Search

CN-121212135-B - AI interview method, electronic device, storage medium, and program product

CN121212135BCN 121212135 BCN121212135 BCN 121212135BCN-121212135-B

Abstract

The application relates to the technical field of artificial intelligence and discloses an AI interview method, electronic equipment, a storage medium and a program product, wherein the method comprises the steps of obtaining interview text; the method comprises the steps of determining a first speech speed and a first pause frequency of a candidate according to an interview text, wherein the first pause frequency is used for describing pause times of the candidate in the process of inputting interview audio corresponding to the interview text, determining semantic integrity probability of the interview text, determining probability that the candidate does not continue speaking according to the first speech speed, the first pause frequency and the semantic integrity probability, and determining whether to continue asking questions according to the probability that the candidate does not continue speaking. Therefore, the timing of the continuous question of the AI interviewee can be adaptively adjusted according to the speaking style of the candidate, so that the AI interviewee takes over the dialogue at a reasonable timing, and the interview experience of the candidate is improved.

Inventors

  • LI ZHONGWEI

Assignees

  • 北京牛客科技有限公司

Dates

Publication Date
20260505
Application Date
20250918

Claims (14)

  1. 1. A method of AI interviewing, comprising: Acquiring an interview text; Determining a first speech speed and a first pause frequency of the candidate according to the interview text, wherein the first pause frequency is used for describing pause times of the candidate in the process of inputting interview audio corresponding to the interview text; determining semantic integrity probabilities of the interview text; Determining the probability that the candidate does not continue speaking according to the first speech speed, the first pause frequency and the semantic integrity probability; determining whether to continue asking questions according to the probability that the candidate does not continue speaking; The method comprises the steps of acquiring interview texts, separating voice audio paragraphs from the interview audios by using a preset voice activity detection algorithm, converting the voice audio paragraphs into candidate texts, determining whether the content of each candidate text is effective, and taking the effective candidate texts as the interview texts; The method further comprises the steps of converting the interview audio collected in the preset time length into a text to be judged every other preset time length in the interview audio collecting process, determining a second speech speed and a second pause frequency of the candidate according to the text to be judged, describing pause times of the candidate in the interview audio corresponding to the text to be judged, and adjusting non-speaking interval parameters of the voice activity detection algorithm according to the second speech speed and the second pause frequency.
  2. 2. The method of claim 1, wherein determining whether the content of each of the candidate texts is valid comprises: For each of the candidate texts: removing stop words in the text to be selected; counting the number of the characters remained in the text to be selected; and under the condition that the number of the characters is larger than the preset number of the characters, determining that the content of the text to be selected is valid.
  3. 3. The method of claim 1, wherein adjusting the non-speaking interval parameter of the voice activity detection algorithm based on the second speech rate and the second pause frequency comprises: Determining a first speech speed deviation degree according to the second speech speed and a preset average speech speed; Determining a first frequency deviation degree according to the second pause frequency and a preset pause frequency; And determining a non-speaking interval parameter according to the first speech speed deviation degree and the first frequency deviation degree.
  4. 4. The method of claim 1, wherein determining the semantic integrity probability of the interview text comprises: inputting the interview text into a preset first semantic integrity model for identification to obtain a first semantic integrity score; Inputting the interview text into a preset second semantic integrity model for recognition to obtain a second semantic integrity score, wherein the second semantic integrity model and the first semantic integrity model are different models; And weighting the first semantic integrity score and the second semantic integrity score according to a preset first weight set to obtain semantic integrity probability.
  5. 5. The method of claim 1, wherein determining the probability that the candidate is not speaking based on the first speech rate, the first pause frequency, and the semantic integrity comprises: Acquiring historical correct separation probability of the voice activity detection algorithm for separating voice audio paragraphs; and determining the probability that the candidate does not continue speaking according to the historical correct separation probability, the first speech speed, the first pause frequency and the semantic integrity probability.
  6. 6. The method of claim 5, wherein determining the probability that the candidate does not continue speaking based on the historical correct separation probability, the first speech rate, the first pause frequency, and the semantic integrity probability comprises: Determining the voice characteristic deviation degree according to the first speech speed and the first pause frequency; And weighting the historical correct separation probability, the semantic integrity probability and the voice characteristic deviation degree according to a preset second weight set to obtain the probability that the candidate does not continue speaking.
  7. 7. The method of claim 6, wherein determining a degree of departure of a speech feature based on the first speech rate and the first frequency of pauses comprises: determining a second speech speed deviation degree according to the first speech speed and a preset average speech speed; Determining a second pause frequency deviation degree according to the first pause frequency and a preset pause frequency; and determining the voice characteristic deviation degree according to the second speech speed deviation degree and the second pause frequency deviation degree.
  8. 8. The method of any of claims 1 to 7, wherein determining whether to continue asking questions based on the probability that the candidate does not continue speaking comprises: carrying out intention recognition on the interview text to obtain a first intention recognition result, wherein the first intention recognition result is used for describing whether the intention of the candidate is that the answer is completed; And when the first intention recognition result is that the answer is finished, and the probability that the candidate does not continue speaking is larger than the set probability, determining to continue asking questions.
  9. 9. The method of claim 8, wherein after determining to continue the question, the method further comprises: Determining the answer detailed degree of the candidate according to the interview text; determining the answer comprehensiveness degree of the candidate according to the interview text; and determining the interview question text according to the answer detailed degree and the answer comprehensive degree.
  10. 10. The method of claim 9, wherein determining the answer details of the candidate from the interview text comprises: and inputting a preset evaluation rule of the detail degree and the interview text into a preset first large language model to obtain the answer detail degree.
  11. 11. The method of claim 9, wherein determining the comprehensiveness of the candidate's answer based on the interview text comprises: and inputting a preset examination key point and the interview text into a preset second large language model to obtain the comprehensive degree of the answer.
  12. 12. An electronic device comprising a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the AI interview method of any one of claims 1-11.
  13. 13. A storage medium storing computer-executable instructions that, when invoked and executed by a processor, cause the processor to implement the AI interview method of any one of claims 1-11.
  14. 14. A computer program product, characterized in that it comprises a computer program which, when executed by a processor, implements the AI interview method of any one of claims 1 to 11.

