Interview method, device, computer equipment and storage medium based on artificial intelligence

Through the artificial intelligence interview method, the nine-layer architectural model is used to process interview data, and the problems of low intelligent interview efficiency and poor data versatility are solved, achieving the accuracy and efficiency of interview results.

CN111694937BActive Publication Date: 2024-11-05PING AN TECH (SHENZHEN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010340345.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-04-26
Publication Date
2024-11-05
Estimated Expiration
2040-04-26

AI Technical Summary

Technical Problem

The existing intelligent interview technology is inefficient, the correlation of each model is not strong, the data is poor, and it is difficult to process multiple interview data dimensions at the same time, resulting in inaccurate interview results.

Method used

Using an interview method based on artificial intelligence, semantic recognition and classification by obtaining response text, using a nine-layer architecture intelligent interview model for text classification, named entity recognition and open interview questions generation, combining instant response scores and authenticity scores Determine the interview results and achieve the unity and integrity of multi-task data processing.

Benefits of technology

It improves the efficiency of intelligent interviews and the accuracy of interview results, realizes the generation of intelligent interview questions and interview guidance, and ensures the unity and completeness of data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111694937B_ABST
    Figure CN111694937B_ABST
Patent Text Reader

Abstract

The present invention discloses an interview method, device, computer equipment and storage medium based on artificial intelligence, including: obtaining a response text, semantically identifying and classifying the response text according to preset conditions, obtaining first data, second data and third data, inputting the first data, second data and third data into a preset intelligent interview model for training, obtaining response text classification, named entities and open-ended interview questions, realizing the use of one model to simultaneously perform processing tasks for multiple interview data, ensuring the uniformity and integrity of data, improving the efficiency of intelligent interviews and the accuracy of interview results, and at the same time, determining an instant response score according to the response text classification, determining an authenticity score of the response according to the named entity, continuing to guide the intelligent interview according to the open-ended interview questions, determining an interview result according to the instant response score and the authenticity score finally obtained, and storing the interview result on a blockchain, thereby improving the interview efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to an interview method, device, computer equipment and storage medium based on artificial intelligence. Background Art

[0002] With the rapid development of computer technology, the application of artificial intelligence is becoming more and more extensive. In some large enterprises, it is necessary to select suitable personnel from a large number of interviewees. The use of intelligent interviews has become a good choice. In the intelligent interview scenario, there are many dimensions involved in evaluating the interview results. For example, it is necessary to judge the interviewee's response ability based on the semantic understanding of the interview answer text, and it is necessary to verify the consistency of the description by obtaining the entity name from the answer. At the same time, it is also necessary to interact with the interviewee by intelligently generating questions based on the answers. In the existing technology, each dimension needs to be implemented in a specific way or model, and the correlation between different models is not strong, and the data universality is poor, which leads to low efficiency of intelligent interviews. Summary of the invention

[0003] The embodiments of the present invention provide an interview method, apparatus, computer equipment and storage medium based on artificial intelligence to improve the efficiency of intelligent interviews.

[0004] An interview method based on artificial intelligence, comprising:

[0005] Obtaining a response text, wherein the response text includes the interviewer's resume text and the answer text obtained by the interviewer answering the interview questions;

[0006] According to preset conditions, semantic recognition and classification are performed on the response text to obtain first data, second data and third data, wherein the first data is text classification task data, the second data is named entity recognition task data, and the third data is text generation task data;

[0007] Inputting the first data, the second data and the third data into a preset intelligent interview model to obtain answer text classification, named entity and open interview questions respectively, wherein the intelligent interview model includes a nine-layer architecture, the first layer is a data input layer, the second layer is a data encoding layer, the third to seventh layers are transformer encoding layers, the eighth layer is a fully connected layer, and the ninth layer is a softmax output layer;

[0008] Determine an instant answer score according to the answer text classification, determine an answer authenticity score according to the named entity, continue to guide the intelligent interview according to the open-ended interview question, and return to the step of obtaining the answer text to continue executing until the generated open-ended interview question is empty;

[0009] The interview result is determined based on the instant response score and authenticity score obtained.

[0010] Optionally, before inputting the first data, the second data and the third data into a preset intelligent interview model to obtain answer text classification, named entities and open-ended interview questions respectively, the artificial intelligence-based interview method further includes:

[0011] Input the sample data into the initial intelligent interview model according to a preset input method, and output the corresponding output result, wherein the sample data is the historical answer text in the interview database;

[0012] Determine a loss function according to the output result;

[0013] Calculate the loss value according to the determined loss function;

[0014] If the loss value exceeds the preset threshold range, iterative fitting is performed according to the loss value, the parameters of the transformer encoding layer are updated, and the step of inputting the sample data into the initial intelligent interview model according to the preset input method and outputting the corresponding output result is returned to continue execution;

[0015] If the loss value does not exceed the preset threshold range, the obtained intelligent interview model will be used as the preset intelligent interview model.

