Interview answer evaluation method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and discloses an interview answer evaluation method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining an interview question and an answer text corresponding to the interview question; determining a scoring detailed rule model according to the question type of the interview question; generating evaluation rules for the interview question by using a scoring rule model; and determining the evaluation of the answer text according to the evaluation rules. In this way, different evaluation rules are generated for different types of interview questions, and then the corresponding evaluation rules are used for evaluating the answer text of the interview questions. The evaluation details of interview evaluation can be more targeted, so that the finally obtained evaluation result of the answer text can be more accurate.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an interview answer evaluation method and device, electronic equipment, and storage medium. Background Art

[0002] In the process of job hunting, users usually go through the interview stage. In order to make the interview evaluation more objective, recruiters began to explore the use of models to automatically evaluate interview answers. In related technologies, the classification model or regression model is trained in advance with labeled training data to obtain a review model. When in use, the interview answer text of the job seeker is obtained; the text features in the interview answer text are extracted using methods such as Bag of Words, TF-IDF (Term Frequency-Inverse Document Frequency) or Word Embeddings; the text features are input into the review model to obtain the evaluation score. However, the evaluation results of interview answers in the above manner are less accurate.

[0003] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present application, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0004] In order to provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. The summary is not an extensive review, nor is it intended to identify key / critical components or delineate the scope of protection of these embodiments, but rather serves as a prelude to the detailed description that follows.

[0005] The embodiments of the present application provide a method and device for evaluating interview answers, an electronic device, and a storage medium to improve the accuracy of the evaluation of interview answers.

[0006] An embodiment of the present application provides a method for evaluating interview answers, comprising: obtaining interview questions and answer texts corresponding to the interview questions; determining a scoring criterion model according to the question type of the interview questions; generating evaluation criteria for the interview questions using the scoring criterion model; and determining an evaluation of the answer text according to the evaluation criteria.

[0007] In the above implementation, different evaluation criteria are generated for different types of interview questions, and the corresponding evaluation criteria are used to evaluate the answer texts of the interview questions. Since the evaluation criteria for interview evaluation can be more targeted, the evaluation results of the answer texts obtained in the end can be more accurate.

[0008] Furthermore, the question type of the interview question is determined in the following manner: the interview question and a preset first prompt word are input into a preset first large language model to obtain the question type corresponding to the interview question; the first prompt word is used to indicate the question type of the content input into the first large language model determined by the first large language model with reference to the preset type and the definition of the preset type.

[0009] In the above implementation, the first language model can be used to automatically classify interview questions, which can make it more efficient to determine the types of interview questions compared to manually determining the types of interview questions.

[0010] Furthermore, determining a scoring criterion model according to the question type of the interview question includes: using a preset model database to perform a table lookup operation on the question type to obtain a scoring criterion model corresponding to the question type; the model database stores the correspondence between the question type and the scoring criterion model.

[0011] In the above implementation, by predetermining the scoring rule model corresponding to each question type and storing the corresponding relationship in the model database, in subsequent use, the scoring rule model corresponding to the question type can be quickly and accurately determined by searching directly through the model database.

[0012] Furthermore, the scoring criteria model is used to generate evaluation criteria for the interview question, including: obtaining a second prompt word corresponding to the scoring criteria model; inputting the interview question and the second prompt word into the scoring criteria model to obtain evaluation criteria corresponding to the interview question; the second prompt word is used to indicate the evaluation criteria of the content input into the scoring criteria model determined by the scoring criteria model with reference to preset scoring dimensions.

[0013] In the above implementation, the evaluation criteria corresponding to the interview questions are generated by using the scoring criteria model, which can automatically determine the evaluation criteria corresponding to the interview questions. Compared with manually determining the evaluation criteria, the efficiency of determining the evaluation criteria for the interview questions can be higher.

[0014] Furthermore, the evaluation of the answer text is determined according to the evaluation criteria, including: inputting the answer text, the evaluation criteria and a preset third prompt word into a preset second largest language model to obtain a comment corresponding to the answer text; the third prompt word is used to indicate that the comment of the answer text input into the second largest language model is determined by the second largest language model with reference to the evaluation criteria.

