CN111027305A - Intelligent interviewing method and device based on text matching and computer equipment - Google Patents
Intelligent interviewing method and device based on text matching and computer equipment Download PDFInfo
- Publication number
- CN111027305A CN111027305A CN201911037921.4A CN201911037921A CN111027305A CN 111027305 A CN111027305 A CN 111027305A CN 201911037921 A CN201911037921 A CN 201911037921A CN 111027305 A CN111027305 A CN 111027305A
- Authority
- CN
- China
- Prior art keywords
- vector
- interview
- text
- word
- matrix
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 239000013598 vector Substances 0.000 claims abstract description 421
- 238000006243 chemical reaction Methods 0.000 claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims description 216
- 238000012545 processing Methods 0.000 claims description 26
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 12
- 230000009466 transformation Effects 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000004364 calculation method Methods 0.000 description 25
- 230000008569 process Effects 0.000 description 12
- 238000012216 screening Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 238000011176 pooling Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000007115 recruitment Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 241001394244 Planea Species 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Economics (AREA)
- Marketing (AREA)
- Data Mining & Analysis (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Abstract
The application provides an intelligent interviewing method, an intelligent interviewing device, computer equipment and a computer readable storage medium based on text matching, and relates to the field of semantic analysis, wherein the method comprises the following steps: acquiring an interview text and a standard text; respectively carrying out vector conversion on the interview text and the standard text according to a first preset rule to obtain a first vector corresponding to the interview text and a second vector corresponding to the standard text; calculating a similarity between the first vector and the second vector; and matching corresponding interview scores according to the similarity. According to the method and the device, the first vector and the second vector obtained after the interview text and the standard text are processed can show the word meaning to be expressed by the text to the maximum extent, so that the accuracy of text similarity matching on the vector level is greatly improved, and the high accuracy and objectivity of intelligent interview are realized.
Description
Technical Field
The application relates to the technical field of semantic analysis, in particular to an intelligent interview method and device based on text matching and computer equipment.
Background
At present, the post recruitment in various industries generally carries out interview manually, and particularly for the posts with high mobility and high recruitment requirement, a human department needs to spend a great deal of energy and resources to carry out frequent interview, so that the recruitment cost is high, and a great amount of human resources are consumed. Moreover, manual interviewing is affected by subjective consciousness of interviewer individuals, and the unification and objectivity of screening standards of applicants cannot be realized.
Disclosure of Invention
The application mainly aims to provide an intelligent interview method, an intelligent interview device and computer equipment based on text matching, and aims to overcome the defect that the existing interview method is lack of uniformity and objectivity.
In order to achieve the above object, the present application provides an intelligent interview method based on text matching, comprising:
acquiring an interview text and a standard text, wherein the interview text is formed after an interviewer answers to interview questions, and the standard text is a text of a standard answer corresponding to the interview questions;
respectively carrying out vector conversion on the interview text and the standard text according to a first preset rule to obtain a first vector corresponding to the interview text and a second vector corresponding to the standard text;
calculating a similarity between the first vector and the second vector;
and matching corresponding interview scores according to the similarity.
Further, the step of performing vector transformation on the interview text according to a first preset rule to obtain a first vector corresponding to the interview text includes:
carrying out complex value embedding on each word in the interview text to obtain a word matrix consisting of complex value vectors corresponding to the words respectively;
converting the word matrix into a mixed density matrix by adopting a sliding window;
calculating according to a preset algorithm to obtain a plurality of probability vectors respectively corresponding to the mixed density matrix in different sliding windows;
and respectively selecting the largest probability vector in each sliding window, wherein each largest probability vector forms the first vector.
Further, the step of converting the word matrix into a mixed density matrix using a sliding window includes:
sequentially and progressively selecting a preset number of first complex value vectors from the word matrix by adopting the sliding window to form a matrix according to the arrangement sequence of the words corresponding to the complex value vectors in the interview text, until the selection of all the complex value vectors is completed, and obtaining a plurality of first word matrices;
respectively calculating the outer product of each first complex value vector in each first word matrix and the vector of the conjugate device corresponding to each first complex value vector, converting each first word matrix into a corresponding word density matrix, and calculating the first probability corresponding to each first complex value vector;
and calculating to obtain a mixed density matrix according to each word density matrix and each first probability.
Further, the step of calculating the first probability corresponding to each of the first complex-valued vectors includes:
respectively substituting each first complex value vector into a first preset formula, and calculating to obtain a norm corresponding to each first complex value vector, wherein the first preset formula is as follows:π(wi) Is the norm, x is the value of the first complex-valued vector;
substituting each norm into a second preset formula respectively, and calculating to obtain the first probability corresponding to each norm, wherein the second preset formula is as follows:p(wi) And e is a natural base number, l represents total l w, and j represents the jth w.
Further, the step of calculating a mixed density matrix according to the word density matrix and each of the first probabilities includes:
multiplying the vectors in the word density matrixes by the corresponding first probability respectively to obtain corresponding weighted word density matrixes respectively;
and adding the weighted word density matrixes to obtain the mixed density matrix.
Further, the step of calculating a plurality of probability vectors corresponding to the mixed density matrix in different sliding windows according to a preset algorithm includes:
substituting the mixed density matrix into a third preset formula, and calculating to obtain a plurality of second probabilities, wherein the third preset formula is as follows: : p is a radical ofx(p)=<xi|ρ|xi>=tr(ρ|xi><xi| x) where | xi>The initial value is the orthogonal one-hot coded vector expressed by Dirac symbol, | xi><xiIf is | xi>Outer product of px(p) is a second probability, i represents the ith projection plane;
and combining the second probabilities to obtain the probability vector.
Further, the step of calculating the similarity between the first vector and the second vector includes:
calculating a cosine value between the first vector and the second vector;
and taking the cosine value as the similarity.
The application also provides an intelligent interview device based on text matching, including:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring an interview text and a standard text, the interview text is formed after an interviewer answers to interview questions, and the standard text is a text of a standard answer corresponding to the interview questions;
the conversion module is used for respectively carrying out vector conversion on the interview text and the standard text according to a first preset rule to obtain a first vector corresponding to the interview text and a second vector corresponding to the standard text;
a calculating module, configured to calculate a similarity between the first vector and the second vector;
and the matching module is used for matching the corresponding interview scores according to the similarity.
Further, the conversion module comprises:
the embedding submodule is used for carrying out complex value embedding on each word in the interview text to obtain a word matrix formed by complex value vectors corresponding to the words respectively;
the conversion submodule is used for converting the word matrix into a mixed density matrix by adopting a sliding window;
the first calculation submodule is used for calculating according to a preset algorithm to obtain a plurality of probability vectors respectively corresponding to the mixed density matrix in different sliding windows;
and the selection submodule is used for respectively selecting the maximum probability vector in each sliding window, and each maximum probability vector forms the first vector.