[0016] Optionally, the step of inputting the sample data into the initial intelligent interview model according to a preset input method and outputting the corresponding output result includes:

[0017] Parsing the first data into a character M+word sequence, and inputting the character M into the initial intelligent interview model for training, and using the output result of the character M as the first output result, wherein the character M is any predefined character;

[0018] Parsing the second data into a character M+word sequence, and inputting the result into the initial intelligent interview model for training, and taking the output result of each character other than the character M as the second output result;

[0019] Parsing the third data into a character M+first dialogue sequence+second dialogue sequence, and inputting the result into the initial intelligent interview model for training, and using the output result of the second dialogue sequence as the third output result, wherein the first dialogue sequence is a dialogue sequence corresponding to the interview question, and the second dialogue sequence is a dialogue sequence corresponding to the interview answer;

[0020] Randomly select a segmentation from the first data and the second data as a target segmentation, and use a preset masking symbol to block the target segmentation, and use the initial intelligent interview model to train the masked first data and second data to obtain a fourth output result;

[0021] The first output result, the second output result, the third output result and the fourth output result are aggregated to obtain the corresponding output result.

[0022] Optionally, the parsing of the third data into a method of character M+first dialogue sequence+second dialogue sequence, and inputting the method into the initial intelligent interview model for training, and taking the output result of the method of the second dialogue sequence as the corresponding output result includes:

[0023] When the initial intelligent interview model is training the first dialogue sequence, an occlusion matrix is ​​used to perform occlusion processing on the second dialogue sequence, and then the first dialogue sequence is trained;

[0024] When the initial intelligent interview model is training the second dialogue sequence, for each character in the second dialogue sequence, an occlusion matrix is ​​used to perform occlusion processing on characters following the character, and then the character is trained.

[0025] Optionally, determining a loss function according to the output result includes:

[0026] By means of cross entropy calculation, respectively calculate the loss values ​​corresponding to the first output result, the second output result, the third output result, and the fourth output result to obtain a first loss value, a second loss value, a third loss value, and a fourth loss value;

[0027] An average of the first loss value, the second loss value, the third loss value, and the fourth loss value is calculated, and the loss function is determined according to the average value.

[0028] An interview device based on artificial intelligence, comprising:

[0029] A data acquisition module, used to acquire a response text, wherein the response text includes the interviewer's resume text and the answer text obtained by the interviewer answering the interview questions;

[0030] A data classification module, used to perform semantic recognition and classification on the response text according to preset conditions to obtain first data, second data and third data, wherein the first data is text classification task data, the second data is named entity recognition task data, and the third data is text generation task data;

[0031] A data training module, used for inputting the first data, the second data and the third data into a preset intelligent interview model to obtain answer text classification, named entities and open interview questions respectively, wherein the intelligent interview model includes a nine-layer architecture, the first layer is a data input layer, the second layer is a data encoding layer, the third to seventh layers are transformer encoding layers, the eighth layer is a fully connected layer, and the ninth layer is a softmax output layer;

[0032] A loop iteration module, used to determine the instant answer score according to the answer text classification, determine the answer authenticity score according to the named entity, continue to guide the intelligent interview according to the open-ended interview question, and return to the step of obtaining the answer text to continue execution until the generated open-ended interview question is empty;

[0033] The result determination module is used to determine the interview result based on the obtained instant answer score and authenticity score.

[0034] Optionally, the artificial intelligence-based interview device further includes:

[0035] A data processing module, used to input sample data into the initial intelligent interview model according to a preset input method, and output corresponding output results, wherein the sample data is the historical answer text in the interview database;

[0036] A function determination module, used to determine a loss function according to the output result;

[0037] The loss calculation module is used to calculate the loss value according to the determined loss function;

[0038] An iterative fitting module, for iteratively fitting according to the loss value if the loss value exceeds a preset threshold range, updating the parameters of the transformer coding layer, and returning to the step of inputting the sample data into the initial intelligent interview model according to a preset input method and outputting the corresponding output result to continue to execute;

[0039] The model determination module is used to use the obtained intelligent interview model as the preset intelligent interview model if the loss value does not exceed the preset threshold range.

[0040] Optionally, the data processing module includes:

[0041] A first processing unit is used to parse the first data into a character M+word sequence, and input the character M into the initial intelligent interview model for training, and use the output result of the character M as the first output result, wherein the character M is any predefined character;

[0042] A second processing unit is used to parse the second data into a character M+word sequence, and input the result into the initial intelligent interview model for training, and use the output result of each character other than the character M as the second output result;

[0043] A third processing unit is used to parse the third data into a character M+first dialogue sequence+second dialogue sequence, and input the third data into the initial intelligent interview model for training, and use the output result of the second dialogue sequence as the third output result, wherein the first dialogue sequence is a dialogue sequence corresponding to the interview question, and the second dialogue sequence is a dialogue sequence corresponding to the interview answer;

[0044] a fourth processing unit, configured to randomly select a segmentation from the first data and the second data as a target segmentation, block the target segmentation with a preset masking symbol, and use the initial intelligent interview model to train the masked first data and second data to obtain a fourth output result;

[0045] The result determination unit is used to summarize the first output result, the second output result, the third output result and the fourth output result to obtain the corresponding output result.