[0015] It is understandable that these methods cannot fully capture the semantic information of the text by relying on traditional feature extraction methods such as bag-of-words models, TF-IDF or word vectors. In the above implementation, the large language model can better understand and process natural language, capture the deep semantic information of the text, and thus improve the accuracy of the score. At the same time, the review model obtained by training the classification model or regression model using the labeled training data has insufficient generalization ability when facing unseen interview answers, resulting in unstable evaluation results. The large language model has learned a lot of language knowledge and contextual information during the pre-training process, has stronger generalization ability, and can handle various types of interview answers more stably. At the same time, when pre-training the large language model, a large amount of labeled data is not required, which can reduce the dependence on a large amount of labeled data and reduce labor costs.

[0016] Furthermore, the evaluation criteria include multiple criteria and weights corresponding to each of the criteria; the comments include scores corresponding to each of the criteria; determining the evaluation of the answer text based on the evaluation criteria also includes: weighting the scores corresponding to each of the criteria according to the weights corresponding to each of the criteria to obtain a target score; and determining the target score as the evaluation of the answer text.

[0017] It is understandable that the same question may have multiple dimensions to answer, and the importance of different dimensions may be different. Therefore, among the evaluation criteria for a question, some evaluation criteria are more important and some are less important. By weighting the scores corresponding to different evaluation criteria, the evaluation of the answer text can be made more objective and more in line with the user's expectations.

[0018] Furthermore, the answer text corresponding to the interview question is obtained in the following manner: receiving the interviewee's answer audio; converting the answer audio into alternative text; correcting the alternative text to obtain the answer text.

[0019] In the above-mentioned embodiment, by correcting the alternative texts converted from the answer audio and then using the corrected answer text for evaluation, it is possible to avoid problems such as typos in the audio-to-text conversion process from affecting the interviewer's interview score, thereby making the evaluation of the answer text more objective and accurate.

[0020] An embodiment of the present application provides an electronic device, including a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the above-mentioned interview answer evaluation method.

[0021] In some embodiments, the storage medium stores computer executable instructions, and when the computer executable instructions are called and executed by the processor, the computer executable instructions enable the processor to implement the above-mentioned interview answer evaluation method.

[0022] The above general description and the following description are exemplary and explanatory only and are not intended to limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] One or more embodiments are exemplarily described by corresponding drawings, and these exemplary descriptions and drawings do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are considered similar elements, and the drawings do not constitute a scale limitation, and wherein:

[0024] Figure 1 It is a schematic diagram of an interview answer evaluation method provided in an embodiment of the present application;

[0025] Figure 2 is a schematic diagram of an interview answer evaluation device provided in an embodiment of the present application;

[0026] Figure 3 It is a schematic diagram of an electronic device provided in an embodiment of the present application.

[0027] Reference numerals:

[0028] 1: Acquisition module; 2: Model determination module; 3: Detail determination module; 4: Evaluation module; 5: Processor; 6: Bus; 7: Memory; 8: Communication interface. DETAILED DESCRIPTION

[0029] In order to be able to understand the features and technical contents of the embodiments of the present application in more detail, the implementation of the embodiments of the present application is described in detail below in conjunction with the accompanying drawings. The attached drawings are for reference only and are not used to limit the embodiments of the present application. In the following technical description, for the convenience of explanation, a full understanding of the disclosed embodiments is provided through multiple details. However, one or more embodiments can still be implemented without these details. In other cases, to simplify the drawings, well-known structures and devices can be simplified for display.

[0030] The terms "first", "second", etc. in the specification and claims of the embodiments of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the embodiments of the present application described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0031] Unless otherwise stated, the term "plurality" means two or more.

[0032] The term "correspondence" may refer to an association relationship or a binding relationship. The correspondence between A and B means that there is an association relationship or a binding relationship between A and B.

[0033] Embodiment 1

[0034] The present application embodiment provides a method for evaluating interview answers. Figure 1 As shown, Figure 1 A flowchart of a method for evaluating interview answers provided in an embodiment of the present application includes:

[0035] Step S101, obtaining interview questions and answer texts corresponding to the interview questions.

[0036] In some embodiments, the interview question may be obtained by receiving a question text from an interviewer or obtaining a question text from a preset interview system.

[0037] In the above embodiment, receiving the question text of the interviewer may include: receiving the question audio of the interviewer; converting the question audio into question text; and correcting the question text to obtain the question text.