Further, the transformation module comprises:
the selection unit is used for sequentially selecting a preset number of first complex-value vectors from the word matrix by adopting the sliding window to form a matrix according to the arrangement sequence of the words corresponding to the complex-value vectors in the interview text, until the selection of all the complex-value vectors is completed, and obtaining a plurality of first word matrices;
a first calculating unit, configured to calculate an outer product of each first complex-valued vector in each first word matrix and a corresponding conjugate device vector, convert each first word matrix into a corresponding word density matrix, and calculate a first probability corresponding to each first complex-valued vector;
and the second calculation unit is used for calculating to obtain a mixed density matrix according to each word density matrix and each first probability.
Further, the first computing unit includes:
a first calculating subunit, configured to substitute each of the first complex-valued vectors into a first preset formula, and calculate to obtain a norm corresponding to each of the first complex-valued vectors, where the first preset formula is:π(wi) Is the norm, x is the value of the first complex-valued vector;
a second calculating subunit, configured to substitute each norm into a second preset formula, and calculate to obtain the first probability corresponding to each norm, where the second preset formula is:p(wi) And e is a natural base number, l represents total l w, and j represents the jth w.
Further, the second calculation unit includes:
the third calculation subunit is configured to multiply the vector in each word density matrix by the corresponding first probability to obtain a corresponding weighted word density matrix;
and the combination subunit is used for adding the weighted word density matrixes to obtain the mixed density matrix.
Further, the computation submodule includes:
a third calculating unit, configured to substitute the mixed density matrix into a third preset formula, and calculate to obtain a plurality of second probabilities, where the third preset formula is: : p is a radical ofx(p)=<xi|ρ|xi>=tr(ρ|xi><xi| x) where | xi>The initial value is the orthogonal one-hot coded vector expressed by Dirac symbol, | xi><xiIf is | xi>Outer product of px(p) is a second probability, i represents the ith projection plane;
and the combination unit is used for combining the second probabilities to obtain the probability vector.
Further, the calculation module includes:
a second calculation submodule for calculating cosine values between the first vector and the second vector;
and the marking module is used for taking the cosine value as the similarity.
The present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.
The present application also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, carries out the steps of the method of any of the above.
According to the intelligent interview method, the intelligent interview device and the intelligent interview computer equipment based on the text matching, interview texts made by an applicant and pre-entered standard texts are converted into corresponding complex-value vectors, then the complex-value vectors are correspondingly calculated to obtain mixed density matrixes respectively corresponding to the interview texts and the standard texts, maximum probability vectors are selected from the mixed density matrixes of all sliding windows to form corresponding first vectors and second vectors, then the similarity between the interview texts and the standard texts is obtained by calculating the cosine between the first vectors and the second vectors, and finally corresponding interview scores are obtained according to the similarity matching. According to the method and the device, the first vector and the second vector obtained after the interview text and the standard text are processed can show the word meaning to be expressed by the text to the maximum extent, so that the accuracy of text similarity matching on the vector level is greatly improved, and the high accuracy and objectivity of intelligent interview are realized.
Drawings
FIG. 1 is a schematic diagram illustrating steps of an intelligent interview method based on text matching according to an embodiment of the present application;
FIG. 2 is a block diagram of an overall structure of an intelligent interview apparatus based on text matching according to an embodiment of the present application;
fig. 3 is a block diagram schematically illustrating a structure of a computer device according to an embodiment of the present application.
The implementation, functional features and advantages of the objectives of the present application will be further explained with reference to the accompanying drawings.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
Referring to fig. 1, an embodiment of the present application provides an intelligent interview method based on text matching, including:
s1: acquiring an interview text and a standard text, wherein the interview text is formed after an interviewer answers to interview questions, and the standard text is a text of a standard answer corresponding to the interview questions;
s2: respectively carrying out vector conversion on the interview text and the standard text according to a first preset rule to obtain a first vector corresponding to the interview text and a second vector corresponding to the standard text;
s3: calculating a similarity between the first vector and the second vector;
s4: and matching corresponding interview scores according to the similarity.
In this embodiment, the interview system can sequentially output interview questions to interviewers for answering according to the preset setting, and the interviewers can answer the interview questions through voice or manual input and other modes. After collecting the answers of the interviewer to the interview questions, the interview system forms interview texts and calls the pre-recorded standard answers corresponding to the interview questions, namely the standard texts. The interview system needs to convert the interview text and the standard text into corresponding vectors so as to compare the similarity between the interview text and the standard text in the following process, wherein the interview system converts the interview text and the standard text into the corresponding vectors in the same method, and the conversion actions of the interview text and the standard text are not limited in sequence, for example, the interview text and the standard text can be converted at the same time, or the standard text and the standard text can be converted first and then the conversion actions of the interview textThe interview text is converted, even the standard text can be converted in advance before an interviewer conducts interview and then is recorded in the interview system for storage, so that the interview system only needs to process the interview text in the interview process, and the processing efficiency is improved. In this embodiment, a vector conversion example of an interview text is specifically described, first, an interview system applies unique hot coding and performs Complex-valued Embedding (Complex-v affected Embedding) on each word in the interview text, so that each word generates a corresponding Complex-valued vector, and each Complex-valued vector is combined to form a word matrix. The interview system adopts a sliding window for the word matrix, and sequentially selects a preset number of complex-value vectors as first complex-value vectors and forms the first word matrix from the word matrix each time according to the arrangement sequence of the words corresponding to the complex-value vectors in the interview text, until the selection of all the complex-value vectors is completed. For example, if the complex-valued vector in the word matrix is (a, b, c, d, e) and the preset number is 3, the first word matrix obtained by selection is: (a, b, c), (b, c, d), (c, d, e). The interview system needs to perform the same data processing on the first word matrix obtained by screening each time, and the specific processing steps are as follows: the interview system multiplies the first complex-valued vectors by the corresponding conjugate transpose vectors to obtain outer products, so that a first word matrix formed by the outer products of the first complex-valued vectors is converted into a word density matrix. And the interview system substitutes each first complex value vector in the first word matrix into a first preset formula respectively to calculate the norm corresponding to each first complex value vector. Then, the interview system substitutes the norms corresponding to the first complex value vectors into a second preset formula respectively, and first probabilities corresponding to the norms are obtained through calculation. After the interview system finishes the data processing, the vectors in the word density matrix are multiplied by the corresponding first probabilities respectively, and the corresponding weighted word density matrix formed by the calculated vectors is obtained. And the interview system adds the weighted word density matrixes to obtain a mixed density matrix. The interview system calls a third preset formula and converts the projection length of the mixed density matrix on each different projection plane into corresponding second probability so as to obtain the mixed density matrix which is composed of a plurality of second probabilities projected on each projection planeA probability vector. Wherein, the third preset formula is: p is a radical ofx(p)=〈xi|ρ|xi>=tr(ρ|xi><xi| x) where | xi>For word vectors represented by dirac symbols, | xi><xiIf is | xi>Outer product of px(p) is a second probability, the plurality of probabilities constituting a probability vector, i representing the ith projection plane. The interview system extracts the first vector from the largest probability vector of the plurality of probability vectors in each sliding window through a pooling operation, for example, if the probability vector of the sliding window a is (1,2,3), the probability vector of the sliding window B is (4,5,6), and the probability vector of the sliding window C is (7,8,9), then 3,6,9 are selected from the sliding window A, B, C to form the first vector (3,6, 9). And the interview system converts the standard text into a corresponding second vector according to the same conversion method. And the interview system calculates a cosine value between the first vector and the second vector according to the angle between the first vector and the second vector, and the calculated cosine value is the similarity between the interview text and the standard text. The interview system is pre-recorded with a mapping relation table of the similarity and the interview score, so that the interview system can obtain the evaluation score of the interviewer in the interview question according to the similarity obtained by the current calculation.