[0046] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned artificial intelligence-based interview method when executing the computer program.

[0047] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned artificial intelligence-based interview method.

[0048] The artificial intelligence-based interview method, device, computer equipment and storage medium provided by the embodiments of the present invention obtain a response text, and semantically identify and classify the response text according to preset conditions to obtain first data, second data and third data, wherein the first data is text classification task data, the second data is named entity recognition task data, and the third data is text generation task data. The first data, the second data and the third data are then input into a preset intelligent interview model to obtain response text classification, named entities and open-ended interview questions, respectively, so as to realize the processing tasks of multiple interview data at the same time with one model, ensure the uniformity and integrity of the data, improve the efficiency of the intelligent interview and the accuracy of the interview results, and at the same time, determine the instant response score according to the response text classification, determine the authenticity score of the response according to the named entity, continue to guide the intelligent interview according to the open-ended interview question, and return to the step of obtaining the response text to continue iterative execution until the generated open-ended interview question is empty and stops, and finally determine the interview result according to the obtained instant response score and authenticity score, realize the generation of intelligent interview questions and interview guidance, and improve the interview efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0050] Figure 1 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0051] Figure 2 is a flowchart of an implementation of an interview method based on artificial intelligence provided by an embodiment of the present invention;

[0052] Figure 3 is a schematic diagram of an interview device based on artificial intelligence provided by an embodiment of the present invention;

[0053] Figure 4 It is a schematic diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0055] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] See also Figure 1 ,like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0058] Users can use terminal devices 101 , 102 , 103 to interact with server 105 via network 104 to receive or send messages, etc.

[0059] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture E interface display perts Group Audio Layer III, Moving Picture Experts Compression Standard Audio Layer 3), MP4 (Moving Picture E interface display perts Group Audio Layer IV, Moving Picture Experts Compression Standard Audio Layer 4) players, laptop computers and desktop computers, etc.

[0060] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0061] It should be noted that the electronic file control method provided in the embodiment of the present application is executed by a server, and accordingly, the electronic file control device is set in the server.

[0062] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. According to the implementation requirements, there may be any number of terminal devices, networks and servers. The terminal devices 101, 102, 103 in the embodiment of the present application may specifically correspond to application systems in actual production.

[0063] See also Figure 2 , Figure 2 An artificial intelligence-based interview method provided by an embodiment of the present invention is shown. The method is applied in Figure 1 The server in is used as an example to illustrate the details as follows:

[0064] S10: Obtaining a response text, wherein the response text includes the interviewer's resume text and the answer text obtained by the interviewer answering the interview questions.

[0065] Specifically, the interviewer's response text is obtained, and the response text includes the interviewer's resume text and the answer text obtained by the interviewer answering the interview questions.

[0066] It should be understood that the interviewee can answer the interview questions by voice, obtain a recording text, and then convert the recording text into an answer text, or answer the questions by paper or electronic test questions to obtain an answer text.

[0067] S20: According to preset conditions, semantic recognition and classification are performed on the response text to obtain first data, second data and third data, wherein the first data is text classification task data, the second data is named entity recognition task data, and the third data is text generation task data.

[0068] Among them, the preset conditions refer to classification conditions set according to the needs of multiple tasks. For example, in an intelligent interview task, one of the tasks is to judge the interviewee's skill mastery through named entity recognition. For this task, the set condition is to select sentences containing nouns and relative words in the answer text as task data. The specific conditions can be set according to the type of data actually needed, and there is no restriction here.

[0069] In this embodiment, as a preferred method, the first data is the data required for the text classification task, the second data is the data required for the named entity recognition task, and the third data is the data required for the text generation task.

[0070] Among them, semantic recognition adopts the method of natural language semantic recognition, mainly by segmenting the response sentence, converting the segmented words into word vectors, clustering the word vectors, and then determining the semantics corresponding to the response sentence based on the obtained cluster center.

[0071] Among them, Natural Language Processing (NLP) is a subfield of Artificial Intelligence (AI). It uses machine learning to understand and parse natural language, thereby solving some problems in the field of natural language. The main application scope of NLP includes but is not limited to: text reading (Text to speech) / speech synthesis (Speech synthesis), speech recognition (Speech recognition), Chinese automatic word segmentation (Chinese words segmentation), part-of-speech tagging (Part-of-speech tagging), syntactic analysis (Parsing), text classification (Text categorization), information retrieval (Information retrieval), automatic summarization (Automatic summarization) and text proofing (Text-proofing), etc.

