CN110347787B - Interview method and device based on AI auxiliary interview scene and terminal equipment - Google Patents
Interview method and device based on AI auxiliary interview scene and terminal equipment Download PDFInfo
- Publication number
- CN110347787B CN110347787B CN201910504604.2A CN201910504604A CN110347787B CN 110347787 B CN110347787 B CN 110347787B CN 201910504604 A CN201910504604 A CN 201910504604A CN 110347787 B CN110347787 B CN 110347787B
- Authority
- CN
- China
- Prior art keywords
- interview
- question
- engine
- key information
- word segmentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/3332—Query translation
- G06F16/3334—Selection or weighting of terms from queries, including natural language queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3343—Query execution using phonetics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/105—Human resources
- G06Q10/1053—Employment or hiring
Abstract
The invention is suitable for the technical field of computer application, and provides an interview method, an interview device and terminal equipment based on an AI auxiliary interview scene, wherein the method comprises the following steps: acquiring voice content expressed by an interviewer, and converting the voice content into text content; performing word segmentation processing on the text content to obtain a word segmentation database; acquiring key information from a word segmentation database; judging the type of the current interview scene according to the key information, selecting whether to start a similar question engine according to the type of the interview scene, acquiring a question template according to the processing result of the similar question engine when the similar question engine is started, and selecting a scoring model to score the answer of an interviewee; and when the similar question engine is not started, obtaining a question template according to the key information, and selecting a scoring model to score the answer of the interviewer. The invention can improve the accuracy of AI interview evaluation while maintaining the flexibility and subjective initiative of real human interview.
Description
Technical Field
The invention relates to the technical field of computer application, in particular to an interview method and device based on an AI auxiliary interview scene and terminal equipment.
Background
And the AI auxiliary interview is the behavior of performing on-site or remote interview on the candidate by the real-person interviewer with the assistance of the AI auxiliary interview system. Compared with the traditional interview scene, the AI auxiliary interview can provide candidate information for interviewers, recommend interview problems, record an interview process and generate an evaluation report for interview results, and has the advantages of process structuralization and scoring standardization.
In the AI auxiliary interview scene, the AI auxiliary interview system generates a corresponding question list from the question bank based on the current interview job capability portrait and plays the roles of prompting interviewers and scoring questions.
However, due to the limitation of the AI technology at the present stage, the scoring function is only suitable for the existing problems in the question bank, and the scoring function does not have the scoring capability for the problems temporarily and instantly asked by the interviewer. On the other hand, in the actual interviewing process, even with questions in the question bank, the interviewer's question order and specific language expression are likely to be different from the standard question method, which poses a challenge to the stability of AI-assisted interviews.
Disclosure of Invention
The invention mainly aims to provide an interview method, an interview device and terminal equipment based on an AI auxiliary interview scene, and aims to solve the problems that in the prior art, when an AI technology is used for assisting interview, the problem of temporary improvision of interviewers does not have scoring capability, and in the interview of the interviewers, the existing problems in an interview base are inaccurate in score judgment and the analysis of interview results is unstable and reliable.
In order to achieve the above object, a first aspect of the embodiments of the present invention provides an interview method based on an AI-assisted interview scene, including:
acquiring voice content expressed by an interviewer, and converting the voice content into text content;
performing word segmentation processing on the text content to obtain a word segmentation database;
acquiring key information from the word segmentation database;
judging the type of the current interview scene according to the key information, and selecting whether to start a similar question engine according to the type of the interview scene, wherein the similar question engine is used for matching a question template in a question database according to word segmentation word vectors;
when the similar question engine is started, a question template is obtained according to the processing result of the similar question engine, and a scoring model is selected to score the answer of the interviewee;
and when the similar question engine is not started, acquiring a question template according to the key information, and selecting a scoring model to score the answer of the interviewer.
With reference to the first aspect of the present invention, in a first implementation manner of the first aspect of the present invention, the determining a current interview scene type according to the key information, and selecting whether to enable a similar question engine according to the interview scene type includes:
detecting the quantity of short sentences in the word segmentation database and the quantity of problem IDs of the key information;
if the number of the question IDs of the key information is the same as the number of the short sentences in the word segmentation database, the type of the current interview scene is a human-computer interview scene, and a similar question engine is not started;
and if the number of the question IDs of the key information is different from the number of the short sentences in the word segmentation database, the current interview scene type is a field auxiliary interview scene, and a similar question engine is started.
