Interaction method and device based on large model, intelligent agent and electronic equipment

Abstract

The invention provides an interaction method and device based on a large model, an intelligent agent and electronic equipment, and relates to the technical field of artificial intelligence, in particular to the technical field of large language models and AI assistants. The specific implementation scheme of the interaction method comprises the following steps: analyzing demand information of a target post and resume information of a target object by utilizing a large model to generate an initial problem; according to the reply state of the target object for the initial question, optimizing the initial question by using the large model to generate a target question; and interacting with the target object based on the target question so as to determine the matching degree of the target object and the target post based on an interaction result.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, particularly to the fields of large language models and AI assistant technologies, and specifically to an interaction method, device, intelligent agent, and electronic device based on large models. Background Art

[0002] A large language model (LLM), also known as a large-scale language model, refers to a computer model that can process and generate natural language. It can not only generate natural language text but also deeply understand the meaning of natural language text, handle various natural language tasks, and is widely used in AI assistants to perform processing tasks such as AI search and AI chat. Summary of the Invention

[0003] The present disclosure provides an interaction method, device, intelligent agent, and electronic device based on large models.

[0004] According to one aspect of the present disclosure, there is provided an interaction method based on a large model, including: using the large model to analyze the requirement information of the target position and the resume information of the target object to generate an initial question; optimizing the initial question using the large model according to the response status of the target object to the initial question to generate a target question; and interacting with the target object based on the target question so as to determine the matching degree between the target object and the target position based on the interaction result.

[0005] According to another aspect of the present disclosure, there is provided an interaction device based on a large model, including: an analysis module, an optimization module, and an interaction module.

[0006] The analysis module is configured to use the large model to analyze the requirement information of the target position and the resume information of the target object to generate an initial question.

[0007] The optimization module is configured to optimize the initial question using the large model according to the response status of the target object to the initial question to generate a target question.

[0008] The interaction module is configured to interact with the target object based on the target question so as to determine the matching degree between the target object and the target position based on the interaction result.

[0009] According to another aspect of the present disclosure, there is provided an intelligent agent for interaction, including: an input module, a processing module, and an output module.

[0010] An input module for receiving input information. A processing module for determining a target task based on the input information received by the input module, determining a large model based on the target task, and obtaining output information by invoking the large model to execute the interaction method based on the large model described above. An output module for outputting the output information obtained by the processing module.

[0011] According to another aspect of the present disclosure, there is provided an electronic device, including: 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 when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the method as described above.

[0012] According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the method as described above.

[0013] According to another aspect of the present disclosure, there is provided a computer program product including a computer program, which implements the method as described above when executed by a processor.

[0014] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used 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

[0015] The drawings are used to better understand the solution and do not constitute a limitation to the present disclosure. Among them:

[0016] Figure 1 Schematically shows an exemplary system architecture to which the interaction method and apparatus based on a large model according to an embodiment of the present disclosure can be applied;

[0017] Figure 2 Schematically shows a flowchart of the interaction method based on a large model according to an embodiment of the present disclosure;

[0018] Figure 3 Schematically shows a schematic diagram of generating an initial question using a large model according to an embodiment of the present disclosure;

[0019] Figure 4A Schematically shows an interaction schematic diagram based on a large model according to an embodiment of the present disclosure;

[0020] Figure 4B Schematically shows an interaction schematic diagram based on a large model according to another embodiment of the present disclosure;

[0021] Figure 4CSchematically shows an interaction schematic diagram based on a large model according to another embodiment of the present disclosure.

[0022] Figure 5 Schematically shows a block diagram of an interaction device based on a large model according to an embodiment of the present disclosure;

[0023] Figure 6 Schematically shows a block diagram of an agent for interaction according to an embodiment of the present disclosure; and

[0024] Figure 7 Schematically shows a block diagram of an electronic device suitable for implementing an interaction method based on a large model according to an embodiment of the present disclosure. Detailed implementation manners

[0025] The following describes exemplary embodiments of the present disclosure in conjunction with the accompanying drawings. Various details of the embodiments of the present disclosure are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize 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, the description of well-known functions and structures is omitted below.

[0026] In the recruitment scenario, the supply-demand relationship between job requirements and job supplies is not balanced. For job suppliers, how to quickly identify candidates who match the job requirements among numerous applicants is the actual demand of job suppliers in the interview interaction process.

[0027] Large language models (LLMs) have strong natural language understanding capabilities. Therefore, in the embodiments of the present invention, by utilizing the natural language understanding capabilities of large models, on the premise of fully understanding job requirements and the resume information of the target object, initial questions for preliminarily examining whether the target object has the skills required for the target job are generated. And during the interaction process, based on the response status of the target object, the large model is used to dynamically optimize the questions, and optimized questions for in-depth examination of the target object are generated, improving the matching degree between the questions generated by the large model and the candidates and job requirements, and further improving the interaction flexibility and interaction efficiency in the interview scenario.