[0072] Among them, the text classification task refers to the extraction and classification of sentences in the answer text, the named entity recognition task refers to the recognition of named entities in the sentences in the answer text, and the text generation task refers to the intelligent generation of the next question sentence based on the previous round or multiple rounds of questions and answers. In this embodiment, the text classification task is used to identify and extract the interviewee's experience or skills, mentality and other related content, which is used to evaluate the interviewee's comprehensive ability. The named entity recognition task can determine the consistency and accuracy of the interviewee's previous and subsequent answers and resume, and is used to identify the authenticity of the materials provided by the interviewee. The text generation task is to generate additional questions based on the interviewee's existing answers when some content is uncertain, so as to better understand the interviewee.

[0073] It should be noted that the text generation task in this embodiment can further generate interview questions for named entity recognition.

[0074] In a specific implementation, the text classification task is to classify the answer content of different skills involved in the answer according to the skill category in the resume, and after classification, score it through a subsequent intelligent interview model.

[0075] It should be understood that the same response text may belong to one task data or to two or more task data. For example, a response text is "I won the first prize in spatial modeling in the 2015 MapGIS competition". After semantic recognition, it is confirmed that the response text belongs to both the data required for named entity recognition and the data belonging to the text classification task.

[0076] In a specific embodiment, the intelligent interview terminal is preset with an interview question bank, which includes several similar questions, and each similar question is marked with the question type corresponding to the similar question, for example, identity information, experience, skills, salary conditions, values, etc. The question type has a mapping relationship with the task data classification. According to semantic recognition, the similar questions corresponding to the response statement are judged, and then the task data classification to which the response statement belongs is determined according to the mapping relationship between the question type corresponding to the similar questions and the task data. At the same time, in the process of semantic recognition, when it is recognized that the response statement contains naming recognition, the response statement is also used as the data of the second data. In the process of semantic recognition, when it is detected that the semantics of the response statement does not meet the preset expectations (it is incomplete, or contains too little information), the response statement is used as the third data, and new interview questions are generated according to the similar questions and the third data corresponding to the response statement.

[0077] S30: Input the first data, the second data and the third data into a preset intelligent interview model to obtain answer text classification, named entities and open-ended interview questions respectively, wherein the intelligent interview model includes a nine-layer architecture, the first layer is a data input layer, the second layer is a data encoding layer, the third layer to the seventh layer are transformer encoding layers, the eighth layer is a fully connected layer, and the ninth layer is a softmax output layer.

[0078] Specifically, this embodiment inputs the first data, the second data, and the third data obtained in step S20 into a preset intelligent interview model, and performs text classification processing on the first data through the intelligent interview model to obtain a response text classification, so that the interviewee's instant response score during the interview can be evaluated through the response text classification later, and the second data is subjected to named entity recognition through the intelligent interview model. It is necessary to extract the named entities based on the candidate's answer to compare whether it is consistent with the description in his resume. For example, when asking about his past work experience, the name of the company he has served can be identified from the candidate's answer to verify the authenticity of the response text provided by the interviewee, and open questions for interacting with the interviewer are generated through the third data.

[0079] Preferably, this embodiment uses five transformer encoding layers, and performs multiple weight calculations and weightings on word segmentation to improve the accuracy of data parsing for each task.

[0080] Among them, the transformer layer in this embodiment is a neural network with an Encoder-Decoder structure, where Encoder is the encoder and Decoder is the decoder. In order to ensure the semantic coherence of the generated poetry, in this embodiment, position encoding (Position Embedding) is added to the encoder, and the position information of the word is represented by position encoding.

[0081] Among them, the preset intelligent interview model refers to a model that is pre-trained with sample data and can simultaneously execute the three tasks mentioned in step S20. The specific training process can refer to the description of subsequent embodiments. To avoid repetition, it will not be repeated here.

[0082] It is worth noting that, in general intelligent interview models, one or two transformer layers are generally used to train text data to obtain corresponding results, while this embodiment uses multi-layer transformers. By pre-training the intelligent interview model, transformer parameters suitable for three tasks are obtained, so that the three tasks can be processed simultaneously with the same model. The conversion and use efficiency between data is higher, which is conducive to improving the efficiency of processing the response text.

[0083] Among them, the softmax output layer is used to output the results obtained after model training.

[0084] S40: Determine the instant answer score based on the answer text classification, determine the answer authenticity score based on the named entity, continue to guide the intelligent interview based on the open-ended interview questions, and return to the step of obtaining the answer text to continue execution until the generated open-ended interview questions are empty.

[0085] Specifically, the instant answer score is determined according to the category to which the obtained answer text belongs, the authenticity score of the answer is determined according to the consistency of the identified named entities, and the intelligent interview is continued to be guided according to the open-ended interview questions. If the generated open-ended interview questions are empty, step S50 is executed; otherwise, the interviewee is guided to answer the generated open-ended interview, and the obtained answer text is executed in a loop according to steps S20 to S40 until the generated open-ended interview questions are empty.

[0086] Among them, the instant response score refers to the evaluation value obtained based on the interviewee's response, the authenticity score refers to the authenticity of the information in the interviewee's resume and interview responses, and the open-ended interview questions refer to judging what information about the interviewee that needs to be known based on the previous questions and the interviewee's responses, and then generating corresponding interview questions.