With reference to the first aspect of the present invention, in a second embodiment of the first aspect of the present invention, when the similar question engine is started, the step of obtaining a question template according to a processing result of the similar question engine, and selecting a scoring model to score answers of an interviewee includes:
when an affinity engine is started, calling a trained matching neural network model in the affinity engine, and converting the key information into the word segmentation word vector through the matching neural network model;
matching a question template in the question database according to the word segmentation word vector;
and calling a scoring model corresponding to the successfully matched question template, and scoring the answer of the interviewer according to the scoring model.
With reference to the second implementation manner of the first aspect of the present invention, in a third implementation manner of the present invention, matching a question template in the question database according to the word segmentation word vector includes:
acquiring a template word vector of each question template;
comparing the word segmentation word vector with the template word vector, and calculating cosine similarity;
and judging whether the word segmentation word vector is matched with the problem template or not according to the cosine similarity.
With reference to the first aspect of the present invention, in a fourth implementation manner of the first aspect of the present invention, when the similar question engine is not started, obtaining a question template according to the key information, and selecting a scoring model to score the answer of the interviewer, includes:
when the similar question engine is not started, matching a question template in the question database according to the question ID in the key information;
and calling a scoring model corresponding to the successfully matched question template, and scoring the answer of the interviewee according to the scoring model.
With reference to the first aspect of the present invention, in a fifth embodiment of the first aspect of the present invention, the scoring model includes a regular scoring model and a composite scoring model;
when the interview scene type is a human-computer interview scene, the scoring model comprises a rule scoring model;
and when the interview scene is a scene of field auxiliary interview, the scoring model comprises a comprehensive scoring model.
With reference to the first aspect of the present invention, in a sixth implementation manner of the first aspect of the present invention, before performing a word segmentation process on the text content to obtain a word segmentation database, the method includes:
and carrying out noise information replacement on the text content, and filtering non-key information.
A second aspect of the present invention provides an interview apparatus based on an AI-assisted interview scenario, including:
the text acquisition module is used for acquiring the voice content expressed by the interviewer and converting the voice content into text content;
the word segmentation processing module is used for carrying out word segmentation processing on the text content to obtain a word segmentation database;
the key information acquisition module is used for acquiring key information from the word segmentation database;
the question matching module is used for judging the type of the current interview scene according to the key information and selecting whether to start a similar question engine according to the type of the interview scene, and the similar question engine is used for matching a question template in a question database according to the word segmentation word vector;
the scoring module is used for acquiring a question template according to the processing result of the similar question engine when the similar question engine is started, selecting a scoring model and scoring the answer of the interviewer;
and when the similar question engine is not started, acquiring a question template according to the key information, and selecting a scoring model to score the answer of the interviewer.
A third aspect of embodiments of the present invention provides a terminal device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method provided in the first aspect when executing the computer program.
A fourth aspect of embodiments of the present invention provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the steps of the method as provided in the first aspect above.
The embodiment of the invention provides an interview method based on an AI auxiliary interview scene, which comprises the steps of converting questions asked by an interviewer through voice into text content, performing word segmentation processing on the text content to obtain a word segmentation database based on the interview text content, analyzing the word segmentation database to obtain key information which can be used for judging the type of the current interview scene, and finally selecting a similar question engine for use according to the interview scene, so that any interview scene can be accurately matched with a question template in a question library according to the questions of the interviewer, and simultaneously selecting a proper scoring model, so that the interviewer can ask the interviewer with high freedom, and the interviewer can answer the questions described by the interviewer and obtain corresponding scores in an AI interview system.
Drawings
Fig. 1 is a schematic flow chart illustrating an implementation of an interview method based on an AI-assisted interview scene according to an embodiment of the present invention;
fig. 2 is a schematic flow chart illustrating an implementation of an interview method based on an AI-assisted interview scene according to a second embodiment of the present invention;
fig. 3 is a schematic structural diagram of a matching neural network model according to a third embodiment of the present invention;
fig. 4 is a schematic flow chart illustrating an implementation of an interview method based on an AI-assisted interview scene according to a third embodiment of the present invention;
fig. 5 is a schematic structural diagram of a component of an interview apparatus based on an AI-assisted interview scene according to a fourth embodiment of the present invention.
The implementation, functional features and advantages of the present invention will be further described with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one of 8230, and" comprising 8230does not exclude the presence of additional like elements in a process, method, article, or apparatus comprising the element.
Suffixes such as "module", "component" or "unit" used to indicate elements are used herein only for facilitating the description of the present invention, and do not have a specific meaning in themselves. Thus, "module" and "component" may be used in a mixture.