[0028] Figure 1 Schematically shows an exemplary system architecture to which an interaction method and device based on a large model can be applied according to an embodiment of the present disclosure.

[0029] It should be noted that Figure 1The figure shown is only an example of the system architecture to which the embodiments of the present disclosure can be applied, to help those skilled in the art understand the technical content of the present disclosure, but it does not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios. For example, in another embodiment, the exemplary system architecture to which the interaction method and device based on a large model can be applied may include a terminal device, but the terminal device can implement the interaction method and device based on a large model provided by the embodiments of the present disclosure without interacting with the server.

[0030] As Figure 1 shown, the system architecture 100 according to this embodiment may include a terminal device 101, a network 102, and a server 103. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired and / or wireless communication links, etc.

[0031] A user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications may be installed on the terminal device 101, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (only as examples).

[0032] The terminal device 101 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, and desktop computers, etc.

[0033] The server 103 may be a server providing various services, such as a background management server that provides support for the content browsed by the user using the terminal device 101 (only as an example). The background management server may analyze and process data such as user requests received, and feedback the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0034] It should be noted that the interaction method based on a large model provided by the embodiments of the present disclosure can generally be executed by the terminal device 101. Correspondingly, the interaction device based on a large model provided by the embodiments of the present disclosure can also be set in the terminal device 101.

[0035] Alternatively, the large model-based interaction method provided by the embodiments of the present disclosure can generally be executed by the server 103. Correspondingly, the large model-based interaction device provided by the embodiments of the present disclosure can generally be disposed in the server 103. The large model-based interaction method provided by the embodiments of the present disclosure can also be executed by a server or a server cluster different from the server 103 and capable of communicating with the terminal device 101 and / or the server 103. Correspondingly, the large model-based interaction device provided by the embodiments of the present disclosure can also be disposed in a server or a server cluster different from the server 103 and capable of communicating with the terminal device 101 and / or the server 103.

[0036] For example, when a user conducts an online interview using a terminal device, the terminal device 101 can obtain the job requirement information and the resume information of the job applicant uploaded by the user. Then, the terminal device 101 can send the above information to the server 103. The server 103 generates an initial question by invoking a large model and sends the initial question to the terminal device 101 to interact with the job applicant through the interaction interface 1011 of the terminal device 101. Next, when the terminal device receives the reply content of the job applicant, it can send the reply content to the server 103. The server 103 optimizes the initial question based on the reply status of the job applicant to generate a question for the next round of interaction. And the generated question is sent to the terminal device 101 for the next round of interaction with the job applicant. And so on, questions for the next round of interaction can be dynamically generated based on the reply status of the job applicant in the previous interaction to determine the matching degree between the job applicant and the target position based on the interaction result.

[0037] It should be understood that Figure 1 the numbers of the terminal devices, networks, and servers in

[0038] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision, disclosure, and application, etc. of the user's personal information all comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good customs.

[0039] In the technical solution of the present disclosure, the authorization or consent of the user is obtained before obtaining or collecting the user's personal information.

[0040] Figure 2 Schematically shows a flowchart of a large model-based interaction method according to an embodiment of the present disclosure.

[0041] As Figure 2 shown, the interaction method 200 may include operations: operation S210 to operation S230.

[0042] In operation S210, using a large model, an initial question is generated by analyzing the requirement information of the target position and the resume information of the target object.

[0043] In operation S220, according to the response status of the target object to the initial question, the initial question is optimized using a large model to generate a target question.

[0044] In operation S230, an interaction is carried out with the target object based on the target question, so as to determine the matching degree between the target object and the target position based on the interaction result.

[0045] According to an embodiment of the present disclosure, the requirement information of the target position may include information such as the position name, skill requirements, and professional experience. For example: the requirement information may be a data analyst, who needs to master Pandas and Numpy, have at least 3 years of relevant experience, and those with data visualization ability are preferred. The target object may be a job seeker intending to apply for the target position. The resume information may include: information such as the name, professional skills, and professional experience of the job seeker. For example: the resume information may be Zhang, with 4 years of data analysis experience, proficient in using Python, Pandas, and Numpy, and has been responsible for large-scale data cleaning and visualization projects.

[0046] In some embodiments, a Prompt (prompt information) A can be directly constructed based on the requirement information of the target position and the resume information of the target object. The Prompt A may also include exemplary text for generating reference questions based on reference position requirements and reference resumes. Then, the Prompt A is input into the large model, and the large model generates an initial question based on a deep analysis of the requirements of the target position and the resume of the target object, with reference to the exemplary text. The initial question may be, for example, Can you briefly explain how Pandas performs data deduplication?

[0047] According to an embodiment of the present disclosure, the response status of the target object to the initial question indicates the matching degree between the response content of the target object and the initial question. The large model is used to optimize the initial question based on the matching degree to generate a target question.

[0048] In some embodiments, the direction for guiding the large model to optimize the initial question can be determined by configuring a matching degree threshold between the response content and the initial question.