[0087] S50: Determine the interview result based on the obtained instant response score and authenticity score.

[0088] Specifically, the interview result is determined by the obtained instant response score and authenticity score.

[0089] In one embodiment, after the interview result is determined, the interview result is stored in the blockchain network node, including the interviewer's information, the answer text and the final interview result. Through blockchain storage, data information can be shared between different platforms, and data can also be prevented from being tampered with.

[0090] Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0091] In this embodiment, by obtaining the answer text, semantic recognition and classification are performed on the answer text according to preset conditions to obtain first data, second data and third data, wherein the first data is text classification task data, the second data is named entity recognition task data, and the third data is text generation task data. The first data, the second data and the third data are input into the preset intelligent interview model to obtain answer text classification, named entities and open-ended interview questions respectively, so as to realize the processing tasks of multiple interview data at the same time with one model, ensure the uniformity and integrity of the data, improve the efficiency of the intelligent interview and the accuracy of the interview results, and at the same time, determine the instant answer score according to the answer text classification, determine the authenticity score of the answer according to the named entity, continue to guide the intelligent interview according to the open-ended interview questions, and return to the step of obtaining the answer text to continue iterative execution until the generated open-ended interview questions are empty and stop, and finally determine the interview result according to the obtained instant answer score and authenticity score, realize the generation of intelligent interview questions and interview guidance, and improve the interview efficiency.

[0092] In some optional implementations of this embodiment, before step S30, the artificial intelligence-based interview method further includes a training process of a preset intelligent interview model, which is described in detail as follows:

[0093] Input the sample data into the initial intelligent interview model according to the preset input method, and output the corresponding output result, wherein the sample data is the historical answer text in the interview database;

[0094] According to the output results, determine the loss function;

[0095] Calculate the loss value according to the determined loss function;

[0096] If the loss value exceeds the preset threshold range, iterative fitting is performed according to the loss value, the parameters of the transformer encoding layer are updated, and the sample data is input into the initial intelligent interview model according to the preset input method, and the step of outputting the corresponding output result is continued;

[0097] If the loss value does not exceed the preset threshold range, the obtained intelligent interview model will be used as the preset intelligent interview model.

[0098] Specifically, the server pre-trains the initial intelligent interview model to obtain a preset intelligent interview model. The training is to continuously adjust the parameters of the transformer layer by performing multiple feature extraction and verification on the sample data, so that the final intelligent interview model can parse the interview task data with an accuracy rate exceeding the preset threshold range.

[0099] Among them, the sample data is the historical answer text in the interview database.

[0100] The output result is the semantic information corresponding to the input sample data, and the semantic information is feature information that can be understood by the computer model and can be used for subsequent semantic analysis.

[0101] Among them, the initial intelligent interview model is a general transformer model, whose parameters are initialized random parameters. By using sample data related to the interview scenario, the finally trained model can recognize questions in various interview scenarios with higher accuracy.

[0102] It should be noted that the output result includes the model training results corresponding to different task data. In this embodiment, the respective task data are used for training, and the corresponding loss functions are obtained, which are then integrated to obtain the final loss function of the entire model. The model is iteratively trained using the obtained loss function. Each training will adjust the parameters of the transformer layer according to the calculated loss by back propagation, and continue to iteratively calculate the loss after adjustment until the calculated loss is within a preset range. It is considered that the parameters of the transformer layer have been adjusted to the optimal parameters that can take into account the three tasks at the same time, and the model at this time is used as the final intelligent interview model.

[0103] It should be understood that the initial intelligent interview model in this embodiment is a transformer model, which includes an encoder, an attention mechanism and a decoder.

[0104] Among them, the attention mechanism originates from the study of human vision. In cognitive science, due to the bottleneck of information processing, humans selectively focus on part of all information and ignore other visible information.

[0105] The attention mechanism includes but is not limited to: multi-head attention, key-value pair attention and structured attention, etc. Preferably, this embodiment adopts the key-value pair attention mechanism, which represents the input information by using the key-value pair format, where the "key" is used to calculate the attention distribution and the "value" is used to generate the selected information.

[0106] In this embodiment, the data of historical interview texts is used as sample data to train the initial intelligent interview model, and the training results are compared with the actual values, and then the model is adjusted and optimized to obtain a preset intelligent interview model. This is beneficial for the subsequent execution of various tasks in the intelligent interview through the preset intelligent interview model, thereby improving the interview efficiency.

[0107] In some optional implementations of this embodiment, the sample data is input into the initial intelligent interview model according to a preset input method, and the corresponding output results are output, including:

[0108] Parse the first data into a character M+word sequence, input the character M into the initial intelligent interview model for training, and use the output result of the character M as the first output result, wherein the character M is any predefined character;

[0109] Parse the second data into a character M+word sequence, input the sequence into the initial intelligent interview model for training, and use the output result of each character other than the character M as the second output result;

[0110] Parse the third data into a character M+first dialogue sequence+second dialogue sequence, and input the result into the initial intelligent interview model for training, and use the output result of the second dialogue sequence as the third output result, wherein the first dialogue sequence is the dialogue sequence corresponding to the interview question, and the second dialogue sequence is the dialogue sequence corresponding to the interview answer;

[0111] Randomly select a segmentation from the first data and the second data as a target segmentation, and use a preset masking symbol to block the target segmentation, and use the initial intelligent interview model to train the masked first data and second data to obtain a fourth output result;

[0112] The first output result, the second output result, the third output result and the fourth output result are summarized to obtain corresponding output results.