Further, the step of performing vector transformation on the interview text according to a first preset rule to obtain a first vector corresponding to the interview text includes:
s21: carrying out complex value embedding on each word in the interview text to obtain a word matrix consisting of complex value vectors corresponding to the words respectively;
s22: converting the word matrix into a mixed density matrix by adopting a sliding window;
s23: calculating according to a preset algorithm to obtain a plurality of probability vectors respectively corresponding to the mixed density matrix in different sliding windows;
s24: and respectively selecting the largest probability vector in each sliding window, wherein each largest probability vector forms the first vector.
In this embodiment, the interview system is usedAnd coding by using the one-hot code and embedding the complex value of each word in the interview text to ensure that each word generates a corresponding complex value vector, and combining the complex value vectors to form a word matrix. The complex vector in this embodiment is expressed as z ═ r (cos θ + i sin θ), which is advantageous in that the complex vector can make the word vector express more implicit word senses than the conventional real vector. The word vector expressed by the complex value vector only considers the common amplitude addition and subtraction, and also considers higher-order semantics brought by phase information of the word vector, so that the effect that two words have more word senses when added can be achieved, and the effect that the two words are added to generate a reaction can also be realized. The interview system adopts a sliding window for a word matrix, sequentially and progressively selects a preset number of complex value vectors as first complex value vectors and forms the first word matrix from the word matrix each time according to the arrangement sequence of words corresponding to the complex value vectors in an interview text, and the first word matrix is formed until the selection of all the complex value vectors is completed, for example, if the complex value vectors in the word matrix are (a, b, c, d, e) and the preset number is 3, the first word matrix obtained by selection is respectively: (a, b, c), (b, c, d), (c, d, e). The interview system needs to perform the same data processing on the first word matrix obtained by screening each time, and the specific processing steps are as follows: the interview system respectively leads each first complex-valued vector a and the corresponding conjugate transpose vector aTThe multiplication results in an outer product, so that a first word matrix consisting of the outer products of the first complex-valued vectors is converted into a word density matrix. And the interview system substitutes each first complex value vector in the first word matrix into a first preset formula respectively to calculate the norm corresponding to each first complex value vector. Then, the interview system substitutes the norms corresponding to the first complex value vectors into a second preset formula respectively, and the first probabilities corresponding to the norms are obtained through calculation. After the interview system finishes the data processing, the vectors in the word density matrix are multiplied by the corresponding first probabilities respectively, and the corresponding weighted word density matrix formed by the calculated vectors is obtained. And the interview system adds the weighted word density matrixes to obtain a mixed density matrix. The interview system calls a third preset formula and converts the projection lengths of the mixed density matrix on different projection planesAnd obtaining a probability vector composed of a plurality of second probabilities of the mixed density matrix projected on each projection plane. The interviewing system respectively extracts the maximum probability vector from the probability vectors of each sliding window to form a first vector through a pooling operation.
Further, the step of converting the word matrix into a mixed density matrix using a sliding window includes:
s221: sequentially and progressively selecting a preset number of first complex value vectors from the word matrix by adopting the sliding window to form a matrix according to the arrangement sequence of the words corresponding to the complex value vectors in the interview text, until the selection of all the complex value vectors is completed, and obtaining a plurality of first word matrices;
s222: respectively calculating the outer product of each first complex value vector in each first word matrix and the vector of the conjugate device corresponding to each first complex value vector, converting each first word matrix into a corresponding word density matrix, and calculating the first probability corresponding to each first complex value vector;
s223: and calculating to obtain a mixed density matrix according to each word density matrix and each first probability.
In this embodiment, the interview system uses a sliding window, and sequentially selects a preset number of complex-valued vectors as first complex-valued vectors and forms a first word matrix from the word matrix each time according to the arrangement sequence of the words corresponding to the complex-valued vectors in the interview text, until the selection of all the complex-valued vectors is completed, for example, if the complex-valued vectors in the word matrix are (a, b, c, d, e), and the preset number is 3, the first word matrix obtained by selection is: (a, b, c), (b, c, d), (c, d, e). The interview system needs to perform the same data processing on the first word matrix obtained by screening each time, and the specific processing steps are as follows: the interview system respectively leads each first complex-valued vector a and the corresponding conjugate transpose vector aTThe multiplication results in an outer product, so that a first word matrix consisting of the outer products of the first complex-valued vectors is converted into a word density matrix. And the interview system substitutes each first complex value vector in the first word matrix into a first preset formula respectively,and calculating to obtain norms corresponding to the first complex value vectors respectively. Then, the interview system substitutes the norms corresponding to the first complex value vectors into a second preset formula respectively, and first probabilities corresponding to the norms are obtained through calculation. After the interview system finishes the data processing, the vectors in the word density matrix are multiplied by the corresponding first probabilities respectively, and the corresponding weighted word density matrix formed by the calculated vectors is obtained. And the interview system adds the weighted word density matrixes to obtain a mixed density matrix. In application, compared with the ordinary average weighting, the embodiment weights according to the local words, so that the system has different weights for different words and can make judgment on the words according to the context.
Further, the step of calculating the first probability corresponding to each of the first complex-valued vectors includes:
s2221: respectively substituting each first complex value vector into a first preset formula, and calculating to obtain a norm corresponding to each first complex value vector, wherein the first preset formula is as follows:π(wi) Is the norm, x is the value of the first complex-valued vector;
s2222: substituting each norm into a second preset formula respectively, and calculating to obtain the first probability corresponding to each norm, wherein the second preset formula is as follows:p(wi) And e is a natural base number, l represents total l w, and j represents the jth w.
In this embodiment, the interview system substitutes each first complex value vector in the first word matrix into a first preset formula respectively, and calculates to obtain a norm corresponding to each first complex value vector, where the second preset formula is:π(wi) To needAnd (5) obtaining a norm, wherein x is the value of the first complex-valued vector. Then, the interview system substitutes the norms corresponding to the first complex value vectors into a second preset formula respectively, and calculates to obtain first probabilities corresponding to the norms respectively, wherein the second preset formula is as follows:p(wi) For the first probability, e is a natural base, l denotes a total of l w, and j denotes the jth w.
Further, the step of calculating a mixed density matrix according to the word density matrix and each of the first probabilities includes:
s2231: multiplying the vectors in the word density matrixes by the corresponding first probability respectively to obtain corresponding weighted word density matrixes respectively;
s2232: and adding the weighted word density matrixes to obtain the mixed density matrix.