[0049] For example: when the matching degree is greater than the matching degree threshold, it means that the response of the target object to the initial question is relatively accurate. The optimization direction can be to further increase the difficulty of the question, or to ask questions related to the initial question to further examine the target object's mastery of this technology or related technologies.

[0050] For example, when the matching degree is less than the matching degree threshold, it indicates that the response of the target object to the initial question deviates from the questioning direction. The optimization direction can be to reduce the difficulty of the question or adjust the questioning direction to examine the target object's mastery of other skills required for the position.

[0051] In some embodiments, Prompt B can be constructed based on the response content of the target object and the initial question. The Prompt B may also include exemplary text for optimizing the reference question based on the matching degree between the reference response and the reference question. Then, Prompt B is input into the large model. Based on analyzing the matching degree between the response content and the initial question, the large model generates an optimized target question with reference to the exemplary text. When the response content matches the question, the target question can be, for example, "Please optimize the computational efficiency of Pandas?" When the response content does not match the question, the target question can be, for example, "What is the most basic use of Pandas?"

[0052] During the interaction process, the large model can be used to dynamically optimize the interaction questions with the target object based on the response status of the target object in the previous round of interaction, generating optimized questions for in-depth examination of the target object, improving the matching degree between the questions generated by the large model and the applicant's and position requirements, and further enhancing the interaction flexibility and efficiency in the interview scenario.

[0053] In related examples, usually the same questions are generated based on the position requirements, and the examination angle is relatively one-sided, making it difficult to conduct personalized interactions for different job seekers to comprehensively examine the matching degree between the job seeker and the target position.

[0054] In view of this, the embodiments of the present disclosure utilize the semantic understanding ability of the large model to deeply analyze the matching degree between the position requirements and the skills demonstrated by the job seeker in the resume, and generate personalized initial questions for different job seekers.

[0055] In the embodiments of the present disclosure, by using the large model, through analyzing the requirement information of the target position and the resume information of the target object, initial questions can be generated, which may include the following operations: extracting requirement features from the requirement information of the target position; extracting skill features from the resume information; matching the requirement features and the skill features to obtain the skill matching degree; and using the large model to process the requirement features and the skill features based on the skill matching degree to generate initial questions.

[0056] In some embodiments, after removing special characters and segmenting the requirement information, the TF-IDF (term frequency-inverse document frequency) algorithm can be used to extract keywords from the requirement information, and the keywords can be used as requirement features. A trained language model, such as the BERT model, can also be used to generate semantic vectors that are more easily understood by large models based on the keywords, and the semantic vectors can be used as requirement features.

[0057] For example, the requirement information can be that a data analyst needs to master Pandas and Numpy, have at least 3 years of relevant experience, and those with data visualization skills are preferred. The keywords can include: python, data analysis, pandas, numpy, experience, etc.

[0058] Similarly, skill features can be extracted from the resume information based on the method described above. Then, the skill matching degree can be obtained by calculating the semantic similarity between the requirement features and the skill features.

[0059] In some embodiments, the skill matching degree can include matching degrees in multiple dimensions. For example: professional technology, work experience, etc. Different weights can be configured for different dimensions, and then the skill matching degree can be obtained after weighting according to the matching degree of each dimension.

[0060] For example, the weight of professional skills can be 50%, the weight of work experience can be 30%, and the weight of other relevant experience can be 20%. Other relevant experience includes but is not limited to: team collaboration experience, etc.

[0061] In some embodiments, the questioning direction of the initial question can be determined based on the skill matching degree. For example, different thresholds can be configured for the skill matching degree. When the skill matching degree is greater than 0.85, it indicates that the applicant has a high skill matching degree with the target position, and the questioning direction of the initial question can be mainly used to examine whether the applicant has the ability to lead project R & D. When the skill matching degree is between 0.6 and 0.85, it means that the applicant's skills can meet most of the requirements of the target position but are still lacking, and the questioning direction of the initial question can be mainly used to verify the applicant's in-depth understanding and comprehensive application ability of the skills. When the skill matching degree is less than 0.6, it means that the applicant's skills have a low matching degree with the target position, but there is still a possibility of being competent for the target position, and the questioning direction of the initial question can be mainly used to examine the applicant's basic understanding and application ability of the skills required for the target position.

[0062] Therefore, Prompt C can be constructed based on the skill matching degree, skill characteristics, and demand characteristics. The exemplary text of the initial questions that can generate different questioning directions based on different skill matching degrees can be included in Prompt C. Input Prompt C into the large model to generate the initial questions for the target object.

[0063] Figure 3 Schematically shows a schematic diagram of generating initial questions using a large model according to an embodiment of the present disclosure.

[0064] As Figure 3 shown, in this embodiment 300, the skill matching degree 313 can be calculated first based on the job requirements 311 and the resume 312. Then, call the large model based on the skill matching degree to generate the initial questions 314.

[0065] For example: The job requirements 311 can include a data analyst, and the resume 312 can include being good at python data analysis and proficient in Pandas. The calculated skill matching degree 313 can be 0.7. The initial question 314 generated by calling the large model can be Please describe the working principle of Pandas groupby.