[0113] The character M is a predefined arbitrary identifier.

[0114] The character sequence refers to a word sequence obtained by segmenting data according to the frequency of words in a preset corpus. The preset corpus can be selected according to actual needs.

[0115] In this embodiment, in combination with the characteristics of each task, the data corresponding to the task are trained in different ways to obtain the training results of the data corresponding to the task as output results. The output results are more consistent with the task requirements, which is conducive to improving the rationality and accuracy of the loss function when the loss function is determined through the output results in the future.

[0116] In some optional implementations of this embodiment, the third data is parsed into a method of character M+first dialogue sequence+second dialogue sequence, and input into the initial intelligent interview model for training, and the output result of the method of the second dialogue sequence is used as the corresponding output result, including:

[0117] When the initial intelligent interview model is training the first dialogue sequence, the occlusion matrix is ​​used to perform occlusion processing on the second dialogue sequence, and then the first dialogue sequence is trained;

[0118] When the initial intelligent interview model is training the second dialogue sequence, for each character in the second dialogue sequence, an occlusion matrix is ​​used to occlude the characters following the character, and then the characters are trained.

[0119] The mask matrix is ​​a matrix composed of 0 and 1. For example, in natural language processing (NLP), sentences are of different lengths, but because sentences are often grouped into mini-batches for training, shorter sentences are padded with 0 at the end, which is a padding operation. A mask matrix is ​​used to indicate which are real data and which are padding data.

[0120] In this embodiment, the occlusion processing of the occlusion matrix is ​​used to ensure that the semantic recognition of the question is not interfered by the response statement. At the same time, it is ensured that each character of the response statement is not interfered by the semantics of subsequent characters, which is conducive to improving the accuracy of the third data analysis and recognition.

[0121] In some optional implementations of this embodiment, determining the loss function according to the output result includes:

[0122] By means of cross entropy calculation, the loss values ​​corresponding to the first output result, the second output result, the third output result and the fourth output result are calculated respectively to obtain a first loss value, a second loss value, a third loss value and a fourth loss value;

[0123] An average of the first loss value, the second loss value, the third loss value, and the fourth loss value is calculated, and a loss function is determined according to the average value.

[0124] Specifically, the loss between the output result corresponding to each training task and the real data is calculated by cross entropy calculation to obtain the loss value corresponding to each training task. Then, the universal loss function of the model is determined according to each loss value, and the loss function is continuously optimized in subsequent iterative training, so that the loss function has better processing capabilities for each training task.

[0125] Among them, cross entropy is an important concept in Shannon information theory, which is mainly used to measure the difference information between two probability distributions. The performance of language models is usually measured by cross entropy and complexity (perplexity). The meaning of cross entropy is the difficulty of text recognition using the model, or from the perspective of compression, how many bits are used to encode each word on average. The meaning of complexity is the average number of branches that the model uses to represent this text, and its reciprocal can be regarded as the average probability of each word. Smoothing refers to assigning a probability value to the unobserved N-gram combination to ensure that the word sequence can always get a probability value through the language model.

[0126] In this embodiment, by determining the loss function suitable for each training task and continuously optimizing and adjusting it, the model is made universal, which is beneficial to improving the efficiency of intelligent interviews.

[0127] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0128] Figure 3 The principle block diagram of the interview device based on artificial intelligence is shown, which corresponds to the interview method based on artificial intelligence in the above embodiment. Figure 3 As shown, the interview device based on artificial intelligence includes a data acquisition module 31, a data classification module 32, a data training module 33, a loop iteration module 34 and a result determination module 35. The functional modules are described in detail as follows:

[0129] The data acquisition module 31 is used to acquire the response text, wherein the response text includes the interviewer's resume text and the answer text obtained by the interviewer answering the interview questions;

[0130] The data classification module 32 is used to perform semantic recognition and classification on the response text according to preset conditions to obtain first data, second data and third data, wherein the first data is text classification task data, the second data is named entity recognition task data, and the third data is text generation task data;

[0131] A data training module 33 is used to input the first data, the second data and the third data into a preset intelligent interview model to obtain answer text classification, named entities and open interview questions respectively, wherein the intelligent interview model includes a nine-layer architecture, the first layer is a data input layer, the second layer is a data encoding layer, the third layer to the seventh layer are all transformer encoding layers, the eighth layer is a fully connected layer, and the ninth layer is a softmax output layer;

[0132] A loop iteration module 34 is used to determine the instant answer score according to the answer text classification, determine the answer authenticity score according to the named entity, continue to guide the intelligent interview according to the open-ended interview question, and return to the step of obtaining the answer text to continue execution until the generated open-ended interview question is empty;

[0133] The result determination module 35 is used to determine the interview result according to the obtained instant answer score and authenticity score.

