Interview dialogue audio processing method and device, electronic equipment and storage medium

Abstract

The invention provides an interview dialogue audio processing method and device, electronic equipment and a storage medium, and relates to the technical field of AI interview. In the method, a first vocabulary and a second vocabulary are respectively extracted from resumes of interview candidates and standard answers of interview questions by utilizing a pre-training language model, and the first vocabulary and the second vocabulary comprise common words and non-daily common words related to posts on which the interview candidates face; generating an intervention hot word list according to the first vocabulary and the second vocabulary; and text conversion processing is performed on the interview dialogue audio based on the intervention hot word list to obtain the conversion text corresponding to the interview dialogue audio, so that when the interview candidates use some professional nouns or industry terms, recognition errors are reduced, and the accuracy of the obtained conversion text is improved.

Description

Technical Field

[0001] The present invention relates to the field of AI interview technology, and in particular to a method, device, electronic device and storage medium for processing interview dialogue audio. Background Art

[0002] In traditional interviews, recording interview conversations typically relies on manual note-taking or audio recording, a time-consuming and prone to inaccuracies. In recent years, automatic speech recognition technology has been introduced to transcribe interview conversations, automatically converting audio into text. However, existing automatic speech recognition technology remains less than ideal when faced with interview questions using specialized terminology or specific fields. This is particularly true when candidates use specialized terms or industry jargon, which can lead to recognition errors. Consequently, the converted text often requires manual proofreading and revision, further increasing time and labor costs. Summary of the Invention

[0003] The present invention provides an interview conversation audio processing method, device, electronic device and computer-readable storage medium, which can improve the accuracy of converted text corresponding to the interview conversation audio.

[0004] The technical solution of the present invention can be implemented as follows:

[0005] In a first aspect, the present invention provides a method for processing audio of an interview conversation, the method comprising:

[0006] Extracting a first vocabulary and a second vocabulary from the interview candidate's resume and the standard answer to the interview question using a pre-trained language model, wherein the first vocabulary and the second vocabulary include common words and non-daily words related to the position of the interview candidate;

[0007] generating an intervention hot word list according to the first vocabulary and the second vocabulary;

[0008] The interview conversation audio is subjected to text conversion processing based on the intervention hot word list to obtain a conversion text corresponding to the interview conversation audio.

[0009] Optionally, the step of generating an intervention hot word list based on the first vocabulary and the second vocabulary includes:

[0010] If the sum of the number of words in the first vocabulary and the number of words in the second vocabulary does not exceed a preset upper limit, each word in the first vocabulary and the second vocabulary is added as a hot word to the intervention hot word table;

[0011] If the sum of the number of words in the first vocabulary and the number of words in the second vocabulary exceeds the preset upper limit, each word in the first vocabulary is added as a hot word to the intervention hot word table.

[0012] Optionally, a speech recognition system is deployed on the electronic device, and the step of performing text conversion processing on the interview conversation audio based on the intervention hot word list to obtain the converted text corresponding to the interview conversation audio includes:

[0013] Inputting the interview conversation audio and the intervention hot word list into the speech recognition system, so that the speech recognition system performs speech recognition on the interview conversation audio according to each hot word in the intervention hot word list, and outputs a recognition text corresponding to the interview conversation audio;

[0014] The recognized text is used as the converted text.

[0015] Optionally, the speech recognition system performs speech recognition on the interview conversation audio according to each hot word in the intervention hot word list, and the step of outputting the recognition text corresponding to the interview conversation audio includes:

[0016] Before performing speech recognition, the speech recognition system sets a recognition weight for each hot word according to the frequency of occurrence of each hot word in the intervention hot word table;

[0017] The speech recognition system performs speech recognition on the interview conversation audio according to the recognition weight of each hot word, and dynamically adjusts the recognition weight of each recognition hot word in real time according to the recognition situation during the speech recognition process, until the recognition text is output when the speech recognition is completed.

[0018] Optionally, the step of dynamically adjusting the recognition weight of each hot word according to the recognition situation in real time during the speech recognition process includes:

[0019] For each of the hot words, if the speech recognition system detects that the hot word appears in the recognition text generated up to the current moment, the recognition weight of the hot word is increased.