[0066] In addition to determining the questioning direction of the initial questions based on the skill matching degree, the questioning difficulty of the initial questions can also be determined. For example: When the skill matching degree is relatively high, it can be determined that the questioning difficulty of the initial questions is relatively high.

[0067] In some embodiments, using a large model, based on the skill matching degree, processing the demand characteristics and skill characteristics to generate initial questions may include the following operations: determining the questioning difficulty of the initial questions according to the skill matching degree; and using the large model, based on the questioning difficulty, processing the demand characteristics and skill characteristics to generate initial questions.

[0068] For example: For the same skill requirements of a data analyst position, when the skill matching degree is greater than 0.9, it can be determined that the questioning difficulty of the initial questions is advanced. The initial questions can be, for example, How to optimize the Pandas groupby operation to improve the calculation performance? When the skill matching degree is between 0.7 and 0.9, it can be determined that the questioning difficulty of the initial questions is intermediate. The initial questions can be, for example, How to use NumPy to process large matrix calculations? When the skill matching degree is less than 0.7, it can be determined that the questioning difficulty of the initial questions is primary. The initial questions can be, for example, Please explain the Pandas DataFrame structure?

[0069] By using the big model to deeply analyze the job requirements and the skills described in the target object's resume, initial questions are generated in a targeted manner. Compared with using the same questions to examine the applicant's ability, it is more capable of exploring the applicant's comprehensive skills, thereby comprehensively analyzing whether the target object matches the target position, further improving the comprehensiveness and objectivity of the target object's skill assessment.

[0070] The actual interaction process usually includes multiple rounds. In order to further improve the interaction efficiency, the initial question is optimized using the big model according to the response status of the target object to the initial question to generate the target question. The following operations may be performed: in response to determining that the response status indicates that the response content matches the initial question, the big model is used to optimize the initial question according to the questioning direction of the initial question to generate the target question; and in response to determining that the response status indicates that the response content does not match the initial question, the big model is used to adjust the questioning direction of the initial question, and the target question is generated based on the adjusted questioning direction.

[0071] In some embodiments, the reply status may indicate the degree of match between the reply content and the initial question, or may indicate the degree of match between the reply content and the answer to the initial question.

[0072] In an embodiment of the present disclosure, before the initial question is optimized using a large model based on the target object's response status to the initial question and the target question is generated, the following operations may be included: obtaining the answer to the initial question and the target object's response content to the initial question; matching the response content and the answer using the large model to generate a response status.

[0073] For example: The initial question could be Please explain the main purpose of the Pandas groupby() method. The reply could be The Pandas groupby() method can be used for data grouping operations and can be used to calculate mean, sum, etc. The answer to the initial question could be The Pandas groupby() method is used for data aggregation, such as calculating mean and sum.

[0074] In some embodiments, based on the semantic understanding ability of the large model, keyword matching, semantic matching and logical matching can be performed on the reply content and the answer respectively, and the final reply status can be generated by configuring weights for the matching content of different dimensions.

[0075] For example, the keyword matching degree can be 0.85, the semantic matching degree can be 0.92, and the logical matching degree can be 0.98. Based on the configuration weights, the weighted sum of the keyword matching degree, semantic matching degree, and logical matching degree is obtained, and the matching degree between the answer content and the initial question is 0.95. The final generated answer status can indicate that the answer content matches the initial question.

[0076] The large model is used to perform multi-dimensional matching between the reply content and the answer to generate a reply status, which further improves the generalization of the reply status of the target object in terms of semantic understanding and enhances the accuracy of reply status evaluation.

[0077] Based on the reply status, the large model is used to dynamically optimize the initial question, generating personalized questions that can not only meet the job requirements but also specifically examine the abilities of job seekers, further improving the flexibility in the interaction process.

[0078] Figure 4A Schematically shows an interaction schematic diagram based on a large model according to an embodiment of the present disclosure.

[0079] As Figure 4A shown, in this embodiment 400A, in the interaction interface 410, the initial question 411 generated based on the resume and job requirements using the large model is displayed, "Please describe the working principle of xx", and the reply content 412 of the job seeker to the initial question 411.

[0080] In some embodiments, in response to determining that the reply status indicates that the reply content matches the initial question, using the large model, the initial question is optimized in accordance with the questioning direction of the initial question to generate a target question, which may include the following operations: using the large model to analyze the difficulty of the initial question to obtain the questioning difficulty of the initial question; and using the large model, in accordance with the questioning direction, by increasing the questioning difficulty of the initial question, generating a target question.

