," respectively. While the intent of both is the same, the entities do not match.

[0048] This embodiment can select historical query information that is highly relevant to the query information and of greater interest to the user from candidate historical query information based on predetermined selection criteria, thereby ensuring the accuracy of the question dimensions identified based on the subsequent historical query information.

[0049] Figure 3A is a schematic diagram of the principle of a method for identifying interaction information according to an embodiment of the present disclosure, and Figure 3B is a schematic diagram of an interaction page according to an embodiment of the present disclosure.

[0050] According to other embodiments of the present disclosure, a method for identifying interaction information includes operations S310-S340, where operation S340 is performed before operation S310.

[0051] In operation S340, determine whether a question is necessary.

[0052] In operation S310, multiple question dimensions are identified.

[0053] In operation S320, the target question dimension is selected.

[0054] In operation S330, interaction information is identified.

[0055] In the above operation S340, determining whether a question is necessary is a preprocessing process, primarily determining whether there are complex needs and whether query information needs to be supplemented by the question format. The necessity of a question can be determined by a preprocessing suggestion information template. For example, the query information entered by the user and the preprocessing suggestion information template may be combined and input into a large-scale model, and the large-scale model may output the determination results. The preprocessing suggestion information template may be pre-placed and stored in a predetermined storage area and read from the storage area as needed.

[0056] For example, a preprocessing information template may include a thought chain, which consists of multiple steps, and which instructs the large-scale model to reason according to the steps in the thought chain. For example, the steps included in the thought chain are as follows. The core of the analysis is to determine whether the query result needs to consider different contextual distinctions or user personalization analysis, such as personalization preferences or user personal images. If so, it is determined that the current query belongs to a complex need and needs to be questioned; otherwise, it is determined that there is no need to question it.

[0057] Furthermore, for example, a pre-processing information template may include examples, primarily by providing examples to explain which situations are individualized and which are distinct contexts. For instance, if the query information in the pre-processing information template is "recommended interesting novels," when answering this question, it is necessary to consider individualized elements such as the user's reading habits and preferred literary types. If the query information is "Beijing house prices," there are significant differences in prices between different areas and types of houses, so the answer will involve specific contextual elements such as area and house type. These issues belong to complex needs and are open to questioning. Conversely, if the query information is "what time is it?", this need is clear and does not change depending on the context or the individual's specific situation, so it does not belong to complex needs and does not require questioning.

[0058] This embodiment ensures the accuracy of questions by determining the question needs, thereby allowing for the acquisition of more user information from the questions when appropriate, and providing more accurate personalized answers.

[0059] In the operation S310 described above, the process of identifying multiple question dimensions can be performed based on the target query information and historical query information. Furthermore, the identification process can consider two perspectives: pre-decomposition dimension analysis and post-decomposition dimension mining. For example, by presenting information templates to a large-scale model, the contextual and personalization elements that influence answering the query information can be decomposed from both pre- and post-decomposition perspectives, and the set of question dimensions can be compiled. This generates smearable options according to the needs dimension, forming an interactive user selection interface.

[0060] For example, regarding the query information "Recommended English Courses," a preliminary analysis is first performed to determine that different learners have different personalization needs regarding course content and format, thus consolidating the subdivided dimensions of course content and teaching method. Afterwards, based on the historical query information formed based on the problem, such as "Beijing Online English Teaching" and "English Courses Suitable for Elementary School Students," the supplementary information is analyzed to include Beijing, online, and elementary school students, and the question dimensions are summarized to include learner group, teaching method, and region. In this way, the question dimensions include {"Dimension Name": "Course Content," "Options": ["Daily," "Business," "Exam"]}, {"Dimension Name": "Teaching Method," "Options": ["Online," "Offline"]}, {"Dimension Name": "Learner Group," "Options": ["Elementary School Students," "Middle School Students," "Adults"]}, and {"Dimension Name": "Region," "Options": ["Beijing," "Shanghai," "Guangzhou"]}.

[0061] In this embodiment, prior knowledge from a large-scale model is used to perform a direct analysis on the original needs, providing insights and suggestions into the overall direction of the user's needs. Meanwhile, the true, specific needs of the user are extracted in conjunction with the user's historical search behavior, strengthening the reliability and applicability of the questions. In this way, subdivided answer choices can be generated for each question dimension according to the needs. An interactive user selection area can be formed in the search interface, guiding the user to clarify their complex needs.

[0062] In the above operation S320, in the process of selecting the target question dimension, the target question dimension can be identified from the multiple question dimensions based on whether the semantic information of the multiple question dimensions matches the semantic information of the query result related to the query information, and based on the evaluation values ​​of the multiple question dimensions.

[0063] In the above operation S330, in the process of identifying interaction information, interaction information can be identified based on the dimension name and multiple choices in the target question dimension.