In the following description, the serial numbers of the embodiments of the invention are only for description, and do not represent the advantages and disadvantages of the embodiments.
Example one
As shown in fig. 1, an embodiment of the present invention provides an interview method based on an AI-assisted interview scenario, which assists interview by providing a suitable scoring model, and the method includes, but is not limited to, the following steps:
s101, acquiring voice content expressed by the interviewer, and converting the voice content into text content.
The step S101 may be implemented by using a front-end system; the front-end system can be realized by a PC end or a mobile phone end. The front-end system realizes the conversion from the voice content to the text content through an Automatic Speech Recognition (ASR) technology.
In one embodiment, the resulting textual content may be identified where errors in the textual content, e.g., semantic, grammatical, or/and logical errors, are by text recognition techniques, e.g., natural Language Processing (NLP) techniques; and correct for associated errors.
In the embodiment of the present invention, after the voice content is converted into the text content, the text content is a plurality of short sentences, such as: what your name is/,/what is your last position of employment/.
And S102, performing word segmentation processing on the text content to obtain a word segmentation database.
In the step S102, the word segmentation process may be completed by the front-end system, or may be completed by a central control engine that subsequently interfaces with the front-end system; the central control engine is a background data system.
In a specific application, the word segmentation processing task can be executed by a word segmentation device or a word segmentation algorithm, such as a forward maximum matching method, a reverse maximum matching method, a bidirectional matching word segmentation method, and the like.
In practical application, the text content after word segmentation takes a single word as a node, and each word has part-of-speech tagging, so that in the subsequent key information extraction, grammatical filtering is used, for example, only words with specific parts-of-speech (such as adjectives and nouns) are reserved.
In the embodiment of the invention, after the word segmentation processing is carried out on each short sentence, the data in the word segmentation database is a plurality of groups of word segments based on a plurality of groups of short sentences,/name/,/last time/position/.
In an embodiment, before the step S102, the method may include:
and carrying out noise information replacement on the text content, and filtering non-key information.
Through the noise information replacement, the text content can be simplified, so that the key information in the step S103 can be acquired more accurately and has pertinence.
In a specific application, the noise information can be name, place name, date, money and the like, and the name, place name, date, money and the like can be replaced by the replacement of the noise information by regular matching; then, chinese word segmentation (e.g., recognizing sentence components, main, predicate, object, definite, shape, etc.) is performed, and stop words are removed. Stop Words refer to that in information retrieval, in order to save storage space and improve search efficiency, some characters or Words are automatically filtered before or after processing natural language data (or text), and the characters or Words are called Stop Words.
And S103, acquiring key information in the word segmentation database.
In step S103, after the word segmentation process is completed, key information such as interviewer ID, candidate ID, company ID, scene ID, job ID, question ID, and question text can be obtained in a targeted manner.
In the embodiment of the invention, the extraction of the key information can be based on the segmentation database, and can be based on the segmentation database with filtered non-key information.
In the embodiment of the invention, a keyword extraction algorithm irrelevant to the field is selected to acquire the key information, such as a supervised learning algorithm and an unsupervised learning algorithm.
The supervised learning algorithm regards the keyword extraction process as a binary classification problem, extracts candidate words, then defines a label for each candidate word, and trains a keyword extraction classifier. When a new document comes, all candidate words are extracted, then the trained keyword extraction classifier is used for classifying all the candidate words, and finally the candidate words marked with the keywords are used as the keywords.
And the unsupervised learning algorithm firstly extracts candidate words, then scores each candidate word, and then outputs the candidate words with the highest topK scores as the keywords. According to different scoring strategies, different algorithms exist, such as TF-IDF, textRank and the like.
And S104, judging the type of the current interview scene according to the key information, and selecting whether to start a similar question engine according to the type of the interview scene.
The similarity question engine is used for matching question templates in the question database according to the word segmentation word vectors.
In one embodiment, one implementation manner of the step S104 may be:
detecting the number of short sentences in the word segmentation database and the number of problem IDs of the key information;
if the number of the question IDs of the key information is the same as the number of the short sentences in the participle database, the current interview scene type is a human-machine interview scene, and a similar question engine is not started;
and if the number of the question IDs of the key information is different from the number of the short sentences in the word segmentation database, the current interview scene type is a field auxiliary interview scene, and a similar question engine is started.