[0081] For example: First, Prompt D can be constructed based on the initial question, and the exemplary text for analyzing the difficulty to obtain the questioning difficulty based on the reference question can also be included in Prompt D. Input Prompt D into the large model to output the questioning difficulty of the initial question. For example: it can be intermediate. Then, Prompt E can be constructed based on the initial question and the questioning difficulty, and the exemplary text for increasing the difficulty in accordance with the questioning direction based on the reference question can also be included in Prompt E. Input Prompt E into the large model to output the target question. The questioning difficulty of the target question is greater than that of the initial question. For example: the questioning difficulty of the target question can be advanced. Different thresholds can also be configured for the questioning difficulty. For example: the threshold for intermediate questioning difficulty is 0.5 - 0.8, and the threshold for advanced questioning difficulty is above 0.8. The questioning difficulty of the target question can be indicated in Prompt E. For example: 0.9, or the difficulty difference of the target question relative to the initial question can be indicated in Prompt E. For example: the questioning difficulty difference is 0.3. The large model can generate a target question with a questioning difficulty of 0.8 by combining the initial questioning difficulty of 0.5 with the questioning difficulty difference of 0.3.

[0082] As Figure 4AAs shown, when the reply matches the question, by invoking the large model, in accordance with the questioning direction of the initial question, the difficulty level of the initial question is increased to generate a target question. The target question can be "How to improve the efficiency of xx algorithm?" shown in the interactive interface 420A. 421.

[0083] Based on the semantic understanding ability of the large model, in accordance with the questioning direction of the initial question, the question difficulty is increased, and a question for the new round of interaction is dynamically generated, achieving a more in-depth examination of the job seeker's skills and realizing intelligent interaction.

[0084] In some embodiments, in response to determining that the reply status indicates that the reply content does not match the initial question, the large model is used to adjust the questioning direction of the initial question, and based on the adjusted questioning direction, a target question is generated, which may include the following operations: using the large model, by analyzing the requirement information of the target position and the resume information of the target object, generating an adjusted questioning direction; using the large model, in accordance with the adjusted questioning direction, generating a target question.

[0085] For example: First, a Prompt F can be constructed based on the position requirements and resume information. In Prompt F, there can also be exemplary texts for adjusting the questioning direction based on the position requirements and resume analysis. Input Prompt F into the large model to output an adjusted questioning direction. For example: It can be the direction regarding the principle of the yy algorithm. Then, based on the called questioning direction, construct Prompt G. In Prompt G, there can also be exemplary texts for generating reference questions based on the reference questioning direction. Input Prompt G into the large model to output the target question.

[0086] As Figure 4A shown, when the reply does not match the question, the questioning direction can be adjusted by invoking the large model to generate a target question. The target question can be "What is the most basic use of the question yy in the interactive interface 430A?" 431.

[0087] By using the large model to generate an adjusted questioning direction based on the in-depth analysis of the position requirements and resume information, the flexibility of dynamically optimizing the question and the adaptability of the optimized question to the position requirements are further improved.

[0088] Figure 4B Schematically shows an interactive schematic diagram based on another embodiment of the present disclosure.

[0089] As Figure 4B shown, the difference between this embodiment 400B and 400A is that when the reply matches the question, the supplementary question generated by invoking the large model is related to the initial question. When the reply does not match the question, the target question generated by invoking the large model is a question with a reduced questioning difficulty.

[0090] In some embodiments, in response to determining that the reply status indicates that the reply content matches the initial question, the large model is used to optimize the initial question in the direction of the initial question to generate a target question, including the following operations: using the large model, by analyzing the requirement information of the target position and the resume information of the target object, generating a plurality of supplementary questions associated with the questioning direction; determining the relevance between the supplementary question and the initial question based on the reply status; and determining the target question from the plurality of supplementary questions based on the relevance.

[0091] For example: First, Prompt H can be constructed based on the position requirements and resume information, and the exemplary text for generating associated questions based on the position requirements and resume analysis can also be included in Prompt H. Input Prompt H into the large model to output a plurality of supplementary questions. Then, when the reply status indicates that the reply content matches most of the initial question but there are some defects, it is determined that the relevance between the supplementary question and the initial question is high. When the reply status indicates that the reply content matches only a small part of the initial question, it is determined that the relevance between the supplementary question and the initial question is low. Next, based on the relevance, construct Prompt I, and the exemplary text for generating supplementary questions associated with the reference question based on the relevance can also be included in Prompt I. Input Prompt G into the large model to output the target question.

[0092] Such as Figure 4B As shown, in the interactive interface 420B, the supplementary question "Please supplement how xx is used in combination with zz?" 422 is displayed.

[0093] When the reply content matches the initial question, by supplementing questions associated with the initial question, the depth of the investigation of the applicant's skills is further improved, and intelligent interaction is achieved.

[0094] In some embodiments, in response to determining that the reply status indicates that the reply content does not match the initial question, the large model is used to adjust the questioning direction of the initial question, and based on the adjusted questioning direction, a target question is generated, which may include the following operations: determining the questioning difficulty of the target question based on the reply status; and using the large model to generate the target question according to the adjusted questioning direction and the questioning difficulty of the target question.