In this embodiment, the interview system multiplies the vectors in the word density matrix by the corresponding first probabilities to obtain the corresponding weighted word density matrices formed by the calculated vectors. And the interview system adds the density matrixes of the weighted words to obtain a mixed density matrix. In application, compared with the ordinary average weighting, the embodiment weights according to the local words, so that the system has different weights for different words and can make judgment on the words according to the context.
Further, the step of calculating a plurality of probability vectors corresponding to the mixed density matrix in different sliding windows according to a preset algorithm includes:
s231: substituting the mixed density matrix into a third preset formula, and calculating to obtain a plurality of second probabilities, wherein the third preset formula is as follows: : p is a radical ofx(p)=<xi|ρ|xi>=tr(ρ|xi><xi| x) where | xi>The initial value is the orthogonal one-hot coded vector expressed by Dirac symbol, | xi><xiIf is | xi>Outer product of px(p) isTwo probabilities, i represents the ith projection plane;
s232: and combining the second probabilities to obtain the probability vector.
In this embodiment, the interview system calls a third preset formula, converts the projection length of the mixed density matrix on the projection plane into corresponding second probabilities, and the interview system combines the obtained second probabilities to obtain the probability vector of the mixed density matrix on the projection plane corresponding to the current sliding window. Wherein, the third preset formula is: p is a radical ofx(p)=〈xi|ρ|xi>=tr(ρ|xi〉〈xi| x) where | x>The initial value is an orthogonal one-hot coded vector represented by a Dirac symbol, and the length is always a unit length, | xi><xiIf is | xi>For a projection plane trained as a high-latitude feature for extracting a mixed density matrix, px(p) is the second probability.
Further, the step of calculating the similarity between the first vector and the second vector includes:
s31: calculating a cosine value between the first vector and the second vector;
s32: and taking the cosine value as the similarity.
In this embodiment, the first vector for conversion of the interview text and the second vector for conversion of the standard text are vectors obtained after the respective N-grams of the two texts are projected on the same projection plane, so that the two vectors can have a basis for calculation. The interview system calculates the cosine of the angle between the first vector and the second vector, and the cosine value obtained through calculation can be used as the similarity between the two texts. For example, a and b are two vectors, the cosine between the two vectors is: cos (θ) ═ a × b/(| a | × | b |).
The intelligent interviewing method based on text matching provided by the embodiment of the invention includes the steps of converting an interviewing text made by an applicant and a pre-entered standard text into corresponding complex-value vectors, then correspondingly calculating the complex-value vectors to obtain mixed density matrices corresponding to the interviewing text and the standard text respectively, selecting the maximum probability vectors from the mixed density matrices of sliding windows to form a first vector and a second vector corresponding to each other, then calculating the cosine between the first vector and the second vector to obtain the similarity between the interviewing text and the standard text, and finally obtaining a corresponding interviewing score according to similarity matching. According to the method and the device, the first vector and the second vector obtained after the interview text and the standard text are processed can show the word meaning to be expressed by the text to the maximum extent, so that the accuracy of text similarity matching on the vector level is greatly improved, and the high accuracy and objectivity of intelligent interview are realized.
Referring to fig. 2, an embodiment of the present application further provides an intelligent interview apparatus based on text matching, including:
the system comprises an acquisition module 1, a processing module and a processing module, wherein the acquisition module is used for acquiring an interview text and a standard text, the interview text is formed after an interviewer answers to interview questions, and the standard text is a text of a standard answer corresponding to the interview questions;
the conversion module 2 is used for respectively performing vector conversion on the interview text and the standard text according to a first preset rule to obtain a first vector corresponding to the interview text and a second vector corresponding to the standard text;
a calculating module 3, configured to calculate a similarity between the first vector and the second vector;
and the matching module 4 is used for matching the corresponding interview scores according to the similarity.
In this embodiment, the interview system can sequentially output interview questions to interviewers for answering according to the preset setting, and the interviewers can answer the interview questions through voice or manual input and other modes. After collecting the answers of the interviewer to the interview questions, the interview system forms interview texts and calls the pre-recorded standard answers corresponding to the interview questions, namely the standard texts. The interview system needs to convert the interview text and the standard text into corresponding vectors so as to compare the similarity between the interview text and the standard text in the following process, wherein the interview system converts the interview text and the standard text into the corresponding vectorsThe conversion actions of the two are not limited in sequence, for example, the two can be converted simultaneously, the standard text can be converted first, then the interview text is converted, even the standard text can be converted in advance before the interviewer performs interview and then recorded in the interview system for storage, so that the interview system only needs to process the interview text in the interview process, and the processing efficiency is improved. In this embodiment, a vector conversion example of an interview text is specifically described, first, an interview system applies unique hot coding and performs Complex-valued Embedding (Complex-valued Embedding) on each word in the interview text, so that each word generates a corresponding Complex-valued vector, and each Complex-valued vector is combined to form a word matrix. The interview system adopts a sliding window for the word matrix, and sequentially selects a preset number of complex-value vectors as first complex-value vectors and forms the first word matrix from the word matrix each time according to the arrangement sequence of the words corresponding to the complex-value vectors in the interview text, until the selection of all the complex-value vectors is completed. For example, if the complex-valued vector in the word matrix is (a, b, c, d, e) and the preset number is 3, the first word matrix obtained by selection is: (a, b, c), (b, c, d), (c, d, e). The interview system needs to perform the same data processing on the first word matrix obtained by screening each time, and the specific processing steps are as follows: the interview system multiplies the first complex-valued vectors by the corresponding conjugate transpose vectors to obtain outer products, so that a first word matrix formed by the outer products of the first complex-valued vectors is converted into a word density matrix. And the interview system substitutes each first complex value vector in the first word matrix into a first preset formula respectively to calculate the norm corresponding to each first complex value vector. Then, the interview system substitutes the norms corresponding to the first complex value vectors into a second preset formula respectively, and first probabilities corresponding to the norms are obtained through calculation. After the interview system finishes the data processing, the vectors in the word density matrix are multiplied by the corresponding first probabilities respectively, and the corresponding weighted word density matrix formed by the calculated vectors is obtained. And the interview system adds the weighted word density matrixes to obtain a mixed density matrix. The interview system calls a third preset formula and obtains the projection lengths of the mixed density matrix on different projection planesAnd converting the first probability into corresponding second probabilities, thereby obtaining a probability vector consisting of a plurality of second probabilities of the mixed density matrix projected on each projection plane. Wherein, the third preset formula is: p is a radical ofx(p)=<xi|ρ|xi>=tr(ρ|xi><xi| x) where | xi>For word vectors represented by dirac symbols, | xi><xiIf is | xi>Outer product of px(p) is a second probability, the plurality of probabilities constituting a probability vector, i representing the ith projection plane. The interview system extracts the first vector from the largest probability vector of the plurality of probability vectors in each sliding window through a pooling operation, for example, if the probability vector of the sliding window a is (1,2,3), the probability vector of the sliding window B is (4,5,6), and the probability vector of the sliding window C is (7,8,9), then 3,6,9 are selected from the sliding window A, B, C to form the first vector (3,6, 9). And the interview system converts the standard text into a corresponding second vector according to the same conversion method. And the interview system calculates a cosine value between the first vector and the second vector according to the angle between the first vector and the second vector, and the calculated cosine value is the similarity between the interview text and the standard text. The interview system is pre-recorded with a mapping relation table of the similarity and the interview score, so that the interview system can obtain the evaluation score of the interviewer in the interview question according to the similarity obtained by the current calculation.