[0020] Optionally, the step of dynamically adjusting the recognition weight of each hot word according to the recognition situation in real time during the speech recognition process includes:

[0021] For each of the hot words, if the speech recognition system detects that the matching degree between the phoneme of the hot word and the segment of the interview conversation audio recognized at the current moment meets a preset condition, the recognition weight of the hot word is increased.

[0022] Optionally, before using the recognized text as the converted text, the step of performing text conversion processing on the interview conversation audio based on the intervention hot word list to obtain the converted text corresponding to the interview conversation audio further includes:

[0023] The pre-trained language model is used to perform error correction processing on the recognized text according to preset error correction instructions, and the processed recognized text is used as the converted text.

[0024] In a second aspect, the present invention provides an interview conversation audio processing device, the device comprising:

[0025] An extraction module is configured to extract a first vocabulary and a second vocabulary from the interview candidate's resume and the standard answer to the interview question respectively using a pre-trained language model, wherein the first vocabulary and the second vocabulary include common words and non-daily words related to the position of the interview candidate;

[0026] A generating module, configured to generate an intervention hot word list based on the first vocabulary and the second vocabulary;

[0027] The conversion module is used to perform text conversion processing on the interview dialogue audio based on the intervention hot word list to obtain the converted text corresponding to the interview dialogue audio.

[0028] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method for processing audio of an interview conversation as described in the first aspect is implemented.

[0029] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the interview conversation audio processing method as described in the first aspect above.

[0030] Compared to the prior art, the present invention provides an interview conversation audio processing method: using a pre-trained language model to extract a first vocabulary and a second vocabulary including common words and non-daily words related to the position that the interview candidate is applying for from the interview candidate's resume and the standard answers to the interview questions; generating an intervention hot word list based on the first vocabulary and the second vocabulary; performing text conversion processing on the interview conversation audio based on the intervention hot word list to obtain a converted text corresponding to the interview conversation audio. Since the present invention extracts non-daily words and common words related to the position that the interview candidate is applying for from the interview candidate's resume and the standard answers to the interview questions to generate an intervention hot word list for text conversion processing of the interview conversation audio, it reduces recognition errors when the interview candidate uses some professional terms or industry terms, and improves the accuracy of the converted text obtained.

[0031] Other features and advantages disclosed in the present invention will be described in the following description, or some features and advantages can be inferred or determined without doubt from the description, or can be learned by practicing the above-mentioned technology disclosed in the present invention.

[0032] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1 A flowchart of a method for processing audio of an interview conversation provided by an embodiment of the present invention;

[0035] Figure 2 A schematic diagram of a process for implementing step S103 provided in an embodiment of the present invention Figure 1 ;

[0036] Figure 3 A schematic diagram of a process for implementing step S103 provided in an embodiment of the present invention Figure 2 ;

[0037] Figure 4 A functional unit block diagram of a device for processing audio for interview conversations provided by an embodiment of the present invention;

[0038] Figure 5 A schematic block diagram of the structure of an electronic device provided by an embodiment of the present invention.

[0039] Icon: 100 - interview dialogue audio processing device; 101 - extraction module; 102 - generation module; 103 - conversion module; 200 - electronic device; 210 - memory; 220 - processor. DETAILED DESCRIPTION

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0041] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0042] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0043] In addition, the terms "first", "second", etc., if used, are merely used to distinguish and describe, and should not be understood as indicating or implying relative importance.

[0044] It should be noted that, in the absence of conflict, the features in the embodiments of the present invention may be combined with each other.

[0045] In order to improve the accuracy of the converted text corresponding to the interview conversation audio, the present invention provides an interview conversation audio processing method, which will be introduced in detail below.

[0046] Please refer to Figure 1 The interview dialogue audio processing method includes steps S101 to S103.

[0047] S101, using a pre-trained language model to extract a first word and a second word from the interview candidate's resume and the standard answer to the interview question.

[0048] The first and second words are common and non-daily words related to the position that the interview candidate is facing.

[0049] The first vocabulary includes common words related to the position that the interview candidate is applying for, which may be campus vocabulary, company terms, professional terms, industry terms, etc. extracted from information such as "professional skills", "work experience", "project experience" and "internship experience" in the interview candidate's resume.