[0095] For example, the difficulty of the target question can be determined based on the response status. For example, when the response status indicates that the response content matches most of the content of the initial question but there are some defects, the difficulty of the target question can be the same as or higher than that of the initial question. For example, if the difficulty of the initial question is intermediate, the difficulty of the target question can be intermediate or advanced. When the response status indicates that the response content does not match the initial question, the difficulty of the target question can be lower than that of the initial question. For example, the difficulty of the target question can be elementary. In addition to the difficulty level of the question, difficulty thresholds can also be configured for different difficulty levels of questions. For example, questions with a difficulty below 0.5 belong to the elementary level, questions with a difficulty between 0.5 and 0.8 belong to the intermediate level, and questions with a difficulty above 0.8 belong to the advanced level. Then, the difficulty of the target question can be added to the Prompt G described above to output the target question.

[0096] In some embodiments, when the response content does not match the initial question, the question difficulty can also be only reduced. As Figure 4B shown, in the interaction interface 430B, the target question xx's most basic use after reducing the question difficulty is shown as 432.

[0097] In some embodiments, when the response content does not match the initial question, both the question direction can be adjusted and the question difficulty can be reduced.

[0098] Figure 4C Schematically shows an interaction schematic diagram based on a large model according to another embodiment of the present disclosure.

[0099] As Figure 4C shown, the difference between this embodiment 400C and embodiment 400B is that when the response does not match the question, when calling the large model to generate the target question, both the question direction is adjusted and the question difficulty is reduced. The target question is what is the basic use of the algorithm of question yy shown in the interaction interface 430C?

[0100] When the response content does not match the initial question, based on adjusting the question direction, the question difficulty is dynamically called based on the response status, further improving the flexibility of question generation.

[0101] Figure 5 Schematically shows a block diagram of an interaction device based on a large model according to an embodiment of the present disclosure.

[0102] As Figure 5 shown, the interaction device 500 may include: an analysis module 510, an optimization module 520, and an interaction module 530.

[0103] An analysis module 510, configured to use a large model to generate an initial question by analyzing the requirement information of the target position and the resume information of the target object.

[0104] An optimization module 520, configured to optimize the initial question using a large model according to the response status of the target object to the initial question, and generate a target question.

[0105] An interaction module 530, configured to interact with the target object based on the target question, so as to determine the matching degree between the target object and the target position based on the interaction result.

[0106] According to an embodiment of the present disclosure, the optimization module includes: a first optimization sub-module and a second optimization sub-module.

[0107] The first optimization sub-module is configured to, in response to determining that the response status indicates that the response content matches the initial question, use a large model to optimize the initial question in the questioning direction of the initial question, and generate a target question.

[0108] The second optimization sub-module is configured to, in response to determining that the response status indicates that the response content does not match the initial question, use a large model to adjust the questioning direction of the initial question, and generate a target question based on the adjusted questioning direction.

[0109] According to an embodiment of the present disclosure, the second optimization sub-module includes: a parsing unit and a difficulty increasing unit.

[0110] The parsing unit is configured to use a large model to perform difficulty parsing on the initial question to obtain the questioning difficulty of the initial question.

[0111] The difficulty increasing unit is configured to use a large model to increase the questioning difficulty of the initial question in the questioning direction, and generate a target question. The questioning difficulty of the target question is greater than that of the initial question.

[0112] According to an embodiment of the present disclosure, the second optimization sub-module further includes: a first analysis unit, a first determination unit, and a second determination unit.

[0113] The first analysis unit is configured to use a large model to analyze the requirement information of the target position and the resume information of the target object, and generate a plurality of supplementary questions associated with the questioning direction.

[0114] The first determination unit is configured to determine the correlation degree between the supplementary question and the initial question based on the response status.

[0115] The second determination unit is configured to determine the target question from the plurality of supplementary questions based on the correlation degree.

[0116] According to an embodiment of the present disclosure, the first optimization sub-module includes: a second analysis unit and a first generation unit.

[0117] A second analysis unit, configured to use a large model to analyze the requirement information of the target position and the resume information of the target object, and generate an adjusted questioning direction.

[0118] A generation unit, configured to use a large model to generate a target question according to the adjusted questioning direction.

[0119] According to an embodiment of the present disclosure, the first optimization sub-module further includes: a determination unit and a second generation unit.

[0120] The determination unit is configured to determine the questioning difficulty of the target question based on the reply status.

[0121] The second generation unit is configured to use the large model to generate a target question according to the adjusted questioning direction and the questioning difficulty of the target question.

[0122] According to an embodiment of the present disclosure, the above interaction device further includes: an acquisition module and a generation module.

[0123] The acquisition module is configured to acquire the answer to the initial question and the reply content of the target object to the initial question.

[0124] The generation module is configured to use a large model to match the reply content and the answer, and generate a reply status.

[0125] According to an embodiment of the present disclosure, the matching module includes: a first extraction sub-module, a second extraction sub-module, a matching sub-module, and a generation sub-module.

[0126] The first extraction sub-module is configured to extract requirement features from the requirement information of the target position.