[0038] Among them, ASR (language recognition technology) technology can be used to convert the question audio into question text.

[0039] Among them, correcting the question text to obtain the title text can be: inputting the question text and the fourth prompt word into a preset third language model, and obtaining the corrected question text as the title text; the fourth prompt word is used to instruct the third language model to correct the content input into the third language model.

[0040] In some embodiments, the answer text corresponding to the interview question can be obtained in the following manner: receiving the interviewer's answer audio; converting the answer audio into alternative text; correcting the alternative text to obtain the answer text.

[0041] Among them, ASR technology can be used to convert the answer audio into alternative text.

[0042] Correcting the candidate texts to obtain the answer text may be as follows: inputting the answer text and the fourth prompt word into a preset third language model to obtain the corrected candidate text as the answer text.

[0043] For example, the fourth prompt word is as follows: #Role: A senior judge of a corporate recruitment interview is responsible for reviewing the interview content (speech recognition results) translated from the interview recording and correcting the errors that may be generated by the interview system during the speech-to-text conversion process. The corrected text will be used as the archive record of the interview candidate. #Correction Principles: Correction Prerequisite: Correction can only be carried out if it is very certain that "it will not affect the fairness of the interview result judgment"; the direct goal of correction is to improve the readability of the interview content, and the core is still to serve the objective judgment of the interview results. Based on your understanding of the principles of speech recognition (ASR) and common errors, focus on correcting the errors introduced by the ASR link to improve the readability of the interview text record. For example, if the candidate's answer mentions "database affairs", it should be corrected to "database affairs"; "e-commerce order system" should be corrected to "e-commerce order system". The candidate's answers must not be corrected or changed, even if they contain knowledge or factual errors, and the content cannot be added, modified or inferred based on personal understanding or external knowledge; for example, if the candidate's answer mentions "TCP is connectionless and UDP is connected", the candidate's answer is wrong and we cannot correct it. #Correction Rules:1. If the original sentence looks good, you don't have to modify it. 2. Keep the modal particles in the original sentence and keep their positions in the sentence unchanged; keep the sentence structure and word order unchanged. 3. Determine whether there are typos in common terms, such as translation errors caused by differences in accent pronunciation. If so, make necessary corrections, and give priority to using appropriate words with similar pronunciations when correcting. 4. Determine whether there are translation errors in professional terms. If so, use professional terms with the closest pronunciations in Chinese and English to correct them. 5. Everyone speaks differently in an interview, such as problems with expression coherence. It is necessary to determine whether there are appropriate punctuation divisions in the text. If the punctuation is seriously unreasonable, you can add appropriate punctuation. Before performing any correction operation, you must first check whether the Correction Principles are violated. If there is a possibility of violation, abandon the correction operation. #Output Format: The returned results are output in json format: the text after error correction, represented by the field content, string type. The processing result is represented by the code field. A value of 0 indicates that no error correction is required, and content returns an empty string. A value of 1 indicates that the error has been corrected, and content returns the corrected content. A value of 2 indicates that the processing failed, and content returns an empty string. Example: {"code":1,"content":"XXX"}.

[0044] Specifically, the text selected by the device is "What is the difference between TCP and UDP? TCP is connectionless, UDP is connection-oriented, TCP is unreliable, and UDP is reliable.". The alternative text is corrected by the fourth prompt word, and {"code":1,"content":"What is the difference between TCP and UDP? TCP is connectionless, UDP is connection-oriented, TCP is unreliable, and UDP is reliable."} is obtained. The text selected by the device is "What are the main tasks of this position?" The alternative text is corrected by the fourth prompt word, and {"code":0,"content":"What are the main tasks of this position?"} is obtained.

[0045] Exemplarily, the interview questions may be displayed on a display screen of an electronic device, and the interviewee may enter the answers into the electronic device in the form of text, and the text entered into the electronic device by the interviewee is used as the answer text.

[0046] Step S102, determining a scoring criteria model according to the type of the interview question.

[0047] In some embodiments, the question types may include: general ability questions and professional technical questions. General ability questions include, for example, how to communicate, how to work in a team, etc. Professional technical questions include, for example, programming technical questions, software design questions, etc.