[0050] Input the interview candidate resume into the pre-trained language model and write a suitable prompt sentence so that the pre-trained language model can extract the first word from the interview candidate resume according to the prompt word.

[0051] Exemplarily, when extracting the first vocabulary, the prompt statements written and input into the pre-trained language model can be "Please identify and list all education-related terms and phrases from the following resume text, such as degrees, school names, and educational programs", "Please extract all company-specific terms and phrases from this resume, including job titles, department names, and corporate structures", "Please find and list all industry-specific terms and concepts in the resume, which should reflect the industry and market in which the individual has experience", "Please extract a list of professional vocabulary and technical terms from the resume, which are not commonly used in everyday language and focus on terms in specific fields", and "Please extract and classify the following types of terms from the resume: 1. Campus vocabulary (for example: university, degree, college); 2. Company terms (for example: CEO, department, merger); 3. Professional terms (for example: software engineer, financial analyst); 4. Industry terms (for example: healthcare, automotive, technology); 5. Non-everyday words (for example: quantum computing, bioremediation); please provide structured output with the terms in each category clearly listed."

[0052] Similarly, the second vocabulary includes common words related to the position that the interview candidate is interviewing for, and may be professional terms and industry terms extracted from standard answers to interview questions.

[0053] The standard answers to interview questions can be pre-set by the question setter, or generated by answering interview questions using a pre-trained language model.

[0054] The interview questions and the corresponding standard answers are input into the pre-trained language model, and appropriate prompt sentences are written so that the pre-trained language model can extract the second vocabulary from the interview questions and the standard answers according to the prompt words.

[0055] For example, when extracting the second vocabulary, the prompt statement written and input into the pre-trained language model can be "Please find all industry-specific terms and concepts in the input text, which should reflect the industry and market in which the individual has experience", "Please extract a list of professional vocabulary and technical terms from the input text. These words are not commonly used in everyday language and focus on terms in specific fields", "Please extract and classify the following types of terms from the input text: 1. Professional terms (for example: software engineers, financial analysts); 2. Industry terms (for example: healthcare, automobiles, technology); 3. Non-everyday words (for example: quantum computing, bioremediation); Please provide structured output with the terms in each category clearly listed."

[0056] S102: Generate an intervention hot word list based on the first word and the second word.

[0057] Among them, the intervention hot word list has a capacity upper limit, that is, the maximum number of hot words that can be included in the intervention hot word list. The intervention hot word list for converting audio into text is generated based on the number of words in the first vocabulary, the number of words in the second vocabulary and the capacity upper limit of the intervention hot word list.

[0058] In a possible implementation, step S102 may have the following two situations:

[0059] In case 1, if the sum of the number of words in the first vocabulary and the number of words in the second vocabulary does not exceed a preset upper limit, each word in the first vocabulary and the second vocabulary is added as a hot word into the intervention hot word list.

[0060] In the second scenario, if the sum of the number of words in the first vocabulary and the number of words in the second vocabulary exceeds a preset upper limit, each word in the first vocabulary is added as a hot word to the intervention hot word list.

[0061] The preset upper limit value refers to the capacity upper limit of the intervention hot word list, that is, the number of hot words included in the intervention hot word list cannot exceed the preset upper limit value.

[0062] It is understandable that if the sum of the number of words in the first vocabulary and the number of words in the second vocabulary does not exceed the upper limit of the intervention hot word list, all words in the first vocabulary and the second vocabulary can be added as hot words to the intervention hot word list.

[0063] If the sum of the number of words in the first vocabulary and the number of words in the second vocabulary exceeds the upper limit of the intervention hot word list, then the first vocabulary has a higher priority than the second vocabulary, that is, each word in the first vocabulary is added to the intervention hot word list as a hot word.

[0064] The reason for doing this is that the first word is extracted from the interview candidate's resume, and the interview candidate's resume is written by the interview candidate. Accordingly, the interview candidate will refer to the resume he or she wrote when answering. That is, the first word is more closely related to the interview candidate himself or herself and is more likely to be mentioned by the interview candidate when answering in the interview.

[0065] S103: Perform text conversion processing on the interview conversation audio based on the intervention hot word list to obtain a converted text corresponding to the interview conversation audio.