Further, the conversion module 2 includes:
the embedding submodule is used for carrying out complex value embedding on each word in the interview text to obtain a word matrix formed by complex value vectors corresponding to the words respectively;
the conversion submodule is used for converting the word matrix into a mixed density matrix by adopting a sliding window;
the first calculation submodule is used for calculating according to a preset algorithm to obtain a plurality of probability vectors respectively corresponding to the mixed density matrix in different sliding windows;
and the selection submodule is used for respectively selecting the maximum probability vector in each sliding window, and each maximum probability vector forms the first vector.
In this embodiment, the interview system uses unique hot code encoding and performs complex-valued embedding on each word in the interview text, so that each word generates a corresponding complex-valued vector, and each complex-valued vector is combined to form a word matrix. The complex vector in this embodiment is expressed as z ═ r (cos θ + i sin θ), which is advantageous in that the complex vector can make the word vector express more implicit word senses than the conventional real vector. The word vector expressed by the complex value vector only considers the common amplitude addition and subtraction, and also considers higher-order semantics brought by phase information of the word vector, so that the effect that two words have more word senses when added can be achieved, and the effect that the two words are added to generate a reaction can also be realized. The interview system adopts a sliding window for a word matrix, sequentially and progressively selects a preset number of complex value vectors as first complex value vectors and forms the first word matrix from the word matrix each time according to the arrangement sequence of words corresponding to the complex value vectors in an interview text, and the first word matrix is formed until the selection of all the complex value vectors is completed, for example, if the complex value vectors in the word matrix are (a, b, c, d, e) and the preset number is 3, the first word matrix obtained by selection is respectively: (a, b, c), (b, c, d), (c, d, e). The interview system needs to perform the same data processing on the first word matrix obtained by screening each time, and the specific processing steps are as follows: the interview system respectively leads each first complex-valued vector a and the corresponding conjugate transpose vector aTThe multiplication results in an outer product, so that a first word matrix consisting of the outer products of the first complex-valued vectors is converted into a word density matrix. And the interview system substitutes each first complex value vector in the first word matrix into a first preset formula respectively to calculate the norm corresponding to each first complex value vector. Then, the interview system substitutes the norms corresponding to the first complex value vectors into a second preset formula respectively, and the first probabilities corresponding to the norms are obtained through calculation. After the interview system finishes the data processing, the vectors in the word density matrix are multiplied by the corresponding first probabilities respectively, and the corresponding weighted word density matrix formed by the calculated vectors is obtained. And the interview system adds the weighted word density matrixes to obtain a mixed density matrix. Interview system callAnd a third preset formula is used for converting the projection lengths of the mixed density matrix on different projection planes into corresponding second probabilities, so that a probability vector consisting of a plurality of second probabilities of the mixed density matrix projected on the projection planes is obtained. The interviewing system respectively extracts the maximum probability vector from the probability vectors of each sliding window to form a first vector through a pooling operation.
Further, the transformation module comprises:
the selection unit is used for sequentially selecting a preset number of first complex-value vectors from the word matrix by adopting the sliding window to form a matrix according to the arrangement sequence of the words corresponding to the complex-value vectors in the interview text, until the selection of all the complex-value vectors is completed, and obtaining a plurality of first word matrices;
a first calculating unit, configured to calculate an outer product of each first complex-valued vector in each first word matrix and a corresponding conjugate device vector, convert each first word matrix into a corresponding word density matrix, and calculate a first probability corresponding to each first complex-valued vector;
and the second calculation unit is used for calculating to obtain a mixed density matrix according to each word density matrix and each first probability.
In this embodiment, the interview system uses a sliding window, and sequentially selects a preset number of complex-valued vectors as first complex-valued vectors and forms a first word matrix from the word matrix each time according to the arrangement sequence of the words corresponding to the complex-valued vectors in the interview text, until the selection of all the complex-valued vectors is completed, for example, if the complex-valued vectors in the word matrix are (a, b, c, d, e), and the preset number is 3, the first word matrix obtained by selection is: (a, b, c), (b, c, d), (c, d, e). The interview system needs to perform the same data processing on the first word matrix obtained by screening each time, and the specific processing steps are as follows: the interview system respectively leads each first complex-valued vector a and the corresponding conjugate transpose vector aTThe multiplication results in an outer product, so that a first word matrix consisting of the outer products of the first complex-valued vectors is converted into a word density matrix. And, the interview systemAnd the system substitutes each first complex value vector in the first word matrix into a first preset formula respectively, and the norm corresponding to each first complex value vector is obtained through calculation. Then, the interview system substitutes the norms corresponding to the first complex value vectors into a second preset formula respectively, and first probabilities corresponding to the norms are obtained through calculation. After the interview system finishes the data processing, the vectors in the word density matrix are multiplied by the corresponding first probabilities respectively, and the corresponding weighted word density matrix formed by the calculated vectors is obtained. And the interview system adds the weighted word density matrixes to obtain a mixed density matrix. In application, compared with the ordinary average weighting, the embodiment weights according to the local words, so that the system has different weights for different words and can make judgment on the words according to the context.
Further, the first computing unit includes:
a first calculating subunit, configured to substitute each of the first complex-valued vectors into a first preset formula, and calculate to obtain a norm corresponding to each of the first complex-valued vectors, where the first preset formula is:π(wi) Is the norm, x is the value of the first complex-valued vector;
a second calculating subunit, configured to substitute each norm into a second preset formula, and calculate to obtain the first probability corresponding to each norm, where the second preset formula is:p(wi) And e is a natural base number, l represents total l w, and j represents the jth w.
In this embodiment, the interview system substitutes each first complex value vector in the first word matrix into a first preset formula respectively, and calculates to obtain a norm corresponding to each first complex value vector, where the second preset formula is:π(wi) X is the value of the first complex-valued vector for the norm to be found. Then, the interview system substitutes the norms corresponding to the first complex value vectors into a second preset formula respectively, and calculates to obtain first probabilities corresponding to the norms respectively, wherein the second preset formula is as follows:p(wi) For the first probability, e is a natural base, l denotes a total of l w, and j denotes the jth w.
Further, the second calculation unit includes:
the third calculation subunit is configured to multiply the vector in each word density matrix by the corresponding first probability to obtain a corresponding weighted word density matrix;
and the combination subunit is used for adding the weighted word density matrixes to obtain the mixed density matrix.
In this embodiment, the interview system multiplies the vectors in the word density matrix by the corresponding first probabilities to obtain the corresponding weighted word density matrices formed by the calculated vectors. And the interview system adds the density matrixes of the weighted words to obtain a mixed density matrix. In application, compared with the ordinary average weighting, the embodiment weights according to the local words, so that the system has different weights for different words and can make judgment on the words according to the context.