[0127] The second extraction sub-module is configured to extract skill features from the resume information.

[0128] The matching sub-module is configured to match the requirement features and the skill features to obtain a skill matching degree.

[0129] The generation sub-module is configured to use a large model to process the requirement features and the skill features based on the skill matching degree, and generate an initial question.

[0130] According to an embodiment of the present disclosure, the generation sub-module includes: a determination unit and a generation unit.

[0131] The determination unit is configured to determine the questioning difficulty of the initial question according to the skill matching degree.

[0132] The generation unit is configured to use a large model to process the requirement features and the skill features based on the questioning difficulty, and generate an initial question.

[0133] Figure 6 A block diagram of an agent for interaction according to an embodiment of the present disclosure is schematically shown.

[0134] In an embodiment of the present disclosure, inspired by the von Neumann architecture in modern computer theory, as Figure 6 shown, the AI agent 600 may include three core modules: an input module 610, an output module 620, and a processing module 630. The processing module 630 may include a control unit 631, a storage unit 632, and an arithmetic unit 633.

[0135] The input module 610 is responsible for receiving or sensing information such as queries, requests, instructions, signals, or data from the outside world (e.g., users or the external environment), and converting it into a format that the AI agent 600 can understand and process. The input module 610 is the primary link for the AI agent 600 to interact with the outside world, enabling the AI agent 600 to efficiently and accurately obtain necessary sensory information from the outside world and respond to this information.

[0136] In an example, the input information received by using the input module 610 may include the requirement information of the target position and the resume information of the target object described above.

[0137] In an example, the processing module 630 is the core support for the AI agent 600 to handle complex tasks. The processing module 630 may determine a target task based on the input information received by the input module 1010, determine a large model based on the target task, and output questions for each interaction round by invoking the large model to execute the interaction method based on the large model described above.

[0138] In an example, the control unit 631 in the processing module 630 will continuously interact with the storage unit 632, the arithmetic unit 633, and / or the output module 620 during operation. However, it should be noted that in an embodiment of the present disclosure, the control unit 631 acts as a single initiator to initiate communication with the storage unit 632, the arithmetic unit 633, and / or the output module 620, and there is no communication coupling between the storage unit 632, the arithmetic unit 633, and the output module 620.

[0139] In an example, the performance of the control unit 631 may be closely related to the large model on which the AI agent 600 is based. To fully utilize the capabilities of the large language model, the internal structure of the control unit 631 can be designed to be highly configurable and extensible to cope with various different types of tasks and requirements in real scenarios.

[0140] The storage unit 632 may be responsible for memorizing information such as historical interactions and event streams. The text generated in each round as described above may be included in the storage unit 632.

[0141] In the example, after the AI agent 600 obtains an information optimization request, the AI agent 600 can call a large model to execute a task corresponding to the input information and output corresponding text. The corresponding text can be stored in the storage unit 632. The AI agent 600 can retrieve relevant data resources from the storage unit 632 and feedback them to the control unit 631. Then, the control unit 631 can use the feedback data resources to generate questions for each interaction round. Relevant data resources can also be retrieved from the storage unit 632 and feedback them to the control unit 631. Then, the control unit 631 can use the returned data resources to generate questions for each interaction round. And the questions for each interaction round are passed to the output module 620.

[0142] The operation unit 633 can be regarded as a predefined tool library. As described above, the renderer and display controls can be included in the operation unit 633.

[0143] In the example, when the AI agent 600 needs to render multiple output data, relevant renderers and display controls can be called from the operation unit 633 and feedback them to the control unit 632. Then, the control unit 632 can use the feedback renderers and display controls to render the first search result and pass the first search result to the output module 620. It can be understood that although the large language model has excellent language understanding and generation capabilities, like humans, the tasks it can solve without any tools are very limited. When the AI agent 600 is given the ability to call tools, it can achieve tasks such as performing mathematical operations with the help of a calculator, performing data analysis with the help of python, and performing prediction tasks with the help of a search engine.

[0144] In the example, the output module 620 can output the questions for each interaction round described above.

[0145] The AI agent 600 according to the embodiment of the present disclosure can simply and effectively improve the degree of intelligence, and improve flexibility and versatility.

[0146] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0147] According to an embodiment of the present disclosure, an electronic device includes: 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 to enable the at least one processor to execute the method as described above.

[0148] According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are for causing a computer to execute the method as described above.

[0149] According to an embodiment of the present disclosure, a computer program product includes a computer program which, when executed by a processor, implements the method as described above.

[0150] Figure 7 A schematic block diagram of an exemplary electronic device 700 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as, for example, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, for example, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely exemplary and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0151] As Figure 7 shown, the device 700 includes a computing unit 701 which can perform various appropriate actions and processes according to the computer program stored in a read-only memory (ROM) 702 or the computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0152] A plurality of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0153] The computing unit 701 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 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, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 executes the various methods and processes described above, such as the interaction method based on a large model. For example, in some embodiments, the interaction method based on a large model can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the interaction method based on a large model described above can be executed. Alternatively, in other embodiments, the computing unit 701 can be configured to execute the interaction method based on a large model in any other suitable manner (e.g., by means of firmware).