In a specific application, when the key information in each group of short sentences comprises a question ID, each question of the interviewer is a question in a question database, and the current interview scene type is a human-computer interview scene. If the key information in any group of short sentences does not include the question ID, the current interview scene type is a field auxiliary interview scene, and the interview question may not come from the question database.
S105, when the similar question engine is started, obtaining a question template according to the processing result of the similar question engine, selecting a grading model, and grading the answer of the interviewer;
and when the similar question engine is not started, acquiring a question template according to the key information, and selecting a scoring model to score the answer of the interviewer.
In practical application, text content may be a problem existing in a problem database or a problem not existing in the database, and different interview scene types are different due to different interview scene types, so that different problem template acquisition modes are adopted for different interview scene types.
In a human-computer interview scene, in a segmentation database after segmentation processing, each group of short sentences is the existing problem in the problem database, so that the problem templates corresponding to each group of short sentences can be matched according to the problem ID in the key information to complete the problem matching of the whole text content, and at the moment, a similarity question engine is not started.
In a scene of the on-site auxiliary interview, the corresponding question template cannot be accurately matched in the question database only according to the key information, and at the moment, a similar question engine needs to be started, and the key information is processed through the similar question engine.
The similarity question engine matches the question templates in the question database through word vectors of text contents, the word vectors can reflect position relations between each participle and the participle in the text contents, and when key information is extracted from the participle database, the feature extraction accuracy of the participle is improved, so that when an interview scene is a field auxiliary interview scene, the question templates corresponding to the text contents can be still accurately matched. The processing of the affinity engine is explained and illustrated in the following embodiments.
In one embodiment, the scoring model includes a rule scoring model and a composite scoring model;
when the interview scene type is a human-computer interview scene, the scoring model comprises a rule scoring model; and when the interview scene is a scene of field auxiliary interview, the scoring model comprises a comprehensive scoring model.
In the above-mentioned human-machine interview scenario, the standard questions asked by the system are usually presented, while in the scene of auxiliary interview on site, the problem that the interviewer can freely play is usually presented.
In embodiments of the present invention, the scoring models are used to determine which set of scoring models to select for evaluating the interviewer. Any interview scene type can be assisted by introducing a scoring model, and the difference is that the scoring model is selected by obtaining the question template according to the processing result of the similarity question engine, and the scoring standard or the scoring rule is different from the scoring model selected by obtaining the question template according to the key information.
The embodiment of the invention also takes a human-computer interview scene and a field auxiliary interview scene as interview scene types to exemplify the scoring standard or scoring rule in the scoring model:
when the interview scene is a human-computer interview scene, questions obtained by an interviewee come from a question database, relevant questions can be matched in the question database certainly according to key information extracted from a word segmentation database, then in the interview scene, a question template can be accurately matched according to the key information, a rule scoring model is selected, the rule scoring model can be understood as a machine-recognizable scoring mode, and the interviewee is scored according to dimensions such as answer fluency, whether answer content is matched with scoring keywords in the scoring model, and the like, wherein the dimensions include:
fluency of answer: identifying whether a pause of a longer time (judged by a threshold) occurs in the process of answering by the candidate; and identifying whether the candidate answers store continuously repeated words or phrases (identifying tension, ending or the like).
Whether the answer content matches the scoring keyword in the scoring model: some score keywords for answering the question are preset in the score model, and corresponding scores can be obtained as long as the keywords appear in the content answered by the candidate.
When the interview scene is a scene of field auxiliary interview, the questions obtained by an interviewer are not necessarily from a question database, related questions may possibly not be matched in the question database according to key information extracted from a word segmentation database, in the interview scene, a similar question engine is started to perform word vector processing, a question template can be matched according to a processing result of the word vector processing, and a comprehensive scoring model is selected, wherein the comprehensive scoring model can be understood as a scoring mode in which a machine recognizable class and a manual auxiliary judgment class are combined with each other, namely, the interviewer is scored from dimensions such as answer fluency, whether answer content is matched with a score keyword in the scoring model, and meanwhile, the interviewer is scored from 1, the interviewer evaluates the answer logicality, and 2, the interviewer evaluates the compression resistance when the interviewer answers the questions to a candidate person; 3. and the interviewee is scored according to the dimensionalities such as the evaluation of emergency ability when the candidate answers the question. In practical application, the system can provide a plurality of evaluation grades for the interviewer to select according to the evaluation grades obtained by each dimensionality and the desired weight, and the system calculates the score of the artificial auxiliary judgment class when the candidate answers. And finally, integrating scores of all dimensions of the identifiable class and the artificial auxiliary judgment class of the machine to obtain a final score when the candidate answers.