[0064] According to other embodiments of the present disclosure, the process of identifying the multiple question dimensions may include combining a first presentation information template, query information, historical query information, and query results related to the query information as the first presentation information, inputting the first presentation information into a large-scale model, and obtaining the multiple question dimensions. The first presentation information template may be pre-located and stored in a predetermined storage area and read out from the storage area as needed.

[0065] In one example, the first presentation template includes a first thought chain, which includes several steps, and the first thought chain instructs the large-scale model to reason according to the steps in the first thought chain. For example, the steps included in the first thought chain include identifying the relationship type between query information and historical query information, determining whether a question dimension exists based on the relationship type, and, in the case of a candidate question dimension, generating the question dimension and choices based on a predetermined subdivision direction. In another example, the first presentation template may further indicate the identity and task of the large-scale model. The first presentation template may further indicate the format of the input and output information of the large-scale model. The first presentation template may further include examples. In this embodiment, the first presentation template includes a first thought chain, which guides the large-scale model through the steps in the first thought chain, enabling the large-scale model to accurately output the question dimension.

[0066] For example, in this embodiment, the query information is "movies to watch together," and the historical query information includes multiple pieces of information such as which are videos to watch together before exercising, which are movies to watch alone, which are recommended US movies to watch together, and which are movies to watch as a couple. Based on the above query information and historical query information, the following question dimensions can be identified: {'Dimension Name': 'Movie Type', 'Options': ['Comedy', 'Action', 'Romance', 'Science Fiction']}, {'Dimension Name': 'Movie Origin', 'Options': ['USA', 'China', 'India', 'Other Countries']}, and {'Dimension Name': 'Specific Focus Elements', 'Options': ['Content Richness', 'Visual Effects', 'Word of Mouth and Ratings']}.

[0067] For example, in this embodiment, the query information is "Top 10 in the mobile game ranking," and the historical query information includes multiple pieces of information such as mobile game ranking, top 10 in the mobile game ranking of the martial arts genre, top 10 in the mobile game ranking, and top 10 in the mobile game ranking. Based on the above query information and historical query information, the following question dimensions can be identified: {'Dimension Name': 'Game Type', 'Options': ['Martial Arts', 'Shooting', 'Policy', 'Adventure']}, and {'Dimension Name': 'Ranking Basis', 'Options': ['Download Volume', 'Active User Volume', 'User Usage Rating']}.

[0068] For example, in this embodiment, the first information template is as follows:

[0069] As a search query analysis expert, you need to question the "current query" based on "current query" and "historical query information," break down the user's intent, and guide them to perform multiple interactive searches. "Breakdown" is defined as having a specific meaning.

[0070] The specific steps for breaking down the problem are as follows:

[0071] Step 1, Problem Needs Analysis: Analyze what information users are trying to obtain or what problems they are trying to solve through this problem. Determine whether the needs for this problem are clear or not.

[0072] Step 2, Analysis of Historical Query Information: Analyze the "historical query information" and determine its relationship with the "current query." There are mainly four types of relationships: the first type is that the query has the same meaning as the "current query" and there is no information gain; the second type is that the query is completely unrelated to the "current query"; the third type is that the subject or needs have changed; and the fourth type is that further subdivision has been performed on the "current query." Based on the above analysis, determine whether the user's historical query information can provide a useful subdivision dimension, outputting True if it can, and False otherwise.

[0073] In Step 3, determine whether the problem needs to be subdivided. If Step 2 is not False, refer to the contents of the "History Query Information" during the analysis process. When making a specific determination, consider whether the current problem needs to be subdivided from three directions: "Subdivision to supplement specific context," "Subdivision to match individual user situations," and "Subdivision of needs hierarchy." If subdivision is not needed in any of the three directions, it indicates that the problem is sufficiently specific and further subdivision is unnecessary. Otherwise, it can be determined that the current problem needs to be subdivided further. Combine the problem needs with the "History Query Information" to determine whether the current problem needs to be subdivided.

[0074] In Step 4, list the subdivision dimensions: Enumerate the subdivision dimensions that have a significant impact on the answer to the current problem from three directions: "subdivision to supplement specific context," "subdivision to match individual user situations," and "subdivision of needs hierarchy." Here, for subdivision dimensions in the direction of "subdivision of needs hierarchy," only the subdivision dimension needs to be provided, without any content for the options. For subdivision dimensions in the other two directions, content for the options needs to be provided.

[0075] In step 5, remove duplicates from existing subdivision dimensions in the current query results: Analyze whether the current query results include the decomposed subdivision dimensions mentioned above, and if so, remove the dimension with semantic duplication in the next analysis.

[0076] ## Input format: query{}, Answer{}, History query information{}.

[0077] ## Output format: { "Problem Needs Analysis": "Analyze the user's 'current query'." "History query information analysis": "Analyzing historical query information activity." "Determining whether or not to subdivide the problem": "Determine whether or not to trigger subdivision from three perspectives." “List the subdivision dimensions”: “List the subdivision dimensions and corresponding options from three directions.” "Deduplication of existing subdivision dimensions in the current query results": "". }

[0078] ##Example: AAA. ## Identify the thought process and results according to the requirements and [examples].