[0066] In a possible implementation, a speech recognition system is deployed on the electronic device, and step S103 may include: Figure 2 Sub-steps S103-1 to S103-2 are shown.

[0067] S103-1, input the interview conversation audio and the intervention hot word list into the speech recognition system, so that the speech recognition system performs speech recognition on the interview conversation audio according to each hot word in the intervention hot word list, and outputs the recognition text corresponding to the interview conversation audio.

[0068] Speech recognition systems typically include components such as "sound capture," "preprocessing," "feature extraction," "acoustic model," "language model," "decoder," and "post-processing." By integrating intervention hotword lists into components like the acoustic model, language model, and decoder, the speech recognition system's performance for professional terms and industry jargon can be improved, thus meeting the speech-to-text conversion requirements of interview scenarios.

[0069] The speech recognition system performs speech recognition on the interview conversation audio according to each hot word in the intervention hot word list, and outputs the recognition text corresponding to the interview conversation audio. The implementation process can be as follows:

[0070] Before performing speech recognition, the speech recognition system sets a recognition weight for each hot word according to the frequency of occurrence of each hot word in the intervention hot word list;

[0071] The speech recognition system performs speech recognition on the interview conversation audio according to the recognition weight of each hot word, and dynamically adjusts the recognition weight of each recognition hot word in real time according to the recognition situation during the speech recognition process until the recognition text is output when the speech recognition is completed.

[0072] It is understandable that before speech recognition begins, the speech recognition system will perform word frequency statistics on the input intervention hot word list, and use the frequency of occurrence of each hot word in the intervention hot word list as its initial recognition weight.

[0073] It should be noted that the words in the first vocabulary are more closely related to the interview candidate's own information and are more likely to be mentioned in the interview answer. After determining the initial recognition weight of each hot word based on the word frequency, the initial recognition weight of the hot words belonging to the first vocabulary can be further set, for example, adding a fixed value on the basis of the set initial recognition weight.

[0074] After determining the initial recognition weight of each hot word, the speech recognition system begins to perform speech recognition on the interview conversation audio.

[0075] The recognition weight of a hot word affects the decoding process, speech model, and acoustic model. During decoding, the decoder will prefer paths containing hot words with high recognition weights during the search process. In language models, hot words with high recognition weights increase the probability of sentences or phrases containing these hot words, thus affecting the selection of decoding paths. When training an acoustic model, hot words with high recognition weights may cause the model to focus more on the acoustic features associated with these words.

[0076] During the entire speech recognition process, the speech recognition system needs to dynamically adjust the recognition weight of each hot word in the intervention hot word list based on the recognition status of the interview conversation audio in real time. There are two implementation methods.

[0077] Method 1: For each hot word, if the speech recognition system detects that the hot word appears in the recognition text generated up to the current moment, the recognition weight of the hot word is increased.

[0078] At every moment during the speech recognition process, the speech recognition system monitors the generated recognized text to check whether any hot words in the intervention hot word list appear in the generated recognized text. If a hot word in the intervention hot word list appears in the generated recognized text, the recognition weight of such hot words is increased.

[0079] When improving the recognition weight of hot words, it can be achieved according to the preset weight improvement rules. The weight improvement rules can be to set a weight improvement multiple or increase a fixed value.

[0080] Method 2: For each hot word, if the speech recognition system detects that the matching degree between the phoneme of the hot word and the segment of the interview dialogue audio recognized at the current moment meets the preset conditions, the recognition weight of the hot word is increased.

[0081] The speech recognition system segments the interview audio into phoneme-level segments and constructs a phoneme model for each hot word. Each hot word phoneme model contains the acoustic characteristics of each phoneme in the hot word. During the speech recognition process, the speech recognition system matches the phoneme segments of the interview audio with the phoneme model of each hot word in real time, calculates the degree of match (such as a similarity score), and increases the recognition weight of hot words whose match exceeds a preset threshold.

[0082] When increasing the weight of a hot word, the extent of the weight increase can be determined based on the degree of matching.

[0083] It can be understood that the speech recognition system dynamically adjusts the recognition weight of each hot word so that hot words with high recognition weight are more easily selected during the speech recognition process.

[0084] S103-2, using the recognized text as the converted text.

[0085] In an embodiment of the present invention, the recognition text output by the speech recognition system can be used as the converted text corresponding to the interview conversation audio.

[0086] To further improve the readability of the converted text corresponding to the interview dialogue audio, please refer to Figure 3 The implementation process of step S103 may also include Figure 3 Sub-step S103-3 is shown.

[0087] S103-3, using the pre-trained language model, perform error correction processing on the recognized text according to preset error correction instructions.

[0088] After obtaining the recognized text output by the speech recognition system, the recognized text and interview questions can be input into the pre-trained language model. The pre-trained language model will perform semantic-level error correction on the recognized text according to preset error correction instructions to improve the content accuracy and content structure organization of the recognized text, and use the corrected recognized text as the converted text corresponding to the interview dialogue audio.

[0089] The preset error correction instructions may include the following ordered instructions:

[0090] Instruction 1: If the original sentence already looks good, you don't have to change it.

[0091] Instruction 2: Keep the modal particles in the original sentence and their positions in the sentence unchanged; keep the sentence structure and word order unchanged.

[0092] Command 3: Determine whether there are any common spelling errors, such as translation errors caused by differences in pronunciation and accent. If so, make necessary corrections, prioritizing appropriate words with similar pronunciations.

[0093] Instruction 4: Determine if there are any translation errors in professional terms. If so, correct them by using the professional terms that are closest in pronunciation to the English and Chinese versions.

[0094] Instruction 5: Everyone speaks differently during an interview, which can lead to issues with coherence. Determine whether the text is properly punctuated. If the punctuation is severely inappropriate, add appropriate punctuation.

[0095] Instruction 6: Compare the text content before and after correction. If there is a change in the semantics of the content between the two versions, or if there is a significant change in the candidate's speech expression characteristics, then abandon this correction instruction.

[0096] In order to execute the corresponding steps in the above method embodiment and various possible implementations, an implementation of the interview dialogue audio processing device 100 is provided below.

[0097] Please refer to Figure 4 The interview dialogue audio processing device 100 may include an extraction module 101 , a generation module 102 and a conversion module 103 .

[0098] The extraction module 101 is used to extract a first vocabulary and a second vocabulary from the interview candidate's resume and the standard answer to the interview question respectively using a pre-trained language model. The first vocabulary and the second vocabulary include common words and non-daily words related to the position of the interview candidate.

[0099] The generating module 102 is configured to generate an intervention hot word list based on the first word and the second word.

[0100] The conversion module 103 is used to perform text conversion processing on the interview conversation audio based on the intervention hot word list to obtain the converted text corresponding to the interview conversation audio.

[0101] Optionally, the generation module 102 is specifically used to add each word in the first vocabulary and the second vocabulary as a hot word to the intervention hot word list if the sum of the number of words in the first vocabulary and the number of words in the second vocabulary does not exceed a preset upper limit; if the sum of the number of words in the first vocabulary and the number of words in the second vocabulary exceeds a preset upper limit, then add each word in the first vocabulary as a hot word to the intervention hot word list.

[0102] Optionally, a speech recognition system is deployed on the electronic device, and the conversion module is specifically used to input the interview conversation audio and the intervention hot word list into the speech recognition system, so that the speech recognition system performs speech recognition on the interview conversation audio according to each hot word in the intervention hot word list, and outputs the recognition text corresponding to the interview conversation audio; and uses the recognition text as the conversion text.

[0103] Optionally, the conversion module 103 is further specifically configured to utilize a pre-trained language model to perform error correction processing on the recognized text according to preset error correction instructions, and use the processed recognized text as the converted text.

[0104] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the interview dialogue audio processing device 100 described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0105] Furthermore, the embodiment of the present invention also provides an electronic device 200, please refer to Figure 5 , the electronic device 200 may include a memory 210 and a processor 220 .

[0106] Among them, the processor 220 can be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of a program of an interview dialogue audio processing method provided in the above-mentioned method embodiment.

[0107] The memory 210 can be a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to this. The memory 210 can exist independently and be connected to the processor 220 via a communication bus. The memory 210 can also be integrated with the processor 220. Among them, the memory 210 is used to store machine executable instructions for executing the scheme of the present application. The processor 220 is used to execute the machine executable instructions stored in the memory 210 to implement the above-mentioned method embodiment.