Further, the computation submodule includes:
a third calculating unit, configured to substitute the mixed density matrix into a third preset formula, and calculate to obtain a plurality of second probabilities, where the third preset formula is: : p is a radical ofx(p)=〈xi|ρ|xi>=tr(ρ|xi><xi| x) where | xi>The initial value is the orthogonal one-hot coded vector expressed by Dirac symbol, | xi><xiIf is | xi>Outer product of px(p) is a second probability, i represents the ith projection plane;
and the combination unit is used for combining the second probabilities to obtain the probability vector.
In this embodiment, the interview system calls a third preset formula, converts the projection length of the mixed density matrix on the projection plane into corresponding second probabilities, and the interview system combines the obtained second probabilities to obtain the probability vector of the mixed density matrix on the projection plane corresponding to the current sliding window. Wherein, the third preset formula is: p is a radical ofx(p)=<xi|ρ|xi>=tr(ρ|xi><xi| x) where | x>The initial value is an orthogonal one-hot coded vector represented by a Dirac symbol, and the length is always a unit length, | xi><xiIf is | xi>For a projection plane trained as a high-latitude feature for extracting a mixed density matrix, px(p) is the second probability.
Further, the calculating module 3 includes:
a second calculation submodule for calculating cosine values between the first vector and the second vector;
and the marking module is used for taking the cosine value as the similarity.
In this embodiment, the first vector for conversion of the interview text and the second vector for conversion of the standard text are vectors obtained after the respective N-grams of the two texts are projected on the same projection plane, so that the two vectors can have a basis for calculation. The interview system calculates the cosine of the angle between the first vector and the second vector, and the cosine value obtained through calculation can be used as the similarity between the two texts. For example, a and b are two vectors, the cosine between the two vectors is: cos (θ) ═ a × b/(| a | × | b |).
According to the intelligent interview device based on text matching, the interview text made by an applicant and the pre-entered standard text are converted into corresponding complex-value vectors, then the complex-value vectors are correspondingly calculated to obtain the mixed density matrixes respectively corresponding to the interview text and the standard text, the maximum probability vectors are selected from the mixed density matrixes of the sliding windows to form the first vector and the second vector respectively corresponding to the interview text and the standard text, then the cosine between the first vector and the second vector is calculated to obtain the similarity between the interview text and the standard text, and finally the corresponding interview score is obtained according to the similarity matching. According to the method and the device, the first vector and the second vector obtained after the interview text and the standard text are processed can show the word meaning to be expressed by the text to the maximum extent, so that the accuracy of text similarity matching on the vector level is greatly improved, and the high accuracy and objectivity of intelligent interview are realized.
Referring to fig. 3, a computer device, which may be a server and whose internal structure may be as shown in fig. 3, is also provided in the embodiment of the present application. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the computer designed processor is used to provide computational and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing data such as standard texts. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement an intelligent interview method based on text matching.
The processor executes the intelligent interview method based on text matching, and the steps are as follows:
s1: acquiring an interview text and a standard text, wherein the interview text is formed after an interviewer answers to interview questions, and the standard text is a text of a standard answer corresponding to the interview questions;
s2: respectively carrying out vector conversion on the interview text and the standard text according to a first preset rule to obtain a first vector corresponding to the interview text and a second vector corresponding to the standard text;
s3: calculating a similarity between the first vector and the second vector;
s4: and matching corresponding interview scores according to the similarity.
Further, the step of performing vector transformation on the interview text according to a first preset rule to obtain a first vector corresponding to the interview text includes:
s21: carrying out complex value embedding on each word in the interview text to obtain a word matrix consisting of complex value vectors corresponding to the words respectively;
s22: converting the word matrix into a mixed density matrix by adopting a sliding window;
s23: calculating according to a preset algorithm to obtain a plurality of probability vectors respectively corresponding to the mixed density matrix in different sliding windows;
s24: and respectively selecting the largest probability vector in each sliding window, wherein each largest probability vector forms the first vector.
Further, the step of converting the word matrix into a mixed density matrix using a sliding window includes:
s221: sequentially and progressively selecting a preset number of first complex value vectors from the word matrix by adopting the sliding window to form a matrix according to the arrangement sequence of the words corresponding to the complex value vectors in the interview text, until the selection of all the complex value vectors is completed, and obtaining a plurality of first word matrices;
s222: respectively calculating the outer product of each first complex value vector in each first word matrix and the vector of the conjugate device corresponding to each first complex value vector, converting each first word matrix into a corresponding word density matrix, and calculating the first probability corresponding to each first complex value vector;
s223: and calculating to obtain a mixed density matrix according to each word density matrix and each first probability.
Further, the step of calculating the first probability corresponding to each of the first complex-valued vectors includes:
s2221: respectively substituting each first complex value vector into a first preset publicIn the formula, the norm corresponding to each first complex value vector is obtained through calculation, wherein the first preset formula is as follows:π(wi) Is the norm, x is the value of the first complex-valued vector;
s2222: substituting each norm into a second preset formula respectively, and calculating to obtain the first probability corresponding to each norm, wherein the second preset formula is as follows:p(wi) And e is a natural base number, l represents total l w, and j represents the jth w.
Further, the step of calculating a mixed density matrix according to the word density matrix and each of the first probabilities includes:
s2231: multiplying the vectors in the word density matrixes by the corresponding first probability respectively to obtain corresponding weighted word density matrixes respectively;
s2232: and adding the weighted word density matrixes to obtain the mixed density matrix.
Further, the step of calculating a plurality of probability vectors corresponding to the mixed density matrix in different sliding windows according to a preset algorithm includes:
s231: substituting the mixed density matrix into a third preset formula, and calculating to obtain a plurality of second probabilities, wherein the third preset formula is as follows: : p is a radical ofx(p)=<xi|ρ|xi>=tr(ρ|xi><xi| x) where | xi>The initial value is the orthogonal one-hot coded vector expressed by Dirac symbol, | xi><xiIf is | xi>Outer product of px(p) is a second probability, i represents the ith projection plane;
s232: and combining the second probabilities to obtain the probability vector.
Further, the step of calculating the similarity between the first vector and the second vector includes:
s31: calculating a cosine value between the first vector and the second vector;
s32: and taking the cosine value as the similarity.
An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method for intelligent interview based on text matching is implemented, specifically:
s1: acquiring an interview text and a standard text, wherein the interview text is formed after an interviewer answers to interview questions, and the standard text is a text of a standard answer corresponding to the interview questions;
s2: respectively carrying out vector conversion on the interview text and the standard text according to a first preset rule to obtain a first vector corresponding to the interview text and a second vector corresponding to the standard text;
s3: calculating a similarity between the first vector and the second vector;
s4: and matching corresponding interview scores according to the similarity.
Further, the step of performing vector transformation on the interview text according to a first preset rule to obtain a first vector corresponding to the interview text includes:
s21: carrying out complex value embedding on each word in the interview text to obtain a word matrix consisting of complex value vectors corresponding to the words respectively;
s22: converting the word matrix into a mixed density matrix by adopting a sliding window;
s23: calculating according to a preset algorithm to obtain a plurality of probability vectors respectively corresponding to the mixed density matrix in different sliding windows;
s24: and respectively selecting the largest probability vector in each sliding window, wherein each largest probability vector forms the first vector.