[0079] In summary, the first information template has been explained, and it should be understood that in the first information template, `query` represents query information and `answer` represents the query result.

[0080] According to another embodiment of the present disclosure, the process for selecting target question dimensions may include a deduplication subprocess and a valuation subprocess.

[0081] For the deduplication subprocess, if it is determined that the semantic information of a question dimension matches the semantic information of the query result associated with the query information for each of the multiple question dimensions, the question dimension can be removed from the multiple question dimensions to obtain a deduplication-free question dimension.

[0082] For the evaluation subprocess, the target question dimension can be identified from the deduplication-free question dimensions based on the evaluation value of each deduplication-free question dimension.

[0083] Furthermore, if some of the query results already contain feedback, the user does not need to repeat related questions. Therefore, the deduplication subprocess can avoid duplicate queries and improve the accuracy of the query dimensions. By evaluating the query dimensions, the accuracy of the target query dimensions can be better aligned with the user's query needs.

[0084] According to another embodiment of the present disclosure, the deduplication subprocess in the process of selecting target question dimensions may include combining a second presentation information template, query information, query results, and multiple question dimensions as second presentation information, and then inputting the second presentation information into a large model to obtain deduplication of question dimensions.

[0085] The format of the second presentation template is similar to that of the first presentation template. For example, the second presentation template may be pre-located and stored in a predetermined memory area and read from the memory area when needed. For example, the second presentation template includes a second thought chain, which includes several steps, and the second thought chain instructs the large-scale model to reason according to the steps in the second thought chain. For example, the steps included in the second thought chain include generating summary information for query result information and determining whether the summary information is related to the question dimension. In another example, the second presentation template may further indicate the identity and tasks of the large-scale model. The second presentation template may further indicate the format of the input and output information of the large-scale model. The second presentation template may include examples. In this embodiment, the second presentation template includes a second thought chain, and by guiding the large-scale model through the steps in the second thought chain, the large-scale model can perform deduplication accurately, thereby avoiding asking duplicate questions on existing content in the query results.

[0086] For example, in this embodiment, the second information template is as follows:

[0087] You are a robot that evaluates whether the question dimension matches the meaning of the query result answer. Your role is to evaluate the input <query> 、 <answer>The process involves generating a thought logic chain based on the <question dimension> to determine whether the current question dimension overlaps with the meaning of the query result.

[0088] The specific thought steps and logic are as follows:

[0089] In Step 1, the current <query> 、 <answer>And a deep understanding of the <question dimension>, and the current state of the <question dimension> <query>Deeply consider and understand the value and impact on our understanding of this topic, and summarize the main points already answered in the current query results.

[0090] In step 2, it is determined whether the content of each question dimension was mentioned in step 2 above, based on the given list of question dimensions. If it was mentioned, it is determined that the meaning of that question dimension overlaps with the meaning of the current query result; otherwise, it is determined that there is no overlap.

[0091] In step 3, the corresponding JSON data is output.

[0092] Below, I will provide you with examples, and you will need to learn how to generate the required thought chain through these examples.

[0093] Example: BBB.

[0094] Next, you are given one input, and you use what you have learned above to think deeply, understand, and generate a corresponding chain of thought that eliminates duplication.

[0095] The above explains the second information template.

[0096] According to another embodiment of the present disclosure, a valuation subprocess in the process of selecting target question dimensions may include combining a third-presentation information template, query information, query results, and deduplication of question dimensions as third-presentation information, and then inputting the third-presentation information into a large-scale model to obtain target question dimensions.

[0097] The format of the third presentation information template is the same as that of the first and second presentation information templates. For example, the third presentation information template may be pre-configured and stored in a designated memory area and read from the memory area when needed. For example, the third presentation information template includes a third thought chain, the third thought chain includes several steps, and the third thought chain instructs the large-scale model to reason according to the steps in the third thought chain. For example, the steps included in the third thought chain include identifying evaluation values ​​for question dimensions based on a predetermined evaluation method, and selecting a target question dimension from the deduplication-reduced question dimensions based on the evaluation values. The predetermined evaluation method may include evaluating the importance of the question dimensions in influencing the current answer, and may also include evaluating whether the question dimensions help to further understand the user's needs. In another example, the third presentation information template may further indicate the identity and tasks of the large-scale model. The third presentation information template may further indicate the format of the input and output information of the large-scale model. The third presentation information template may further include examples. In this embodiment, the third presented information template includes a third thought chain, and by guiding the large-scale model through the steps in the third thought chain, the large-scale model can accurately select more valuable question dimensions, thereby improving user satisfaction and ensuring the value and applicability of the question dimensions.

[0098] For example, in this embodiment, the third information template is as follows:

[0099] You are a robot to evaluate the degree of prior value of a question dimension. Your role is to input... <query> 、 <answer>The goal is to generate a thought logic chain that analyzes the prior value of the question dimension using the <question dimension>.