[0154] The various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuitry, integrated circuit systems, field-programmable gate arrays (FPGA), application-specific integrated circuits (ASIC), application-specific standard products (ASSP), systems-on-a-chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments 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, receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0155] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program codes can be executed entirely on the machine, partially on the machine, as an independent software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0156] In the context of this disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0157] In order to provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0158] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can 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.

[0159] A computer system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The client-server relationship is generated by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

[0160] It should be understood that the various forms of processes shown above can be used, with steps reordered, added or deleted. For example, the steps described 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 no limitation is imposed herein.

[0161] The above specific embodiments do not constitute a limitation on the protection scope of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub - combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the protection scope of this disclosure.

Claims

1. An interactive method based on a large model, comprising: Using the big model, we generate initial questions by analyzing the target job requirements and the target person's resume information; According to the answer status of the target object to the initial question, the initial question is optimized by using the large model to generate a target question; as well as Interact with the target object based on the target question, so as to determine the matching degree between the target object and the target position based on the interaction result.

2. The method according to claim 1, wherein: The step of optimizing the initial question using the large model according to the target object's answer status to the initial question to generate the target question includes: In response to determining that the reply status indicates that the reply content matches the initial question, optimizing the initial question according to the questioning direction of the initial question using the large model to generate the target question; and In response to determining that the reply status indicates that the reply content does not match the initial question, the large model is used to adjust the questioning direction of the initial question, and the target question is generated based on the adjusted questioning direction.

3. The method according to claim 2, wherein: In response to determining that the reply status indicates that the reply content matches the initial question, optimizing the initial question according to the questioning direction of the initial question using the large model to generate the target question includes: Analyzing the difficulty of the initial question using the large model to obtain the difficulty of the initial question; and The target question is generated by using the large model and increasing the difficulty of the initial question according to the questioning direction; the difficulty of the target question is greater than the difficulty of the initial question.

4. The method according to claim 2, wherein: In response to determining that the reply status indicates that the reply content matches the initial question, optimizing the initial question according to the questioning direction of the initial question using the large model to generate the target question, further comprising: By using the large model, the requirement information of the target position and the resume information of the target object are analyzed to generate a plurality of supplementary questions associated with the questioning direction; Based on the response status, determining a relevance of the supplemental question to the initial question; and Based on the degree of association, the target question is determined from the plurality of supplementary questions.

5. The method according to claim 2, wherein: In response to determining that the reply status indicates that the reply content does not match the initial question, adjusting the questioning direction of the initial question using the large model, and generating the target question based on the adjusted questioning direction, comprising: Using the large model, by analyzing the requirement information of the target position and the resume information of the target object, an adjusted question direction is generated; and The target question is generated by using the large model according to the adjusted question direction.

6. The method according to claim 5, wherein: In response to determining that the reply status indicates that the reply content does not match the initial question, adjusting the questioning direction of the initial question using the large model, and generating the target question based on the adjusted questioning direction, further comprising: Determining the difficulty of the target question based on the answer status; and The target question is generated by utilizing the large model according to the adjusted question direction and the question difficulty of the target question.

7. According to any one of claims 2 to 6, before optimizing the initial question using the large model according to the target object's answer status to the initial question to generate the target question, the method further comprises: Obtaining the answer to the initial question and the target object's reply to the initial question; as well as The reply content and the answer are matched using the large model to generate the reply status.

8. The method according to claim 1, wherein: The large model is used to analyze the target position's demand information and the target object's resume information to generate initial questions, including: Extracting demand features from demand information of the target position; Extracting skill features from the resume information; Matching the demand characteristics with the skill characteristics to obtain a skill matching degree; and The large model is used to process the demand characteristics and skill characteristics based on the skill matching degree to generate the initial question.

9. The method according to claim 8, wherein: The step of using the large model to process the demand characteristics and skill characteristics based on the skill matching degree to generate the initial question includes: Determining the difficulty of asking the initial question based on the skill match; and The large model is used to process the demand characteristics and skill characteristics based on the question difficulty to generate the initial question.

10. An interactive device based on a large model, comprising: The analysis module is used to generate initial questions by using the big model to analyze the target position’s demand information and the target object’s resume information; An optimization module, configured to optimize the initial question using the large model according to the target object's answer status to the initial question, and generate a target question; as well as An interaction module is used to interact with the target object based on the target question, so as to determine the matching degree between the target object and the target position based on the interaction result.

11. An intelligent agent for interaction, comprising: An input module, used for receiving input information; a processing module, configured to determine a target task based on the input information received by the input module, determine a large model based on the target task, and obtain output information by calling the large model to execute the method according to any one of claims 1 to 9; as well as An output module is used to output the output information obtained by the processing module.

12. 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 9.

13. 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 1-9.

14. 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 9.

Images