[0048] In some embodiments, the question type of an interview question can be determined in the following manner: the interview question and a preset first prompt word are input into a preset first language model to obtain the question type corresponding to the interview question; the first prompt word is used to indicate the question type of the content input into the first language model determined by the first language model with reference to the preset type and the definition of the preset type.

[0049] In the above embodiment, the first language model, for example, ChatGPT (a chatbot program), can collect sample interview questions and sample question types corresponding to the sample interview questions. The first language model is trained using the sample question types corresponding to the sample interview questions. The trained first language model can determine the question type corresponding to the interview question with reference to the previous decomposition logic.

[0050] In the above embodiment, the interview question and the preset first prompt word are input into the preset first language model to obtain the question type corresponding to the interview question, which can be: the interview question and the preset first prompt word are combined to obtain a first combined text. The first combined text is input into the preset first language model to obtain the question type corresponding to the interview question.

[0051] Among them, the first specific character can be set in the preset first prompt word, the interview question can be filled in the first specific character, or the first specific character can be replaced with the interview question to realize the combination of the interview question and the preset first prompt word.

[0052] It can be understood that the first specific character can be a single character or a combination of multiple characters, such as double quotes, single quotes, slashes, colons, percent signs, letters, etc.

[0053] Exemplarily, the first prompt word is such as: The known question types are as follows: Basic knowledge: one or more basic knowledge in the field of Internet development, usually related to concepts, principles, and basic skills. Functional design: design a certain function / system, provide implementation solutions for specified scenarios, ask how to solve specified problems, etc. General ability: questions are used to test candidates' communication skills, teamwork and other abilities. Others: questions that are not included in the above options. Among the above question types, select a type that best matches the question "%s" and output it in Json format, the format is: {"type":String}. Among them, replace %s with the interview question.

[0054] In some embodiments, step S102 may include: using a preset model database to perform a table lookup operation on the question type to obtain a scoring criterion model corresponding to the question type; the model database stores the correspondence between the question type and the scoring criterion model.

[0055] Exemplarily, the scoring criterion model may include: eight-part essay model, competency model, general technology model, communication ability model, teamwork model, etc. Among them, for professional and technical questions, eight-part essay model, competency model, general technology model, etc. may be matched. For general ability questions, communication ability model, teamwork model, etc. may be matched.

[0056] Specifically, suppose the question is "How to work in a team". Input "How to work in a team" and the first prompt word into the preset first language model, and obtain the question type corresponding to the interview question as "Teamwork". Use the preset model database to look up the question type "Teamwork" and obtain the scoring criterion model as "Teamwork Model".

[0057] In some embodiments, sample interview questions and evaluation criteria corresponding to the sample interview questions can be collected. The LLM model (large language model) is trained using the evaluation criteria corresponding to the sample interview questions. The trained large language model can refer to the previous evaluation criteria to determine the logic and corresponding evaluation criteria of the interview questions. It can be understood that different scoring criteria models can be obtained by training the large language model using different training data. For example: for the teamwork model, only sample interview questions of the teamwork type and the corresponding evaluation criteria are used to train the large language model. For the general technology model, only sample interview questions of the general technology type and the corresponding evaluation criteria are used to train the large language model.

[0058] Step S103, generating evaluation criteria for the interview questions using the scoring criteria model.

[0059] Among them, the evaluation details may include scoring criteria, scoring items and scoring weights.

[0060] In some embodiments, step S103 may include: obtaining a second prompt word corresponding to the scoring criteria model; inputting the interview question and the second prompt word into the scoring criteria model to obtain evaluation criteria corresponding to the interview question; the second prompt word is used to indicate the evaluation criteria of the content input into the scoring criteria model determined by the scoring criteria model with reference to preset scoring dimensions.

[0061] In the above embodiment, obtaining the second prompt word corresponding to the scoring rule model may be: using a preset prompt word database to perform a table lookup operation on the scoring rule model to obtain the second prompt word corresponding to the scoring rule model; the prompt word database stores second prompt words corresponding to different scoring rule models.

[0062] In the above embodiment, the interview question and the second prompt word are input into the scoring criterion model to obtain the evaluation criterion corresponding to the interview question, which may be: the interview question and the preset second prompt word are combined to obtain a second combined text. The second combined text is input into the scoring criterion model to obtain the evaluation criterion corresponding to the interview question.