[0134] Optionally, the artificial intelligence-based interview device further includes:

[0135] A data processing module is used to input sample data into the initial intelligent interview model according to a preset input method and output corresponding output results, wherein the sample data is the historical answer text in the interview database;

[0136] A function determination module is used to determine the loss function according to the output result;

[0137] The loss calculation module is used to calculate the loss value according to the determined loss function;

[0138] The iterative fitting module is used to iteratively fit according to the loss value if the loss value exceeds the preset threshold range, update the parameters of the transformer encoding layer, and return to input the sample data into the initial intelligent interview model according to the preset input method, and output the corresponding output result step to continue execution;

[0139] The model determination module is used to use the obtained intelligent interview model as the preset intelligent interview model if the loss value does not exceed the preset threshold range.

[0140] Optionally, the data processing module includes:

[0141] A first processing unit is used to parse the first data into a character M+word sequence, and input the first data into an initial intelligent interview model for training, and use an output result of the character M as a first output result, wherein the character M is any predefined character;

[0142] A second processing unit is used to parse the second data into a character M+word sequence, and input the second data into the initial intelligent interview model for training, and use the output result of each character other than the character M as the second output result;

[0143] A third processing unit is used to parse the third data into a character M+first dialogue sequence+second dialogue sequence, and input the third data into the initial intelligent interview model for training, and use the output result of the second dialogue sequence as the third output result, wherein the first dialogue sequence is a dialogue sequence corresponding to the interview question, and the second dialogue sequence is a dialogue sequence corresponding to the interview answer;

[0144] The fourth processing unit is used to randomly select a segmentation from the first data and the second data as a target segmentation, and block the target segmentation with a preset masking symbol, and use the initial intelligent interview model to train the masked first data and second data to obtain a fourth output result;

[0145] The result determination unit is used to summarize the first output result, the second output result, the third output result and the fourth output result to obtain a corresponding output result.

[0146] Optionally, the third processing unit includes:

[0147] A first occlusion subunit is used to use an occlusion matrix to perform occlusion processing on the second dialogue sequence when the initial intelligent interview model trains the first dialogue sequence, and then train the first dialogue sequence;

[0148] The second occlusion subunit is used to, when the initial intelligent interview model is training the second dialogue sequence, use an occlusion matrix for each character in the second dialogue sequence, perform occlusion processing on the characters following the character, and then train the characters.

[0149] Optionally, the function determination module includes:

[0150] A loss value calculation unit, used to calculate the loss values ​​corresponding to the first output result, the second output result, the third output result and the fourth output result respectively by cross entropy calculation to obtain a first loss value, a second loss value, a third loss value and a fourth loss value;

[0151] The loss function generating unit is used to calculate the average of the first loss value, the second loss value, the third loss value and the fourth loss value, and determine the loss function according to the average.

[0152] For the specific definition of the interview device based on artificial intelligence, please refer to the definition of the interview method based on artificial intelligence in the above text, which will not be repeated here. Each module in the above-mentioned interview device based on artificial intelligence can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0153] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0154] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components connected to the memory 41, the processor 42, and the network interface 43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0155] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0156] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or D interface display memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as the program code for controlling electronic files, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0157] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the program code stored in the memory 41 or process data, such as running the program code for controlling the electronic file.

[0158] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0159] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores an interface display program, and the interface display program can be executed by at least one processor to enable the at least one processor to perform the steps of the electronic file control method as described above.

[0160] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0161] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. An interview method based on artificial intelligence, characterized in that: The AI-based interview method includes: Obtaining a response text, wherein the response text includes the interviewer's resume text and the answer text obtained by the interviewer answering the interview questions; According to preset conditions, semantic recognition and classification are performed on the response text to obtain first data, second data and third data, wherein the first data is text classification task data, the second data is named entity recognition task data, and the third data is text generation task data; Inputting the first data, the second data and the third data into a preset intelligent interview model to obtain answer text classification, named entity and open interview questions respectively, wherein the intelligent interview model includes a nine-layer architecture, the first layer is a data input layer, the second layer is a data encoding layer, the third to seventh layers are transformer encoding layers, the eighth layer is a fully connected layer, and the ninth layer is a softmax output layer; Determine an instant answer score according to the answer text classification, determine an answer authenticity score according to the named entity, continue to guide the intelligent interview according to the open-ended interview question, and return to the step of obtaining the answer text to continue executing until the generated open-ended interview question is empty; The interview result is determined based on the instant response score and authenticity score obtained.