The steps S101 to S103 are implementation processes of acquiring and processing text contents, and the steps S104 and S105 are implementation processes of selecting how to match a question template and selecting a scoring model according to the interview scene type, and scoring candidate persons.
Example two
As shown in fig. 2, an interview method based on an AI-assisted interview scene is provided in the embodiment of the present invention, which shows a detailed implementation flow of step S105 in the first embodiment. When the similar question engine is started in step S105, obtaining the question template according to the processing result of the similar question engine, selecting a scoring model, and scoring the answer of the interviewer, may include:
s10511, when starting the similarity engine, calling the trained matching neural network model in the similarity engine, and converting the key information into word vectors through the matching neural network model.
In step S10511, the key information in the text content expressed by the interviewer is converted into word vectors, and the word vectors are used for matching, so that an accurate matching range can be obtained.
In one embodiment, considering that interviewers may temporarily investigate simpler or more complex questions of an interviewer to determine whether the interviewer is qualified for another position, the comparison logic may be designed to match all questions in the company database under the interview scenario of starting the affinity question engine, i.e., not limited to the question logic for that position.
In the embodiment of the invention, the neural network can be trained by the general semantic understanding and the HR related text, so that a well-trained matched neural network model is obtained.
The embodiment of the invention also shows the process of training the Word vector by using the Word2Vec neural network model:
first, as shown in FIG. 3, word2Vec includes two submodels: CBOW and Skip-Gram, in FIG. 3, part a represents CBOW, part b represents Skip-Gram, and each comprises three neural network layers of input, mapping and output, wherein part a represents output, and part b represents input.
Then, training the two sub-models, wherein the training method comprises the following steps: "surrounding words are superimposed to predict the current word" (P (wt | Context) P (wt | Context)) and "current words are used to predict surrounding words" (P (Wothers | wt)), respectively.
Wherein, the training can be carried out by using a speed-up method; in particular, the level Softmax and the negative sample sampling, respectively.
The level Softmax is a simplification of Softmax, and directly reduces the efficiency of the prediction probability from O (| V |) to O (log 2| V |), but relatively speaking, the precision is slightly lower than that of native Softmax; the negative sample sampling adopts the opposite idea, original input and output are combined to be used as input, then two classifications are made for scoring, and modeling of joint probabilities P (wt, context) and P (Wothers, wt) is formed, wherein positive samples appear by linguistic data, and negative samples are randomly drawn in a plurality of samples.
The embodiment of the invention also shows the training conditions in practical application, which comprise the following steps:
training the corpus: weChat public article, many fields, in Chinese balance corpus, HR given interview corpus and internet crawled interview scene corpus;
the amount of the corpus: 800 pieces, the total word number reaches 650 hundred million;
number of model words: 352196, which is basically a Chinese word, includes the common English words;
model structure: skip-Gram + Huffman Softmax;
vector dimension: 256 dimensions;
word segmentation tool: the Chinese character recognition method has the advantages that (1) words are formed and divided, a dictionary with 50 ten thousand entries is added, and new word discovery is closed;
training tools: word2Vec, gensim, server trained for 7 days;
other cases are as follows: the window size is 10 and the minimum word frequency is 64, 10 iterations.
In a specific application, along with the use of the AI interview system, interview contents of an interviewer and an interviewer are also recorded in the system database and used as training samples of the matching neural network model, so that the matching neural network model is further optimized in use.
In a specific application, after the training of the matching neural network model is completed, when the interview scene type is determined to be a field auxiliary interview scene, namely when the similarity engine is started, the conversion of the word vector is executed by the similarity engine.
And S10512, matching the question template in the question database according to the word segmentation word vector.
In the above step S10512, the matching process between the word vector and the word vector is performed, wherein the word segmentation vector based on the word segmentation database is completed in step S10511, and the word vector of the question template can be directly obtained.
In one embodiment, the step S10512 may include:
acquiring a template word vector of each question template;
comparing the word segmentation word vector with the template word vector, and calculating cosine similarity;
and judging whether the word segmentation word vector is matched with the problem template or not according to the cosine similarity.
In the specific application, the cosine similarity is a variable which is used for reflecting the position of each participle in a sentence in a word vector, and the problem that when an interviewer proposes a random interview, the problem that the matching of a problem template is unsuccessful or inaccurate due to the fact that the participles are reversed can be solved through the cosine similarity in the template word vector and the cosine similarity of the participle word vector based on text content.