Further, the step of converting the word matrix into a mixed density matrix using a sliding window includes:
s221: sequentially and progressively selecting a preset number of first complex value vectors from the word matrix by adopting the sliding window to form a matrix according to the arrangement sequence of the words corresponding to the complex value vectors in the interview text, until the selection of all the complex value vectors is completed, and obtaining a plurality of first word matrices;
s222: respectively calculating the outer product of each first complex value vector in each first word matrix and the vector of the conjugate device corresponding to each first complex value vector, converting each first word matrix into a corresponding word density matrix, and calculating the first probability corresponding to each first complex value vector;
s223: and calculating to obtain a mixed density matrix according to each word density matrix and each first probability.
Further, the step of calculating the first probability corresponding to each of the first complex-valued vectors includes:
s2221: respectively substituting each first complex value vector into a first preset formula, and calculating to obtain a norm corresponding to each first complex value vector, wherein the first preset formula is as follows:π(wi) Is the norm, x is the value of the first complex-valued vector;
s2222: substituting each norm into a second preset formula respectively, and calculating to obtain the first probability corresponding to each norm, wherein the second preset formula is as follows:p(wi) And e is a natural base number, l represents total l w, and j represents the jth w.
Further, the step of calculating a mixed density matrix according to the word density matrix and each of the first probabilities includes:
s2231: multiplying the vectors in the word density matrixes by the corresponding first probability respectively to obtain corresponding weighted word density matrixes respectively;
s2232: and adding the weighted word density matrixes to obtain the mixed density matrix.
Further, the step of calculating a plurality of probability vectors corresponding to the mixed density matrix in different sliding windows according to a preset algorithm includes:
s231: substituting the mixed density matrix into a third preset formula, and calculating to obtain a plurality of second probabilities, wherein the third preset formula is as follows: : p is a radical ofx(p)=〈xi|ρ|xi>=tr(ρ|xi><xi| x) where | xi>The initial value is the orthogonal one-hot coded vector expressed by Dirac symbol, | xi><xiIf is | xi>Outer product of px(p) is a second probability, i represents the ith projection plane;
s232: and combining the second probabilities to obtain the probability vector.
Further, the step of calculating the similarity between the first vector and the second vector includes:
s31: calculating a cosine value between the first vector and the second vector;
s32: and taking the cosine value as the similarity.
It will be understood by those skilled in the art that all or part of the processes of the methods of the above embodiments may be implemented by hardware associated with instructions of a computer program, which may be stored on a non-volatile computer-readable storage medium, and when executed, may include processes of the above embodiments of the methods. Any reference to memory, storage, database, or other medium provided herein and used in the examples may include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), double-rate SDRAM (SSRSDRAM), Enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
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, apparatus, article, or method 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, apparatus, article, or method. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, apparatus, article, or method that includes the element.
The above description is only for the preferred embodiment of the present application and not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings of the present application, or which are directly or indirectly applied to other related technical fields, are intended to be included within the scope of the present application.
Claims (10)
1. An intelligent interview method based on text matching is characterized by comprising the following steps:
acquiring an interview text and a standard text, wherein the interview text is formed after an interviewer answers to interview questions, and the standard text is a text of a standard answer corresponding to the interview questions;
respectively carrying out vector conversion on the interview text and the standard text according to a first preset rule to obtain a first vector corresponding to the interview text and a second vector corresponding to the standard text;
calculating a similarity between the first vector and the second vector;
and matching corresponding interview scores according to the similarity.
2. The intelligent interview method based on text matching according to claim 1, wherein the step of performing vector transformation on the interview text according to a first preset rule to obtain a first vector corresponding to the interview text comprises:
carrying out complex value embedding on each word in the interview text to obtain a word matrix consisting of complex value vectors corresponding to the words respectively;
converting the word matrix into a mixed density matrix by adopting a sliding window;
calculating according to a preset algorithm to obtain a plurality of probability vectors respectively corresponding to the mixed density matrix in different sliding windows;
and respectively selecting the largest probability vector in each sliding window, wherein each largest probability vector forms the first vector.
3. The intelligent interview method based on text matching according to claim 2, wherein said step of converting the word matrix into a mixed density matrix using a sliding window comprises:
sequentially and progressively selecting a preset number of first complex value vectors from the word matrix by adopting the sliding window to form a matrix according to the arrangement sequence of the words corresponding to the complex value vectors in the interview text, until the selection of all the complex value vectors is completed, and obtaining a plurality of first word matrices;
respectively calculating the outer product of each first complex value vector in each first word matrix and the vector of the conjugate device corresponding to each first complex value vector, converting each first word matrix into a corresponding word density matrix, and calculating the first probability corresponding to each first complex value vector;
and calculating to obtain a mixed density matrix according to each word density matrix and each first probability.
4. The intelligent interview method based on text matching according to claim 3, wherein the step of calculating the first probability corresponding to each of the first complex-valued vectors comprises:
respectively substituting each first complex value vector into a first preset formula, and calculating to obtain a norm corresponding to each first complex value vector, wherein the first preset formula is as follows:π(wi) Is the norm, x is the value of the first complex-valued vector;
5. The intelligent interview method based on text matching according to claim 3 wherein said step of calculating a mixed density matrix from said word density matrix and each of said first probabilities comprises:
multiplying the vectors in the word density matrixes by the corresponding first probability respectively to obtain corresponding weighted word density matrixes respectively;
and adding the weighted word density matrixes to obtain the mixed density matrix.
6. The intelligent interview method based on text matching according to claim 2, wherein the step of calculating a plurality of probability vectors corresponding to the mixed density matrix in different sliding windows according to a preset algorithm comprises:
substituting the mixed density matrix into a third preset formula to calculate a plurality of second probabilities,wherein the third preset formula is as follows: : p is a radical ofx(p)=〈xi|ρ|xi>=tr(ρ|xi><xi| x) where | xi>The initial value is the orthogonal one-hot coded vector expressed by Dirac symbol, | xi><xiIf is | xi>Outer product of px(p) is a second probability, i represents the ith projection plane;
and combining the second probabilities to obtain the probability vector.
7. The intelligent interview method based on text matching according to claim 1 wherein said step of calculating the similarity between said first vector and said second vector comprises:
calculating a cosine value between the first vector and the second vector;
and taking the cosine value as the similarity.
8. An intelligent interview device based on text matching, comprising:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring an interview text and a standard text, the interview text is formed after an interviewer answers to interview questions, and the standard text is a text of a standard answer corresponding to the interview questions;
the conversion module is used for respectively carrying out vector conversion on the interview text and the standard text according to a first preset rule to obtain a first vector corresponding to the interview text and a second vector corresponding to the standard text;
a calculating module, configured to calculate a similarity between the first vector and the second vector;
and the matching module is used for matching the corresponding interview scores according to the similarity.