[0100] The specific thought steps and logic are as follows:

[0101] In Step 1, the current <query> 、 <answer>And a deep understanding of the <question dimension>, and the current state of the <question dimension> <query>To deeply consider and understand the value and impact on [something].

[0102] In Step 2, the value of the newly submitted question dimensions is evaluated, with specific evaluation criteria being whether the current subdivision direction is a significant factor influencing the answer to the question, and whether it is effective in gaining a deeper understanding of the current question needs.

[0103] In Step 3, finally, select the three most valuable goal question dimensions from among them, provide up to four options for each goal question dimension, and determine whether the options are mutually exclusive.

[0104] Example: CCC.

[0105] This concludes the explanation of the third presentation template.

[0106] In other embodiments of this disclosure, after obtaining multiple target question dimensions, the multiple target question dimensions may be sorted, or they may be sorted based on how often they are clicked. For example, the user group to which the user who entered the query information belongs may be identified, and then the frequency with which that user group clicks on a target question dimension in the history search process may be identified. Then, the target question dimensions are sorted based on the frequency. In cases where the intentions of multiple users in the same user group are relatively similar, this embodiment can prioritize the presentation of more valuable target question dimensions by determining the ranking of the target question dimensions through user group history operations.

[0107] According to other embodiments of the present disclosure, the process of identifying interaction information may include combining a fourth presentation information template, query information, and target question dimensions as fourth presentation information, then inputting the fourth presentation information into a large-scale model to obtain a question problem in the interaction information.

[0108] The format of the fourth presentation information template is similar to that of the first, second, and third presentation information templates. For example, the fourth presentation information template may be pre-configured and stored in a predetermined memory area and read from the memory area when needed. For example, the fourth presentation information template includes a fourth thought chain, which includes several steps, and the fourth thought chain instructs the large-scale model to reason according to the steps in the fourth thought chain. For example, the steps included in the fourth thought chain include identifying a query scene based on query information and question dimensions, and generating a question problem in interaction information based on the query scene. In another example, the fourth presentation information template may further indicate the identity and task of the large-scale model. The fourth presentation information template may further indicate the format of the input and output information of the large-scale model. The fourth presentation information template may further include examples. In this embodiment, the fourth presentation information template includes a fourth thought chain, which guides the large-scale model through the steps in the fourth thought chain, enabling the large-scale model to accurately output a question problem.

[0109] For example, in this embodiment, the fourth information template is as follows:

[0110] Your task is to reconstruct this question based on the original search problem, the information angles of the question, and the answer choices for the question. Specifically, you need to follow these steps:

[0111] In Step 1, we gain a deep understanding of the information angles of the current search and question problems, analyze the key factors of the questions and the user's needs, and estimate the possible background or scenario in which the user asked the question.

[0112] In Step 2, based on the user's background information, a simple and fluent question is formed, taking into account the information angles and answer choices provided for the question.

[0113] Example: Input: "Search Question": "AA"; "Dimension Name": "BB"; "Option": "CC". output: "Step 1": "User Question DD, the key factor of this problem is EE, and the user need is FF. The background or scenario the user might face is GG, so we need to identify HH."

[0114] "Step 2": "The information angle of the question is II, and based on the current question background, a question can be submitted for keyword JJ, guiding the user to supplement KK."

[0115] "Final Question": "LL".

[0116] This concludes the explanation of the fourth information template.

[0117] Some embodiments of this disclosure combine a large language model with prompt engineering and use presentation templates to guide the large model to output results. The large model may be a large language model (LLM), a generative model, etc., and this disclosure is not limited to large models. The presentation templates (e.g., preprocessing presentation templates, first presentation templates, second presentation templates, third presentation templates, fourth presentation templates, and fifth presentation templates) may all be text information written in natural language. Each presentation template may be pre-located and stored in a predetermined memory area and read from the memory area when needed.

[0118] According to other embodiments of this disclosure, after acquiring interaction information, the interaction information can be applied in multiple ways. In one application scenario, the interaction information can be displayed via a front-end page, allowing the user to make selections based on the question dimension and supplement the query content.

[0119] For example, after a user selects several options, new query information can be identified based on the option the user clicked, and then the query results displayed to the user can be updated by performing a query based on this new query information. For instance, if the query information is "Adjusting the body with traditional Chinese medicine," the question dimension name is "Gender," and the user selected the option "Female," then the new query information could be "Women adjusting their bodies with traditional Chinese medicine."

[0120] Furthermore, after the user selects an option, the selected option may be converted into natural language. For example, if the question dimension name is "gender" and the user selects "female," the converted text information may be "My gender is female," and this converted text information may be entered into the input box on the page.

[0121] According to other embodiments of this disclosure, after acquiring interaction information, the interaction information can be applied in multiple ways. In another application scenario, the interaction information can be used to build training samples, and these training samples can be used to train other models. In this way, when resources are limited, the trained other models can be used to interact with the user.