Description

AI interview method, electronic device, storage medium, and program product Technical Field The application relates to the technical field of artificial intelligence, in particular to an AI interview method, electronic equipment, a storage medium and a program product. Background When recruiting employees, companies typically need to interview candidates for delivering resumes, thereby screening suitable candidates as employees. However, many candidates usually compete for the same post, and only manually interviewing the candidates one by one wastes more time. In the related art, AI (artificial intelligence) interviewees are used to interview candidates to screen out a part of the candidates, and then the screened candidates are interviewed again manually. However, when asking a question about the candidate, the AI interviewee usually directly accumulates the non-speaking time of the candidate, and if the non-speaking time of the candidate reaches the set time, the AI interviewee continues to ask the next question. However, asking questions in this way may result in a longer waiting time for the candidate to answer the next question if the set duration is set too long, and if the set duration is set too short, the AI interviewee may interrupt the candidate to answer the question. In the process of implementing the embodiment of the application, the related art is found to have at least the following problems: AI interviewees cannot take over a conversation at a reasonable opportunity. It should be noted that the information disclosed in the above background section is only for enhancing understanding of the background of the application and thus may include information that does not form the prior art that is already known to those of ordinary skill in the art. Disclosure of Invention The following presents a simplified summary in order to provide a basic understanding of some aspects of the disclosed embodiments. This summary is not an extensive overview, and is intended to neither identify key/critical elements nor delineate the scope of such embodiments, but is intended as a prelude to the more detailed description that follows. The embodiment of the application provides an AI interview method, electronic equipment, storage medium and program product, so that AI interviewee officers take over conversations at reasonable occasions. The embodiment of the application provides an AI interview method, which comprises the steps of obtaining an interview text, determining a first speech speed and a first pause frequency of a candidate according to the interview text, wherein the first pause frequency is used for describing pause times of the candidate in the process of inputting interview audio corresponding to the interview text, determining semantic integrity probability of the interview text, determining probability that the candidate does not continue speaking according to the first speech speed, the first pause frequency and the semantic integrity probability, and determining whether to continue asking according to the probability that the candidate does not continue speaking. In the above embodiment, the speaking style of the candidate can be reflected by determining the first speech rate and the first pause frequency of the candidate. By combining the speaking style of the candidate and the semantic integrity probability, whether the candidate still continues to answer can be accurately deduced. Whether to continue asking is determined according to whether the candidate needs to continue answering or not, and the timing of the AI interviewee to continue asking can be adaptively adjusted according to the speaking style of the candidate, so that the AI interviewee takes over the dialogue at a reasonable timing, and the interview experience of the candidate is improved. Further, the interview text corresponding to the interview audio is obtained, the interview audio is collected, a voice audio paragraph is separated from the interview audio by means of a preset voice activity detection algorithm, the voice audio paragraph is an audio paragraph in which candidate voices exist, the voice audio paragraph is converted into candidate texts, whether the content of each candidate text is effective or not is determined, and the effective candidate texts are used as the interview text. In the above embodiment, the voice audio paragraphs are separated from the interview audio by the voice activity detection algorithm, and the candidate text corresponding to the voice audio paragraphs with valid content is screened as the interview text. More effective information can be retained in the interview text. Meanwhile, partial text to be selected with little meaning is removed, so that the pressure of the subsequent processing of the interview text can be reduced, and whether to continue asking questions can be judged more quickly. Further, the method further comprises the steps of converting the interview audio collected