2. The artificial intelligence-based interview method according to claim 1, characterized in that: Before inputting the first data, the second data and the third data into a preset intelligent interview model to obtain answer text classification, named entities and open-ended interview questions respectively, the artificial intelligence-based interview method further includes: Input the sample data into the initial intelligent interview model according to a preset input method, and output the corresponding output result, wherein the sample data is the historical answer text in the interview database; Determine a loss function according to the output result; Calculate the loss value according to the determined loss function; If the loss value exceeds the preset threshold range, iterative fitting is performed according to the loss value, the parameters of the transformer encoding layer are updated, and the step of inputting the sample data into the initial intelligent interview model according to the preset input method and outputting the corresponding output result is returned to continue execution; If the loss value does not exceed the preset threshold range, the obtained intelligent interview model will be used as the preset intelligent interview model.

3. The artificial intelligence-based interview method according to claim 2, characterized in that: The step of inputting the sample data into the initial intelligent interview model according to a preset input method and outputting the corresponding output results includes: Parsing the first data into a character M+word sequence, and inputting the character M into the initial intelligent interview model for training, and using the output result of the character M as the first output result, wherein the character M is any predefined character; Parsing the second data into a character M+word sequence, and inputting the result into the initial intelligent interview model for training, and taking the output result of each character other than the character M as the second output result; Parsing the third data into a character M+first dialogue sequence+second dialogue sequence, and inputting the result into the initial intelligent interview model for training, and using the output result of the second dialogue sequence as the third output result, wherein the first dialogue sequence is a dialogue sequence corresponding to the interview question, and the second dialogue sequence is a dialogue sequence corresponding to the interview answer; Randomly select a segmentation from the first data and the second data as a target segmentation, and use a preset masking symbol to block the target segmentation, and use the initial intelligent interview model to train the masked first data and second data to obtain a fourth output result; The first output result, the second output result, the third output result and the fourth output result are aggregated to obtain the corresponding output result.

4. The artificial intelligence-based interview method according to claim 3, characterized in that: The method of parsing the third data into character M+first dialogue sequence+second dialogue sequence, and inputting the method into the initial intelligent interview model for training, and using the output result of the second dialogue sequence as the corresponding output result includes: When the initial intelligent interview model is training the first dialogue sequence, an occlusion matrix is ​​used to perform occlusion processing on the second dialogue sequence, and then the first dialogue sequence is trained; When the initial intelligent interview model is training the second dialogue sequence, for each character in the second dialogue sequence, an occlusion matrix is ​​used to perform occlusion processing on characters following the character, and then the character is trained.

5. The artificial intelligence-based interview method according to claim 3, characterized in that: Determining the loss function according to the output result includes: By means of cross entropy calculation, respectively calculate the loss values ​​corresponding to the first output result, the second output result, the third output result, and the fourth output result to obtain a first loss value, a second loss value, a third loss value, and a fourth loss value; An average of the first loss value, the second loss value, the third loss value, and the fourth loss value is calculated, and the loss function is determined according to the average value.

6. The artificial intelligence-based interview method according to claim 1, characterized in that: After determining the interview result based on the obtained instant answer score and authenticity score, it also includes: storing the interview result in a blockchain network node.

7. An interview device based on artificial intelligence, characterized in that: The interview device based on artificial intelligence comprises: A data acquisition module, used to acquire a response text, wherein the response text includes the interviewer's resume text and the answer text obtained by the interviewer answering the interview questions; A data classification module, used to perform semantic recognition and classification on the response text according to preset conditions to obtain first data, second data and third data, wherein the first data is text classification task data, the second data is named entity recognition task data, and the third data is text generation task data; A data training module, used for inputting the first data, the second data and the third data into a preset intelligent interview model to obtain answer text classification, named entities and open interview questions respectively, wherein the intelligent interview model includes a nine-layer architecture, the first layer is a data input layer, the second layer is a data encoding layer, the third to seventh layers are transformer encoding layers, the eighth layer is a fully connected layer, and the ninth layer is a softmax output layer; A loop iteration module, used to determine the instant answer score according to the answer text classification, determine the answer authenticity score according to the named entity, continue to guide the intelligent interview according to the open-ended interview question, and return to the step of obtaining the answer text to continue execution until the generated open-ended interview question is empty; The result determination module is used to determine the interview result based on the obtained instant answer score and authenticity score.

8. The artificial intelligence-based interview device according to claim 7, characterized in that: The artificial intelligence-based interview device also includes: A data processing module, used to input sample data into the initial intelligent interview model according to a preset input method, and output corresponding output results, wherein the sample data is the historical answer text in the interview database; A function determination module, used to determine a loss function according to the output result; The loss calculation module is used to calculate the loss value according to the determined loss function; An iterative fitting module, for iteratively fitting according to the loss value if the loss value exceeds a preset threshold range, updating the parameters of the transformer coding layer, and returning to the step of inputting the sample data into the initial intelligent interview model according to a preset input method and outputting the corresponding output result to continue to execute; The model determination module is used to use the obtained intelligent interview model as the preset intelligent interview model if the loss value does not exceed the preset threshold range.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the artificial intelligence-based interview method as described in any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the artificial intelligence-based interview method as described in any one of claims 1 to 6 is implemented.