[0122] For example, taking another model as an example of a model containing two SFT models, to facilitate distinction, we will refer to one SFT model as Model 1 and the other SFT model as Model 2 below.

[0123] Regarding the first model, it may be a generative model. In the training process, interaction information may be precisely labeled, and then the first model may be trained using the precisely labeled data as training samples. After training, the input to the first model may include query information, historical query information, query results related to the query information, and may further include presentation information templates applied to the first model. The output of the first model may include a question dimension, question choices, and a question obtained by converting the question dimension into natural language.

[0124] For example, an example of the output of the first model is as follows: {"Question Dimension": "Meaning of Name", "Question": "What is the desired meaning for a child's name?", "Options": ["Intelligence and Wisdom", "Health and Peace", "Noble Character"], "Are the options mutually exclusive?": "Yes", {"Question Dimension": "Source of Name", "Question": "What is the source of this desired name?", "Options": ["Tang Poetry", "Song Ci", "Book of Songs", "Chu Ci"], "Are the options mutually exclusive?": "No"}.

[0125] Regarding the second model, it may be a generative model. The second model generates consistent natural language text based on the above question dimensions, question options, and question, and fills it into the input box on the front-end page. For example, the output of the second model might be: "I would like my baby's gender to be a boy, and I would like it to convey the meanings of brave, strong, kind, and harmonious."

[0126] For example, a pre-placed information template may be used to guide the second model to generate a natural language question, such an information template as follows:

[0127] Let's say you are an artificial intelligence assistant, and your primary functions are problem analysis and dimensional interpolation. You provide feedback to the user in a more natural language-based way, guiding continuous interaction between you and the user.

[0128] The overall analysis process is as follows:

[0129] Thought Process: Step 1 involves modifying the natural language guide description for each dimension, based on the questioner's perspective in the interaction guide, to match the question. Step 2 involves randomly selecting one option for each dimension and providing a natural language supplementary explanation of the needs, based on the respondent's perspective in the interaction guide, to match the problem.

[0130] Dimensional Guide Transformation: Based on the first step in the thought process, it provides the modified result of the natural language-based guide formula description for each dimension.

[0131] Needs Supplementary Explanation: Based on the second step in the thinking process, a needs supplementary explanation guide is provided.

[0132] Examples are as follows:

[0133] Question: What is the difference between a futon and a mattress?

[0134] Supplemental dimension: [{"Dimension name":"Aspect of comparison", "Sub-dimension options":["Thickness", "Material", "Cleanliness"]}].

[0135] Overall analysis process: "Thought Process": The supplementary dimension is "Aspect of Contrast," which, depending on the problem, can be used to provide a natural language explanation and translate it into "What aspect of contrast are you more interested in regarding the distinction between a futon and a mattress?" Assuming the user selected "thickness" in the "Aspect of Contrast" dimension, we can organize a supplementary explanation of the user's needs, specifically, "I want to know the difference between a futon and a mattress when the aspect of contrast is thickness."

[0136] "Dimensional Guide Transformation": ["What aspects of the distinction between futon and shikifuton are of more interest to you?"]

[0137] "Needs supplementary explanation": {"Prefix": "Regarding futons and mattresses": "Combination of dimensional options": ["The aspect of contrast is thickness"], "Suffix": "I want to know the difference in cases"}.

[0138] A note of caution is that the output format may be analyzed using a predetermined function.

[0139] Based on the overall analysis process and examples described above, please analyze the following issues and supplementary dimensions.

[0140] Problem: MMM Supplemental dimension: NNN

[0141] The above describes the information template used to guide the second model.

[0142] Figure 4 is a schematic block diagram of an apparatus for identifying interaction information according to an embodiment of the present disclosure.

[0143] As shown in Figure 4, the device 400 for identifying interaction information may include a first identification module 410, a second identification module 420, and a third identification module 430.

[0144] The first identification module 410 identifies multiple question dimensions based on the target query information and historical query information, and each question dimension includes a dimension name and multiple options.

[0145] The second identification module 420 identifies the target question dimension from the multiple question dimensions based on whether the semantic information of the multiple question dimensions matches the semantic information of the query results related to the query information, and based on the evaluation values ​​of the multiple question dimensions.

[0146] The third identification module 430 identifies interaction information based on the dimension name and multiple choices in the target question dimension.

[0147] In another embodiment of the present disclosure, the apparatus further includes a history information identification module that identifies candidate history query information as history query information, which satisfies the following conditions: the intent of the candidate history query information matches the intent of the query information, the search frequency of the candidate history query information satisfies a predetermined frequency condition, and the entities included in the candidate history query information match the entities included in the query information.