In practical applications, the process of comparing the word vector with the template word vector may be:
after the word vectors of the text content are obtained, averaging all the word vectors to obtain the vectors of the sentences. Then comparing the sentence vector of the input interviewer corpus with the sentence vector of each standard problem template, and calculating the similarity of other strings, wherein the formula of the cosine similarity is as follows:
in specific application, the word vector is a 100-5000 dimensional vector, and all keywords can be uniformly converted into a vector with a certain fixed dimension in the same machine learning model, wherein the dimension is 256 in the proposal.
In one embodiment, an adjustable decision threshold may be set to assist in determining whether the segmented word vector matches the template word vector. If the calculated cosine similarity is larger than the threshold value, the data with the maximum similarity is considered to be the matched problem, and if the calculated cosine similarity is not larger than the threshold value, the data with the maximum similarity is considered to be not matched.
And S10513, calling a scoring model corresponding to the successfully matched question template, and scoring the answer of the interviewer according to the scoring model.
In the step S10513, the scoring models corresponding to each question template are different, and the question templates in the embodiment of the present invention are all applicable to the scene of the field-assisted interview, so that the scoring models may include a comprehensive scoring model, that is, a scoring mode in which the machine-recognizable class and the manual-assisted judgment class are combined with each other.
EXAMPLE III
As shown in fig. 4, an interview method based on an AI-assisted interview scene according to an embodiment of the present invention shows a detailed implementation flow of step S105 in the first embodiment. In step S105, when the similar question engine is not started, obtaining a question template according to the key information, and selecting a scoring model to score the answer of the interviewer, may include:
and S10521, when the similar question engine is not started, matching the question template in the question database according to the question ID in the key information.
And S10522, calling the scoring model corresponding to the successfully matched question template, and scoring the answer of the interviewer according to the scoring model.
In the above steps S10521 and S10522, if the similar question engine is not started, it means that the interviewer does not randomly ask questions in the current interview scene, and the interview process is completed by the machine. At the moment, the problem template can be accurately matched according to the key information.
Example four
As shown in fig. 5, the embodiment of the present invention provides an interview apparatus 50 based on an AI-assisted interview scene, comprising:
the text acquisition module 51 is configured to acquire the voice content expressed by the interviewer and convert the voice content into text content;
a word segmentation processing module 52, configured to perform word segmentation processing on the text content to obtain a word segmentation database;
a key information obtaining module 53, configured to obtain key information from the word segmentation database;
a question matching module 54, configured to determine a current interview scene type according to the key information, and select whether to start a similar question engine according to the interview scene type, where the similar question engine is configured to match a question template in a question database according to the word segmentation word vector;
the scoring module 55 is used for acquiring a question template according to the processing result of the similar question engine when the similar question engine is started, selecting a scoring model and scoring the answer of the interviewer;
and when the similar question engine is not started, acquiring a question template according to the key information, and selecting a scoring model to score the answer of the interviewer.
The embodiment of the present invention further provides a terminal device, which includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, where when the processor executes the computer program, each step in the interview method based on the AI-assisted interview scenario as described in the first embodiment is implemented.
An embodiment of the present invention further provides a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium, and when the computer program is executed by a processor, the various steps in the interview method based on the AI-assisted interview scene as described in the first embodiment are implemented.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the foregoing embodiments illustrate the present invention in detail, those of ordinary skill in the art will understand that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.
Claims (9)
1. An interview method based on an AI auxiliary interview scene is characterized by comprising the following steps:
acquiring voice content expressed by an interviewer, and converting the voice content into text content;
performing word segmentation processing on the text content to obtain a word segmentation database;
acquiring key information from the word segmentation database;
detecting the number of short sentences in the word segmentation database and the number of problem IDs of the key information;
if the number of the question IDs of the key information is the same as the number of the short sentences in the participle database, the current interview scene type is a human-computer interview scene, and a similar question engine is not started;
if the number of the question IDs of the key information is different from the number of the short sentences in the participle database, the current interview scene type is a field auxiliary interview scene, and a similar question engine is started;
the similarity engine is used for matching a question template in a question database according to the word segmentation word vectors;
when the similar question engine is started, a question template is obtained according to the processing result of the similar question engine, and a grading model is selected to grade the answer of the interviewee;
and when the similar question engine is not started, acquiring a question template according to the key information, and selecting a scoring model to score the answer of the interviewer.