[0108] An embodiment of the present invention also provides a computer-readable storage medium containing a computer program. When the computer program is executed, it can be used to perform relevant operations in the interview dialogue audio processing method provided in the above method embodiment.

[0109] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for processing interview dialogue audio, characterized in that: The method comprises: Extracting a first vocabulary and a second vocabulary from the interview candidate's resume and the standard answer to the interview question respectively using a pre-trained language model, wherein the first vocabulary and the second vocabulary include common words and non-daily words related to the position of the interview candidate; Generate an intervention hot word list according to the first vocabulary and the second vocabulary; The interview conversation audio is subjected to text conversion processing based on the intervention hot word list to obtain a conversion text corresponding to the interview conversation audio.

2. The method according to claim 1, characterized in that The step of generating an intervention hot word list according to the first vocabulary and the second vocabulary comprises: If the sum of the number of words in the first vocabulary and the number of words in the second vocabulary does not exceed a preset upper limit, each word in the first vocabulary and the second vocabulary is added as a hot word to the intervention hot word list; If the sum of the number of words in the first vocabulary and the number of words in the second vocabulary exceeds the preset upper limit, each word in the first vocabulary is added to the intervention hot word table as a hot word.

3. The method according to claim 1, characterized in that The electronic device is equipped with a speech recognition system, and the step of converting the interview conversation audio into text based on the intervention hot word list to obtain the converted text corresponding to the interview conversation audio includes: Inputting the interview conversation audio and the intervention hot word list into the speech recognition system, so that the speech recognition system performs speech recognition on the interview conversation audio according to each hot word in the intervention hot word list, and outputs a recognition text corresponding to the interview conversation audio; The recognized text is used as the converted text.

4. The method according to claim 3, characterized in that The speech recognition system performs speech recognition on the interview conversation audio according to each hot word in the intervention hot word list, and the step of outputting the recognition text corresponding to the interview conversation audio includes: The speech recognition system sets a recognition weight for each hot word according to the frequency of occurrence of each hot word in the intervention hot word table before performing speech recognition; The speech recognition system performs speech recognition on the interview conversation audio according to the recognition weight of each hot word, and dynamically adjusts the recognition weight of each recognition hot word in real time according to the recognition situation during the speech recognition process, until the recognition text is output when the speech recognition is completed.

5. The method according to claim 4, characterized in that The step of dynamically adjusting the recognition weight of each hot word according to the recognition situation in real time during the speech recognition process of the speech recognition system includes: For each of the hot words, if the speech recognition system detects that the hot word appears in the recognition text generated up to the current moment, the recognition weight of the hot word is increased.

6. The method according to claim 4, characterized in that The step of dynamically adjusting the recognition weight of each hot word according to the recognition situation in real time during the speech recognition process of the speech recognition system includes: For each of the hot words, if the speech recognition system detects that the matching degree between the phonemes of the hot word and the segment of the interview conversation audio recognized at the current moment meets a preset condition, the recognition weight of the hot word is increased.

7. The method according to claim 3, characterized in that Before using the recognized text as the converted text, the step of performing text conversion processing on the interview conversation audio based on the intervention hot word list to obtain the converted text corresponding to the interview conversation audio also includes: The pre-trained language model is used to perform error correction processing on the recognized text according to preset error correction instructions.

8. An interview dialogue audio processing device, characterized in that: The device comprises: An extraction module, configured to extract a first vocabulary and a second vocabulary from the interview candidate's resume and the standard answer to the interview question respectively using a pre-trained language model, wherein the first vocabulary and the second vocabulary include common words and non-daily words related to the position of the interview candidate; A generating module, configured to generate an intervention hot word list according to the first vocabulary and the second vocabulary; The conversion module is used to perform text conversion processing on the interview conversation audio based on the intervention hot word list to obtain the converted text corresponding to the interview conversation audio.

9. An electronic device, characterized in that: It includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the interview dialogue audio processing method as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: It stores a computer program, which, when executed by a processor, implements the interview conversation audio processing method as described in any one of claims 1 to 7.

Images