[0148] In another embodiment of the present disclosure, the first specific module includes a first combination submodule and a first input submodule. The first combination submodule combines a first presentation information template, query information, historical query information, and query results related to the query information as the first presentation information; the first input submodule inputs the first presentation information into a large model to obtain multiple question dimensions; the first presentation information template includes a first thought chain, which instructs the large model to obtain multiple question dimensions in such a manner that it identifies the relationship type between the query information and the historical query information, determines whether a question dimension exists based on the relationship type, and, if it is a candidate question dimension, generates a question and options based on a predetermined subdivision direction.

[0149] In another embodiment of the present disclosure, the second identification module includes a delete submodule and a identification submodule. The delete submodule deletes a question dimension from the multiple question dimensions if it identifies that the semantic information of the question dimension matches the semantic information of the query result associated with the query information, thereby obtaining the deduplication question dimension. The identification submodule then identifies a target question dimension from the deduplication question dimensions based on the evaluation value of each deduplication question dimension.

[0150] In another embodiment of the present disclosure, the deletion submodule includes a first combination unit and a first input unit. The first combination unit combines a second presentation information template, query information, query results, and multiple question dimensions as the second presentation information, and the first input unit inputs the second presentation information into a large model to obtain deduplication question dimensions, where the second presentation information template includes a second thought chain, the second thought chain instructs the large model to obtain deduplication question dimensions in such a way that it generates summary information for the query results and determines whether the summary information is related to the question dimensions.

[0151] In another embodiment of the present disclosure, a particular submodule includes a second combination unit and a second input unit. The second combination unit combines a third presentation information template, query information, query results, and deduplication question dimensions as third presentation information, and the second input unit inputs the third presentation information into a large model to obtain a target question dimension, where the third presentation information template includes a third thought chain, the third thought chain instructs the large model to obtain the target question dimension in a manner that identifies evaluation values ​​for question dimensions based on a predetermined evaluation method, and selects a target question dimension from the deduplication question dimensions based on the evaluation values.

[0152] In another embodiment of the present disclosure, the third specific module includes a second combination submodule and a second input submodule. The second combination submodule combines a fourth presentation information template, query information, and target question dimensions as the fourth presentation information, and the second input submodule inputs the fourth presentation information into a large model to obtain a question problem in interaction information, where the fourth presentation information template includes a fourth thought chain, which instructs the large model to obtain a question problem in interaction information in such a way that it identifies a query scene based on the query information and question dimensions, and generates a question problem in interaction information based on the query scene.

[0153] According to embodiments of the present disclosure, the present disclosure further provides an electronic device comprising at least one processor and a memory communicated with 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 perform the method for identifying the interaction information described above.

[0154] According to embodiments of the present disclosure, the present disclosure further provides a non-temporary computer-readable storage medium in which computer instructions are stored, wherein the computer instructions cause the computer to perform the method for identifying the interaction information described above.

[0155] According to embodiments of the present disclosure, the present disclosure further provides a computer program that, when executed by a processor, implements the method for identifying the interaction information described above.

[0156] Figure 5 is a structural block diagram of an electronic device for implementing a method for identifying interaction information according to an embodiment of the present disclosure. 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 may further represent various forms of mobile devices, such as personal digital assistants, mobile phones, smartphones, wearable devices, and other similar computing devices. The components, their connections and relationships, and their functions shown herein are illustrative and do not limit the implementation of the present disclosure as described herein and / or requested herein.

[0157] As shown in Figure 5, the device 500 includes a computing unit 501, which may perform various appropriate operations and processes based on computer programs stored in read-only memory (ROM) 502 or computer programs loaded from storage unit 508 into random access memory (RAM) 503. The RAM 503 may also store various programs and data necessary for the operation of the device 500. The computing unit 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0158] Multiple components in device 500 are connected to the I / O interface 505 and include, for example, an input unit 506 such as a keyboard or mouse; an output unit 507 such as various types of displays or speakers; a storage unit 508 such as a magnetic disk or optical disk; and a communication unit 509 such as a network card, modem, or wireless communication transceiver. The communication unit 509 enables device 500 to exchange information and data with other devices via computer networks such as the Internet and / or various electrical networks.

[0159] The computing unit 501 may be various general-purpose and / or dedicated processing modules having processing and computational capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a GPU (Graphics Processing Unit), various dedicated artificial intelligence (AI) computing chips, computing units for running various machine learning model algorithms, a DSP (Digital Signal Processor), and any suitable processor, controller, microcontroller, etc. The computing unit 501 performs each of the methods and processes described in the preceding paragraph, for example, the method for identifying interaction information. For example, in some embodiments, the method for identifying interaction information may be implemented as a computer software program tangibly contained in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed into the device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the method for identifying interaction information described in the preceding paragraph may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform a method for identifying interaction information by any other suitable method (e.g., via firmware).

[0160] Various embodiments of the systems and technologies described herein may 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-chip (SOCs), complex-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may be implemented in one or more computer programs which can be executed and / or interpreted on a programmable system which includes at least one programmable processor, which may be a dedicated or general-purpose programmable processor which 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.

[0161] Program code for carrying out the methods of this disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a dedicated computer, or other programmable data processing device, so that when the program code is executed by the processor or controller, the functions and operations defined in the flowcharts and / or block diagrams are performed. The program code may be executed entirely on a device, partially on a device, partially on a device as a standalone software package, partially on a remote device, or entirely on a remote device or server.

[0162] In the context of this disclosure, a machine-readable medium may be a tangible medium that contains or stores programs used in or in combination with an instruction execution system, device, or electronic device. 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, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any appropriate combination thereof. More specific examples of machine-readable storage media include: , mobile Band-type computer disks, hard disks, random access memory (RAM), read-only memory (ROM), and erasable programmable read-only memory (EPROM or flash memory) ,Ko This includes compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0163] To provide interaction with a user, a computer may be made to implement the systems and techniques described herein, the computer comprising a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor), a keyboard and a pointing device (e.g., a mouse or trackball), the user may provide input to the computer via the keyboard and the pointing device. Other types of devices may further provide interaction with the user, for example, feedback provided to the user may be any form of sensing feedback (e.g., visual feedback, auditory feedback, or haptic feedback), and input from the user may be received in any form (including voice input, speech input, or haptic input).

[0164] The systems and technologies described herein can be implemented in computing systems including background components (e.g., a data server), computing systems including middleware components (e.g., an application server), computing systems including front-end components (e.g., a user computer having a graphical user interface or a web browser, through which the user can interact with embodiments of the systems and technologies described herein), or in computing systems including any combination of such background components, middleware components, or front-end components. Components of the system can be connected to one another by digital data communication (e.g., a communication network) in any form or medium. Examples of communication networks include, but are not limited to, local area networks (LANs), wide area networks (WANs), and the Internet.

[0165] A computer system may include clients and servers. Clients and servers are generally geographically separated and typically communicate via a communication network. The client-server relationship is generated by a computer program running on the relevant computer that has a client-server relationship.

[0166] It should be understood that various forms of flows shown above may be used, and operations may be re-sorted, added, or deleted. For example, each operation described herein may be performed in parallel, sequentially, or in a different order, as long as the desired results of the proposed techniques disclosed herein can be achieved.

[0167] The specific embodiments described above do not limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, subcombinations, and substitutions are possible depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.< / query> < / answer> < / query> < / answer> < / query> < / query> < / answer> < / query> < / answer> < / query>

Claims

1. A method for identifying interaction information, which is performed by a computer, Identifying multiple question dimensions based on target query information and historical query information, wherein each question dimension includes a dimension name and multiple options. Based on whether the semantic information of multiple question dimensions matches the semantic information of the query results related to the query information, and based on the evaluation values ​​of the multiple question dimensions, the target question dimension is identified from the multiple question dimensions. This includes identifying interaction information based on the dimension name and multiple options in the aforementioned target question dimension. A method for identifying interaction information characterized by the following:

2. This further includes identifying candidate history query information as the history query information that satisfies the following conditions from among the candidate history query information: the intent of the candidate history query information matches the intent of the query information, the search frequency of the candidate history query information satisfies a predetermined frequency condition, and the entities included in the candidate history query information match the entities included in the query information. The method according to feature 1.

3. Identifying multiple query dimensions based on the aforementioned target query information and historical query information is possible. Combining the first presentation information template, the query information, the historical query information, and the query results related to the query information as the first presentation information, This includes inputting the first presented information into a large-scale model and obtaining the multiple question dimensions, Here, the first presentation information template includes a first thought chain, the first thought chain instructs the large-scale model to obtain the multiple question dimensions in a manner that it identifies the relationship type between the query information and the historical query information, determines whether a question dimension exists based on the relationship type, and, if a question dimension exists, generates the question dimension and choices based on a predetermined subdivision direction. The method according to feature 1.

4. Identifying a target question dimension from the multiple question dimensions based on whether the semantic information of the multiple question dimensions matches the semantic information of the query result related to the query information, and based on the evaluation values ​​of the multiple question dimensions, If, for each of the multiple question dimensions, it is determined that the semantic information of the question dimension matches the semantic information of the query result related to the query information, the question dimension is deleted from the multiple question dimensions to obtain the question dimension after deduplicating. This includes identifying the target question dimension from the deduplication-removed question dimensions based on the evaluation values ​​of each of the deduplication-removed question dimensions. The method according to feature 1.

5. If it is determined that the semantic information of the question dimension matches the semantic information of the query result related to the query information, the question dimension is deleted from the multiple question dimensions, and the question dimension after deduplicating is obtained. Combining the second presentation information template, the query information, the query results, and the multiple question dimensions as the second presentation information, This includes inputting the second presented information into a large-scale model and obtaining the question dimensions after deduplication, Here, the second presentation information template includes a second thought chain, which instructs the large-scale model to obtain the deduplication-reduced question dimensions by generating summary information for the query results and determining whether the summary information is related to the question dimensions. The method according to feature 4.