9. A computer device comprising a memory and a processor, the memory having stored therein a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method according to any one of claims 1 to 7.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911037921.4A CN111027305B (en) | 2019-10-29 | 2019-10-29 | Text matching-based intelligent interview method and device and computer equipment |
PCT/CN2020/098796 WO2021082488A1 (en) | 2019-10-29 | 2020-06-29 | Text matching-based intelligent interview method and apparatus, and computer device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911037921.4A CN111027305B (en) | 2019-10-29 | 2019-10-29 | Text matching-based intelligent interview method and device and computer equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111027305A true CN111027305A (en) | 2020-04-17 |
CN111027305B CN111027305B (en) | 2023-12-19 |
Family
ID=70204652
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911037921.4A Active CN111027305B (en) | 2019-10-29 | 2019-10-29 | Text matching-based intelligent interview method and device and computer equipment |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN111027305B (en) |
WO (1) | WO2021082488A1 (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111694936A (en) * | 2020-04-26 | 2020-09-22 | 平安科技(深圳)有限公司 | Method and device for identifying AI intelligent interview, computer equipment and storage medium |
CN111798217A (en) * | 2020-07-10 | 2020-10-20 | 河北冀联人力资源服务集团有限公司 | Data analysis system and method |
WO2021082488A1 (en) * | 2019-10-29 | 2021-05-06 | 平安科技(深圳)有限公司 | Text matching-based intelligent interview method and apparatus, and computer device |
CN113378065A (en) * | 2021-07-09 | 2021-09-10 | 小红书科技有限公司 | Method for determining content diversity based on sliding spectrum decomposition and method for selecting content |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150269529A1 (en) * | 2014-03-24 | 2015-09-24 | Educational Testing Service | Systems and Methods for Assessing Structured Interview Responses |
CN107818164A (en) * | 2017-11-02 | 2018-03-20 | 东北师范大学 | A kind of intelligent answer method and its system |
US20180150739A1 (en) * | 2016-11-30 | 2018-05-31 | Microsoft Technology Licensing, Llc | Systems and methods for performing automated interviews |
CN109978339A (en) * | 2019-02-27 | 2019-07-05 | 平安科技(深圳)有限公司 | AI interviews model training method, device, computer equipment and storage medium |
CN110347787A (en) * | 2019-06-12 | 2019-10-18 | 平安科技(深圳)有限公司 | A kind of interview method, apparatus and terminal device based on AI secondary surface examination hall scape |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8341095B2 (en) * | 2009-01-12 | 2012-12-25 | Nec Laboratories America, Inc. | Supervised semantic indexing and its extensions |
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 |
-
2019
- 2019-10-29 CN CN201911037921.4A patent/CN111027305B/en active Active
-
2020
- 2020-06-29 WO PCT/CN2020/098796 patent/WO2021082488A1/en active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150269529A1 (en) * | 2014-03-24 | 2015-09-24 | Educational Testing Service | Systems and Methods for Assessing Structured Interview Responses |
US20180150739A1 (en) * | 2016-11-30 | 2018-05-31 | Microsoft Technology Licensing, Llc | Systems and methods for performing automated interviews |
CN107818164A (en) * | 2017-11-02 | 2018-03-20 | 东北师范大学 | A kind of intelligent answer method and its system |
CN109978339A (en) * | 2019-02-27 | 2019-07-05 | 平安科技(深圳)有限公司 | AI interviews model training method, device, computer equipment and storage medium |
CN110347787A (en) * | 2019-06-12 | 2019-10-18 | 平安科技(深圳)有限公司 | A kind of interview method, apparatus and terminal device based on AI secondary surface examination hall scape |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021082488A1 (en) * | 2019-10-29 | 2021-05-06 | 平安科技(深圳)有限公司 | Text matching-based intelligent interview method and apparatus, and computer device |
CN111694936A (en) * | 2020-04-26 | 2020-09-22 | 平安科技(深圳)有限公司 | Method and device for identifying AI intelligent interview, computer equipment and storage medium |
WO2021217866A1 (en) * | 2020-04-26 | 2021-11-04 | 平安科技(深圳)有限公司 | Method and apparatus for ai interview recognition, computer device and storage medium |
CN111694936B (en) * | 2020-04-26 | 2023-06-06 | 平安科技(深圳)有限公司 | Method, device, computer equipment and storage medium for identification of AI intelligent interview |
CN111798217A (en) * | 2020-07-10 | 2020-10-20 | 河北冀联人力资源服务集团有限公司 | Data analysis system and method |
CN113378065A (en) * | 2021-07-09 | 2021-09-10 | 小红书科技有限公司 | Method for determining content diversity based on sliding spectrum decomposition and method for selecting content |
CN113378065B (en) * | 2021-07-09 | 2023-07-04 | 小红书科技有限公司 | Method for determining content diversity based on sliding spectrum decomposition and method for selecting content |
Also Published As
Publication number | Publication date |
---|---|
WO2021082488A1 (en) | 2021-05-06 |
CN111027305B (en) | 2023-12-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111027305A (en) | Intelligent interviewing method and device based on text matching and computer equipment | |
CN110781312B (en) | Text classification method and device based on semantic representation model and computer equipment | |
CN111428467B (en) | Method, device, equipment and storage medium for generating problem questions for reading and understanding | |
CN111651992A (en) | Named entity labeling method and device, computer equipment and storage medium | |
CN113157863B (en) | Question-answer data processing method, device, computer equipment and storage medium | |
CN110457679B (en) | User portrait construction method, device, computer equipment and storage medium | |
CN111339302A (en) | Method and device for training element classification model | |
CN112365385B (en) | Knowledge distillation method and device based on self attention and computer equipment | |
CN112348362A (en) | Method, device, equipment and medium for determining position candidate | |
CN110619124A (en) | Named entity identification method and system combining attention mechanism and bidirectional LSTM | |
CN111583913B (en) | Model training method and device for speech recognition and speech synthesis and computer equipment | |
CN111223476A (en) | Method and device for extracting voice feature vector, computer equipment and storage medium | |
CN112699213A (en) | Speech intention recognition method and device, computer equipment and storage medium | |
CN116739110A (en) | Large language model distillation method based on thinking chain | |
CN112365886A (en) | Compression method and device of speech recognition model and computer equipment | |
CN113255343A (en) | Semantic identification method and device for label data, computer equipment and storage medium | |
CN113220859B (en) | Question answering method and device based on image, computer equipment and storage medium | |
CN114972792A (en) | Question-answering method, device, equipment and storage medium based on bimodal feature fusion | |
CN113255351A (en) | Sentence intention recognition method and device, computer equipment and storage medium | |
CN111078851A (en) | Information processing method, device, equipment and readable storage medium | |
CN111368044A (en) | Intelligent question answering method and device, computer equipment and storage medium | |
CN113934825B (en) | Question answering method and device and electronic equipment | |
CN114897429B (en) | Task allocation method, device, equipment and storage medium suitable for customer service communication | |
CN113988070B (en) | Investigation problem generation method, investigation problem generation device, computer equipment and storage medium | |
CN111797598A (en) | Dialog text generation method and device, computer equipment and storage medium |
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 |