[0063] Among them, the second specific character can be set in the preset second prompt word, so that the interview question can be filled in the second specific character, or the interview question can be replaced in the second specific character to realize the combination of the interview question and the preset second prompt word.

[0064] It can be understood that the second specific character can be a single character or a combination of multiple characters.

[0065] Exemplarily, the second prompt word, for example: Scoring dimension: Basis: Do you have some understanding of the concepts of the knowledge points involved in the question, know what they are, what they are used for, and what problems they are often used to solve. Principle: Do you understand the deep principles behind the knowledge points involved in the question, know how to design them, why they are designed in this way, why they can solve certain problems or achieve certain effects, etc. In the context of campus recruitment, when a candidate (college student) is asked "{{question}}", what is the weight of the above dimensions in determining the candidate's mastery of this question? If the question does not require an answer to the principle, the principle weight should be 0. Usually, for questions that describe a process / flow and the process / flow is long or complex, there is no need to answer the principle. If it is a tool-related question and the principle does not need to be understood in daily use, there is no need to answer the principle. Please output the corresponding weight and specific evaluation content of each dimension. Output in JSON, the format is as follows: {basis:{weight:int,content:string},principle:{weight:int,content:string}}.

[0066] Among them, "{{question}}" can be replaced by the interview question. Specifically, taking the interview question "Do you understand the volatile keyword? What problem does it solve?" as an example, the scoring criterion model generates the following scoring criterion based on the second prompt word: {"basis":{"weight":60,"content":"Understand the basic concept of the volatile keyword and know that it is a keyword in Java used to declare variables. Understand that volatile is mainly used to solve visibility problems in multi-threaded environments and ensure that modifications to shared variables by one thread are immediately visible to other threads."},"principle":{"weight":40,"content":"Understand how the volatile keyword works, including the concept of memory barriers and how it prevents instruction reordering. Understand the happens-before principle in the Java Memory Model (JMM) and how volatile guarantees visibility and ordering but not atomicity. Be able to explain why volatile cannot guarantee thread safety for compound operations."}}.

[0067] It is understandable that different scoring rule models correspond to different second prompt words. For example, different scoring dimensions exist in different second prompt words.

[0068] Step S104, determining the evaluation of the answer text according to the evaluation rules.

[0069] The comments may include the score corresponding to the answer text, or the grade corresponding to the answer text.

[0070] In some embodiments, step S104 may include: inputting the answer text, evaluation criteria and a preset third prompt word into a preset second largest language model to obtain a comment corresponding to the answer text; the third prompt word is used to indicate the comment of the answer text input into the second largest language model determined by the second largest language model with reference to the evaluation criteria.

[0071] In the above embodiment, the second largest language model, for example, ChatGPT, can collect sample answer texts, sample evaluation details, and sample comments corresponding to the sample answer texts and the sample evaluation details. The second largest language model is trained using the aforementioned collected data. The trained second largest language model can determine the comments of the answer text with reference to the previous evaluation logic.

[0072] In the above embodiment, the answer text, the evaluation details and the preset third prompt word are input into the preset second largest language model to obtain the comment corresponding to the answer text, which can be: the answer text, the evaluation details and the preset third prompt word are combined to obtain a third combined text. The third combined text is input into the second largest language model to obtain the comment of the answer text.

[0073] The third specific character and the fourth specific character can be set in the preset third prompt word. The answer text is filled in the third specific character, or the answer text is used to replace the third specific character; the evaluation details are filled in the fourth specific character, or the evaluation details are used to replace the fourth specific character; thereby realizing the combination of the answer text, the evaluation details and the preset third prompt word.

[0074] It can be understood that the third specific character and the fourth specific character can be a single character or a combination of multiple characters.

[0075] In other embodiments, step S104 may include: inputting the interview question, answer text, evaluation criteria and a preset fifth prompt word into a preset third language model to obtain a comment corresponding to the answer text; the fifth prompt word is used to indicate the comment of the answer text input into the third language model determined by the third language model with reference to the interview question and the evaluation criteria.