2. The AI-assisted interview method based on the interview scene as claimed in claim 1, wherein when the similar questions engine is started, the question template is obtained according to the processing result of the similar questions engine, and a scoring model is selected to score the interviewer's answers, comprising:
when an affinity engine is started, calling a trained matching neural network model in the affinity engine, and converting the key information into the word segmentation word vector through the matching neural network model;
matching a question template in the question database according to the word segmentation word vector;
and calling a scoring model corresponding to the successfully matched question template, and scoring the answer of the interviewer according to the scoring model.
3. The AI-assisted interview scenario-based interview method of claim 2, wherein matching the question template in the question database based on the word segmentation vector comprises:
acquiring a template word vector of each question template;
comparing the word segmentation word vector with the template word vector, and calculating cosine similarity;
and judging whether the word segmentation word vector is matched with the problem template or not according to the cosine similarity.
4. The AI-assisted interview method based on the interview scenario of claim 1, wherein the step of obtaining the question template based on the key information and selecting a scoring model to score the interviewer's answers without starting the similar questions engine comprises:
when the similar question engine is not started, matching a question template in the question database according to the question ID in the key information;
and calling a scoring model corresponding to the successfully matched question template, and scoring the answer of the interviewer according to the scoring model.
5. The AI-assisted interview method based on interview scenes of claim 1, wherein the scoring model comprises a regular scoring model and a composite scoring model;
when the interview scene type is a human-computer interview scene, the scoring model comprises a rule scoring model;
and when the interview scene is a scene of field auxiliary interview, the scoring model comprises a comprehensive scoring model.
6. The AI-assisted interview method based on interview scenes as claimed in claim 1, wherein before performing segmentation processing on the text content to obtain a segmentation database, the method comprises:
and carrying out noise information replacement on the text content, and filtering non-key information.
7. An interview device based on an AI assisted interview scene, comprising:
the text acquisition module is used for acquiring the voice content expressed by the interviewer and converting the voice content into text content;
the word segmentation processing module is used for carrying out word segmentation processing on the text content to obtain a word segmentation database;
the key information acquisition module is used for acquiring key information from the word segmentation database;
the question matching module is used for detecting the number of short sentences in the word segmentation database and the number of question IDs of the key information; if the number of the question IDs of the key information is the same as the number of the short sentences in the participle database, the current interview scene type is a human-computer interview scene, and a similar question engine is not started; if the number of the question IDs of the key information is different from the number of the short sentences in the participle database, the current interview scene type is a field auxiliary interview scene, and a similar question engine is started; the similar question engine is used for matching question templates in a question database according to the word segmentation word vectors;
the scoring module is used for acquiring a question template according to the processing result of the similar question engine when the similar question engine is started, selecting a scoring model and scoring the answer of the interviewer; and when the similar question engine is not started, acquiring a question template according to the key information, and selecting a scoring model to score the answer of the interviewer.
8. A terminal device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor, when executing the computer program, carries out the steps of the interview method based on the AI-assisted interview scenario according to any one of claims 1 to 6.
9. A storage medium which is a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps in the AI-assisted interview scenario-based interview method according to any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910504604.2A CN110347787B (en) | 2019-06-12 | 2019-06-12 | Interview method and device based on AI auxiliary interview scene and terminal equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910504604.2A CN110347787B (en) | 2019-06-12 | 2019-06-12 | Interview method and device based on AI auxiliary interview scene and terminal equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110347787A CN110347787A (en) | 2019-10-18 |
| CN110347787B true CN110347787B (en) | 2022-12-09 |
Family
ID=68181789
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910504604.2A Active CN110347787B (en) | 2019-06-12 | 2019-06-12 | Interview method and device based on AI auxiliary interview scene and terminal equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110347787B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111027305B (en) * | 2019-10-29 | 2023-12-19 | 平安科技（深圳）有限公司 | Text matching-based intelligent interview method and device and computer equipment |