6. Based on the evaluation values ​​of each of the question dimensions after deduplication, identifying the target question dimension from the question dimensions after deduplication is: Combining the third presentation information template, the query information, the query results, and the question dimension after deduplication as the third presentation information, This includes inputting the aforementioned third presentation information into a large-scale model and obtaining the aforementioned target question dimension, Here, the third presented information template includes a third thought chain, the third thought chain instructs the large-scale model to obtain the target question dimension by identifying an evaluation value for the question dimension based on a predetermined evaluation method, and selecting the target question dimension from the deduplication-reduced question dimension based on the evaluation value. The method according to feature 4.

7. Identifying interaction information based on the dimension name and multiple options in the aforementioned target question dimension is possible. Combining the fourth presentation information template, the query information, and the target question dimension as the fourth presentation information, This includes inputting the aforementioned fourth presentation information into a large-scale model and obtaining the question in the interaction information, Here, the fourth presentation information template includes a fourth thought chain, which instructs the large-scale model to obtain the question problem in the interaction information by identifying a query scene based on the query information and the question dimension, and generating the question problem in the interaction information based on the query scene. The method according to feature 1.

8. A device for identifying interaction information, Based on the target query information and historical query information, multiple question dimensions are identified, and each question dimension comprises a first identification module containing a dimension name and multiple choices, A second identification module identifies a target question dimension from the multiple question dimensions based on whether the semantic information of the multiple question dimensions matches the semantic information of the query result related to the query information, and based on the evaluation values ​​of the multiple question dimensions. Includes a third identification module that identifies interaction information based on the dimension name and multiple options in the aforementioned target question dimension. A device for identifying interaction information, characterized by the following features.

9. The system further includes a history information identification module that identifies candidate history query information as the history query information if the candidate history query information satisfies the following conditions: the intent of the candidate history query information matches the intent of the query information, the search frequency of the candidate history query information satisfies a predetermined frequency condition, and the entities included in the candidate history query information match the entities included in the query information. The apparatus according to feature 8.

10. The first specific module is, A first combination submodule that combines a first presentation information template, the query information, the historical query information, and query results related to the query information as first presentation information, Includes a first input submodule that inputs the first presented information into a large-scale model and obtains the multiple question dimensions, Here, the first presentation information template includes a first thought chain, the first thought chain instructs the large-scale model to obtain the multiple question dimensions in a manner that it identifies the relationship type between the query information and the historical query information, determines whether a question dimension exists based on the relationship type, and, if a question dimension exists, generates the question dimension and choices based on a predetermined subdivision direction. The apparatus according to feature 8.

11. The second specific module is, For each of the multiple question dimensions, if it is determined that the semantic information of the question dimension matches the semantic information of the query result related to the query information, the delete submodule deletes the question dimension from the multiple question dimensions and obtains the question dimension after deduplication. Includes a specific submodule that identifies the target question dimension from the deduplication-removed question dimensions based on the evaluation value of each of the deduplication-removed question dimensions. The apparatus according to any one of claims 8 to 10, characterized by the features described above.

12. The aforementioned deletion submodule is: A first combination unit that combines a second presentation information template, the query information, the query results, and the multiple question dimensions as second presentation information, The first input unit inputs the second presented information into a large-scale model and obtains the question dimensions after deduplication, Here, the second presentation information template includes a second thought chain, which instructs the large-scale model to obtain the deduplication-reduced question dimensions by generating summary information for the query results and determining whether the summary information is related to the question dimensions. The apparatus according to feature 11.

13. The aforementioned specific submodule is A second combination unit that combines the third presentation information template, the query information, the query results, and the deduplication-removed question dimension as third presentation information, The unit includes a second input unit that inputs the third presented information into a large-scale model and obtains the target question dimension, Here, the third presented information template includes a third thought chain, the third thought chain instructs the large-scale model to obtain the target question dimension by identifying an evaluation value for the question dimension based on a predetermined evaluation method, and selecting the target question dimension from the deduplication-reduced question dimension based on the evaluation value. The apparatus according to feature 11.

14. The aforementioned third specific module is A second combination submodule that combines the fourth presentation information template, the query information, and the target question dimension as the fourth presentation information, The system includes a second input submodule that inputs the aforementioned fourth presentation information into a large-scale model and obtains the question in the interaction information, Here, the fourth presentation information template includes a fourth thought chain, which instructs the large-scale model to obtain the question problem in the interaction information by identifying a query scene based on the query information and the question dimension, and generating the question problem in the interaction information based on the query scene. The apparatus according to feature 8.

15. At least one processor, The memory includes at least one processor and is connected to it in communication, 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 so that the at least one processor can perform the method according to any one of claims 1 to 7. electronic equipment.

16. A non-temporary computer-readable storage medium in which computer instructions are stored, The computer instruction causes the computer to execute the method described in any one of claims 1 to 7. A non-temporary computer-readable storage medium.

17. A computer program, when executed by a processor, that implements the method according to any one of claims 1 to 7.

Images