[0076] In the above embodiment, the interview question, the answer text, the evaluation details and the preset fifth prompt word are input into the preset third language model to obtain the comment corresponding to the answer text, which can be: the interview question, the answer text, the evaluation details and the preset fifth prompt word are combined to obtain a fourth combined text. The fourth combined text is input into the third language model to obtain the comment of the answer text.

[0077] In the above embodiment, the third language model, for example, ChatGPT, can collect sample interview questions, sample answer texts, sample evaluation rules, and sample comments corresponding to the sample interview questions, sample answer texts, and sample evaluation rules. The third language model is trained using the aforementioned collected data. The trained third language model can determine the comments of the answer text with reference to the previous evaluation logic.

[0078] The fifth specific character, the sixth specific character and the seventh specific character can be set in the preset fifth prompt word. The fifth specific character can be filled with an interview question, or the fifth specific character can be replaced with an interview question; the sixth specific character can be filled with an answer text, or the sixth specific character can be replaced with an answer text; the seventh specific character can be filled with evaluation details, or the seventh specific character can be replaced with evaluation details; and the combination of the interview question, the answer text, the evaluation details and the preset fifth prompt word can be realized.

[0079] For example, background: You are a senior interviewer, and you are conducting a campus recruitment interview. You need to score the student's answer in multiple dimensions based on the set scoring criteria and your own professional knowledge. Question: {{question}}, answer: {{answer}}, scoring dimension: {{dimension}}. As a professional interviewer, give the corresponding evaluation reason (reason) and score (score) according to the test points and scoring rules described in the scoring dimension. The evaluation reason content is the analysis of the candidate's answer. The analysis needs to comply with the following rules: Since it is an interview scenario, it is normal for the candidate's answer to be colloquial. The analysis should not contain content such as the candidate's answer being too colloquial. It is normal to have slips of the tongue in professional terms. Do not mention any content related to slips of the tongue in the analysis. Do not mention any content related to the scoring dimension in the analysis. There is no need to give a standard answer in the comment, only a summary evaluation. The score of different dimensions is output in the format of "dimension: score". The result is output in JSON format, and the format is as follows: {"reason": string, "score": {{format}}}.

[0080] The interview question can be filled in {{question}}, the answer text can be filled in {{answer}}, and the evaluation criteria can be filled in {{dimension}}. That is, {{question}} is the fifth specific character, {{answer}} is the sixth specific character, and {{dimension}} is the seventh specific character.

[0081] It is understandable that the evaluation may include a corresponding reasoning process in addition to the comments, wherein the reasoning process may be the reason why the large language model gives the comment, and the reason may include a specific description of the advantages and disadvantages of the interviewee's answer.

[0082] In some embodiments, the evaluation criteria include multiple criteria and weights corresponding to each criterion; the comments include scores corresponding to each criterion; step S104 also includes: weighting the scores corresponding to each criterion according to the weights corresponding to each criterion to obtain a target score; and determining the target score as the evaluation of the answer text.

[0083] For example, interview questions include: "Do you know the volatile keyword? What problem does it solve?" Answer text, for example: "volatile is a keyword, and its main purpose is to modify class member variables and static member variables of a class. It has two main functions. The first is to ensure the visibility of data between threads. That is to say, when a thread modifies a data, the modified data can be immediately seen by other threads, because the volatile keyword will force the modified data to be written to the main memory immediately. Then the volatile keyword, its other function is to prohibit instruction reordering. The underlying principle is to add a memory barrier. The instructions on both sides of the memory barrier cannot be reordered through this memory barrier. Thank you." Scoring rules, for example: {" basis":{"weight":60,"content":"Understand the basic concept of the volatile keyword and know that it is a keyword in Java used to declare variables. Understand that volatile is mainly used to solve visibility problems in multi-threaded environments and ensure that modifications to variables by one thread are immediately visible to other threads. "},"principle":{"weight":40,"content":"Understand how the volatile keyword works, including the concept of memory barriers and how it prevents instruction reordering. Understand the happens-before principle in the Java Memory Model (JMM) and how volatile guarantees visibility and order but not atomicity. Be able to explain why volatile cannot guarantee thread safety for compound operations. "}}.

[0084] The above interview questions, answer texts, scoring rules and preset fifth prompt words are input into the preset third language model to obtain the evaluation given by the large model based on the scoring rules, for example: {"reason":"The candidate has a good understanding of the basic concept of the volatile keyword and can explain its visibility and order issues in a multi-threaded environment. The candidate mentioned the memory visibility problem and the instruction reordering problem, and can explain how volatile solves these problems. However, the candidate's understanding of the working principle of the volatile keyword and the happens-before principle in the Java Memory Model (JMM) is not deep enough, and failed to explain in detail the concept of memory barriers and why volatile cannot guarantee the thread safety of compound operations.","score":{"basis":8.5,"principle":6.0}}. Among them, the "reason" part is the reasoning process.

[0085] Weight the scores for each item according to the weights of each item. For example, the basis item is given 8.5 points, with a weight of 60%, so the score for the basis item is 8.5×0.6=5.1. The principle item is given 6 points, with a weight of 40%, so the score for the principle item is 6×0.4=2.4. The total score is basis+principle, that is, 5.1+2.4=7.5 points. At this time, the target score is 7.5 points. The target score, the basis item score, the principle item score and the reasoning process are used as the evaluation of the answer text.

[0086] Embodiment 2

[0087] Based on the same inventive concept, the present application embodiment provides an interview answer evaluation device, such as Figure 2 As shown, the interview answer evaluation device includes: an acquisition module 1, a model determination module 2, a rule determination module 3 and an evaluation module 4. Among them, the acquisition module 1 is used to obtain interview questions and answer texts corresponding to the interview questions; the model determination module 2 is used to determine the scoring rule model according to the question type of the interview question; the rule determination module 3 is used to generate evaluation rules for the interview question using the scoring rule model; the evaluation module 4 is used to determine the evaluation of the answer text according to the evaluation rules.

[0088] In some embodiments, the model determination module 2 can determine the question type of the interview question in the following manner: input the interview question and a preset first prompt word into a preset first language model to obtain the question type corresponding to the interview question; the first prompt word is used to indicate the question type of the content input into the first language model determined by the first language model with reference to the preset type and the definition of the preset type.

[0089] In some embodiments, the model determination module 2 can determine the scoring criterion model according to the question type of the interview question in the following manner: using a preset model database, performing a table lookup operation on the question type to obtain the scoring criterion model corresponding to the question type; the model database stores the correspondence between the question type and the scoring criterion model.

[0090] In some embodiments, the criteria determination module 3 can generate evaluation criteria for interview questions using the scoring criteria model in the following manner: obtaining a second prompt word corresponding to the scoring criteria model; inputting the interview question and the second prompt word into the scoring criteria model to obtain evaluation criteria corresponding to the interview question; the second prompt word is used to indicate the evaluation criteria of the content input into the scoring criteria model determined by the scoring criteria model with reference to preset scoring dimensions.

[0091] In some embodiments, the evaluation module 4 can determine the evaluation of the answer text according to the evaluation criteria in the following manner: input the answer text, the evaluation criteria and a preset third prompt word into a preset second largest language model to obtain a comment corresponding to the answer text; the third prompt word is used to indicate the comment of the answer text input into the second largest language model determined by the second largest language model with reference to the evaluation criteria.

[0092] In some embodiments, the evaluation criteria include multiple criteria and the weight corresponding to each criteria; the comments include the score corresponding to each criteria; the evaluation module 4 can also determine the evaluation of the answer text according to the evaluation criteria in the following way: weight the score corresponding to each criteria according to the weight corresponding to each criteria to obtain the target score; determine the target score as the evaluation of the answer text.

[0093] In some embodiments, the acquisition module 1 can obtain the answer text corresponding to the interview question in the following manner: receiving the interviewer's answer audio; converting the answer audio into alternative text; correcting the alternative text to obtain the answer text.

[0094] Embodiment 3

[0095] Combination Figure 3 As shown, an embodiment of the present application provides an electronic device, including a processor 5 and a memory 7. Optionally, the device may also include a communication interface 8 and a bus 6. The processor 5, the communication interface 8, and the memory 7 may communicate with each other through the bus 6. The communication interface 8 may be used for information transmission. The processor 5 may call the logic instructions in the memory 7 to execute the interview answer evaluation method of the above embodiment.

[0096] In addition, the logic instructions in the above-mentioned memory 7 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product.

[0097] The memory 7 is a computer-readable storage medium that can be used to store software programs and computer executable programs, such as program instructions / modules corresponding to the method in the embodiment of the present application. The processor 5 executes the function application and data processing by running the program instructions / modules stored in the memory 7, that is, the evaluation method of the interview answer in the above embodiment is realized.

[0098] The memory 7 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function; the data storage area may store data created according to the use of the terminal device, etc. In addition, the memory 7 may include a high-speed random access memory and may also include a non-volatile memory.

[0099] The electronic device may be a computer or a server, etc.

[0100] An embodiment of the present application provides a storage medium storing computer executable instructions, wherein the computer executable instructions are configured to execute the above-mentioned interview answer evaluation method.

[0101] An embodiment of the present application provides a computer program product, which includes a computer program stored on a storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the above-mentioned interview answer evaluation method.

[0102] The computer-readable storage medium mentioned above may be a transient computer-readable storage medium or a non-transitory computer-readable storage medium.

[0103] The technical solution of the embodiment of the present application can be embodied in the form of a software product, which is stored in a storage medium and includes one or more instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in the embodiment of the present application. The aforementioned storage medium can be a non-transient storage medium, including: a USB flash drive, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk, etc., which can store program codes, or a transient storage medium.

[0104] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments indicated above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0105] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application. At the same time, the above embodiments can be combined with each other to form new embodiments without conflict.

Claims

1. A method for evaluating interview answers, characterized in that: include: Obtaining interview questions and answer texts corresponding to the interview questions; Determine a scoring criterion model according to the type of the interview question; Generating evaluation criteria for the interview questions using the scoring criteria model; An evaluation of the answer text is determined according to the evaluation rules.

2. The method according to claim 1, characterized in that The question types of the interview questions are determined by: The interview question and the preset first prompt word are input into a preset first large language model to obtain a question type corresponding to the interview question; the first prompt word is used to indicate the question type of the content input into the first large language model determined by the first large language model with reference to the preset type and the definition of the preset type.

3. The method according to claim 1, characterized in that Determine the scoring criteria model based on the type of interview question, including: A table lookup operation is performed on the question type by using a preset model database to obtain a scoring rule model corresponding to the question type; the model database stores the corresponding relationship between the question type and the scoring rule model.

4. The method according to claim 1, characterized in that Generating evaluation criteria for the interview questions using the scoring criteria model includes: Obtaining a second prompt word corresponding to the scoring rule model; The interview question and the second prompt word are input into the scoring criteria model to obtain evaluation criteria corresponding to the interview question; the second prompt word is used to indicate the evaluation criteria of the content input into the scoring criteria model determined by the scoring criteria model with reference to the preset scoring dimensions.

5. The method according to claim 1, characterized in that Determining the evaluation of the answer text according to the evaluation rules includes: The answer text, the evaluation criteria and a preset third prompt word are input into a preset second largest language model to obtain a comment corresponding to the answer text; the third prompt word is used to indicate the comment of the answer text input into the second largest language model determined by the second largest language model with reference to the evaluation criteria.

6. The method according to claim 5, characterized in that The evaluation rules include multiple rules and weights corresponding to each rule; the comments include scores corresponding to each rule; Determining the evaluation of the answer text according to the evaluation rules also includes: Weight the scores corresponding to the above-mentioned rules according to the weights corresponding to the above-mentioned rules to obtain the target score; The target score is determined as an evaluation of the answer text.

7. The method according to any one of claims 1 to 6, characterized in that: The answer text corresponding to the interview question is obtained in the following way: Receive the interviewee's answer audio; Converting the answer audio into an alternative text; Correct the alternative texts to obtain the answer text.

8. An interview answer evaluation device, characterized in that: include: An acquisition module, used to acquire interview questions and answer texts corresponding to the interview questions; A model determination module, used to determine a scoring rule model according to the type of the interview question; A criterion determination module, used to generate evaluation criteria for the interview question using the scoring criterion model; An evaluation module is used to determine the evaluation of the answer text according to the evaluation rules.

9. An electronic device, characterized in that: The invention comprises a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the interview answer evaluation method according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores computer executable instructions, and when the computer executable instructions are called and executed by the processor, the computer executable instructions enable the processor to implement the interview answer evaluation method described in any one of claims 1 to 7.

Images