| CN111429101A (en) * | 2020-03-30 | 2020-07-17 | 广西通途科技有限公司 | Remote interview system and method based on artificial intelligence |
| CN112466308A (en) * | 2020-11-25 | 2021-03-09 | 北京明略软件系统有限公司 | Auxiliary interviewing method and system based on voice recognition |
| CN114741274A (en) * | 2021-01-08 | 2022-07-12 | 腾讯科技（深圳）有限公司 | Abnormity testing method and device |
| CN113241076A (en) * | 2021-05-12 | 2021-08-10 | 北京字跳网络技术有限公司 | Voice processing method and device and electronic equipment |
| CN113470651A (en) * | 2021-06-04 | 2021-10-01 | 广东外语外贸大学 | Voice scoring method and system based on abstract extraction |
| CN113609864B (en) * | 2021-08-06 | 2022-02-11 | 珠海市鸿瑞信息技术股份有限公司 | Text semantic recognition processing system and method based on industrial control system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104991965A (en) * | 2015-07-23 | 2015-10-21 | 上海智臻网络科技有限公司 | Method and apparatus for creating extended question based on standard question |
| CN106095988A (en) * | 2016-06-21 | 2016-11-09 | 上海智臻智能网络科技股份有限公司 | Automatic question-answering method and device |
| CN106649825A (en) * | 2016-12-29 | 2017-05-10 | 上海智臻智能网络科技股份有限公司 | Voice interaction system, establishment method and device thereof |
| CN107315766A (en) * | 2017-05-16 | 2017-11-03 | 广东电网有限责任公司江门供电局 | A kind of voice response method and its device for gathering intelligence and artificial question and answer |
| CN109492854A (en) * | 2018-09-17 | 2019-03-19 | 平安科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of intelligent robot interview |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10896444B2 (en) * | 2017-01-24 | 2021-01-19 | International Business Machines Corporation | Digital content generation based on user feedback |
-
2019
- 2019-06-12 CN CN201910504604.2A patent/CN110347787B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104991965A (en) * | 2015-07-23 | 2015-10-21 | 上海智臻网络科技有限公司 | Method and apparatus for creating extended question based on standard question |
| CN106095988A (en) * | 2016-06-21 | 2016-11-09 | 上海智臻智能网络科技股份有限公司 | Automatic question-answering method and device |
| CN106649825A (en) * | 2016-12-29 | 2017-05-10 | 上海智臻智能网络科技股份有限公司 | Voice interaction system, establishment method and device thereof |
| CN107315766A (en) * | 2017-05-16 | 2017-11-03 | 广东电网有限责任公司江门供电局 | A kind of voice response method and its device for gathering intelligence and artificial question and answer |
| CN109492854A (en) * | 2018-09-17 | 2019-03-19 | 平安科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of intelligent robot interview |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110347787A (en) | 2019-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110347787B (en) | Interview method and device based on AI auxiliary interview scene and terminal equipment | |
| CN110096570B (en) | Intention identification method and device applied to intelligent customer service robot | |
| CN108304372B (en) | Entity extraction method and device, computer equipment and storage medium | |
| CN108287822B (en) | Chinese similarity problem generation system and method | |
| CN107798140B (en) | Dialog system construction method, semantic controlled response method and device | |
| CN106599032B (en) | Text event extraction method combining sparse coding and structure sensing machine | |
| JP5167546B2 (en) | Sentence search method, sentence search device, computer program, recording medium, and document storage device | |
| CN112069298A (en) | Human-computer interaction method, device and medium based on semantic web and intention recognition | |
| KR20190133931A (en) | Method to response based on sentence paraphrase recognition for a dialog system | |
| CN110414004B (en) | Method and system for extracting core information | |
| CN107797987B (en) | Bi-LSTM-CNN-based mixed corpus named entity identification method | |
| US20060136208A1 (en) | Hybrid apparatus for recognizing answer type | |
| CN107180026B (en) | Event phrase learning method and device based on word embedding semantic mapping | |
| US20230069935A1 (en) | Dialog system answering method based on sentence paraphrase recognition | |
| CN114580382A (en) | Text error correction method and device | |
| Sen et al. | Bangla natural language processing: A comprehensive analysis of classical, machine learning, and deep learning-based methods | |
| CN114020906A (en) | Chinese medical text information matching method and system based on twin neural network | |
| CN112468659A (en) | Quality evaluation method, device, equipment and storage medium applied to telephone customer service | |
| CN114528919A (en) | Natural language processing method and device and computer equipment | |
| CN114818717A (en) | Chinese named entity recognition method and system fusing vocabulary and syntax information | |
| CN115146629A (en) | News text and comment correlation analysis method based on comparative learning | |
| CN112069312A (en) | Text classification method based on entity recognition and electronic device | |
| KR101255957B1 (en) | Method and apparatus for tagging named entity | |
| CN115905187B (en) | Intelligent proposition system oriented to cloud computing engineering technician authentication | |
| Sen et al. | Bangla natural language processing: A comprehensive review of classical machine learning and deep learning based methods |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |