Method, device, equipment and medium for extracting professional terms in intelligent interviews
By using historical interview thesaurus and sequence marking models, the problem of inaccuracy and inefficiency of named entity recognition in interviews is solved, and more efficient professional term extraction is achieved.
Patent Information
- Application Number
- CN202010356753.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-04-29
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2040-04-29
AI Technical Summary
The existing named entity recognition algorithms are difficult to accurately extract professional terms during the interview process, resulting in low recognition accuracy and low computing efficiency, mainly due to models that rely on complex structures and excessive dependence on training corpus.
The historical interview lexicon scans the response sentences, segment the phrases according to preset dimensions, count word frequency, smooth weight information, and combine the sequence marking model to determine the named entity to avoid complex models and excessive dependence on training corpus.
It improves the accuracy and efficiency of professional term extraction, adapts to interview scenarios, and simplifies the naming entity recognition process.
Smart Images

Figure CN111695337B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence, and in particular to a method, device, equipment and medium for extracting professional terms in intelligent interviews. Background Art
[0002] Recruitment interviews are a time-consuming and laborious task, because there are many candidates but limited interviewers, who need to interview a large number of candidates in a day. In order to ask relevant questions about the candidate's experience in a timely manner, understand the candidate's professional skills, extract the professional terms mentioned in the candidate's experience description and resume, and ask further questions.
[0003] However, the extraction of professional terms needs to rely on named entity recognition. Named entity recognition is the extraction of three major categories (entity category, time category and number category) and seven minor categories (personal name, organization name, place name, time, date, currency and percentage) of named entities through natural language processing algorithms. The existing algorithms will directly select the most likely segmentation label for the characters when segmenting words, instead of considering all possibilities. The interview process involves many professional terms, most of which are difficult to be included in the preset named entities, so the recognition accuracy is greatly affected by the training corpus, and there is more or less the problem of insufficient recognition accuracy. At the same time, due to the large amount of corpus used, a complex structure model is required for processing, which makes the calculation efficiency low. Summary of the invention
[0004] The embodiments of the present invention provide a method, an apparatus, a device and a storage medium for extracting professional terms in an intelligent interview, so as to improve the accuracy of extracting professional terms in an intelligent interview.
[0005] In order to solve the above technical problems, the present application embodiment provides a method for extracting professional terms in an intelligent interview, including:
[0006] Use the historical interview vocabulary to scan the answer sentences in the intelligent interview, and segment each character in the answer sentence according to the preset N dimensions to obtain a set of N phrases corresponding to each character, where N is a positive integer;
[0007] Counting the number of occurrences of each phrase in the historical interview vocabulary to obtain the word frequency of each phrase, and determining the weight information of the phrase set corresponding to each character according to the word frequency of each phrase;
[0008] Smoothing the weight information to obtain a vector representation of each character in the N phrase sets;
[0009] Extracting information from the vector representation of each of the characters to obtain representation information of each of the characters;
[0010] In combination with the representation information of the characters, a preset sequence labeling model is adopted to determine the named entities contained in the response sentence, and the named entities are used as the professional terms.
[0011] Optionally, the value of N is 4. Among the preset 4 dimensions, the first dimension is the phrase starting with a character as a named entity, the second dimension is the phrase in the middle of a character as a named entity, the third dimension is the phrase ending with a character as a named entity, and the fourth dimension is the phrase with the character itself as a named entity.
[0012] Optionally, determining the weight information of the phrase set corresponding to each character according to the word frequency of each phrase includes:
[0013] For each character, the weight information of the phrase set corresponding to the character is determined using the following formula:
[0014]
[0015]
[0016] Among them, v s For the phrase set The weight information, L is the length of the four phrases corresponding to the character ω, ω c represents the phrase that makes up the character ω, and z(ω) represents ω c The frequency of words appearing in the historical interview vocabulary, e ω (ω) represents ω c The embedding vector of , B, M, E, S are the phrase sets of four dimensions respectively.
[0017] Optionally, the smoothing process on the weight information to obtain a vector representation of each character in the N phrase sets includes:
[0018] For each character, sort each phrase in the N phrase sets corresponding to the character in order of word frequency from low to high to obtain a sorting result;
[0019] Obtain a preset ratio value, determine the selection quantity M according to the preset ratio value, and take the phrases corresponding to the first M word frequencies of the sorting result as the phrases to be processed;
[0020] According to a preset weighting method, the weight of the phrase to be processed is increased, and the weight information of the phrase set corresponding to the character is updated;
[0021] The weight information of the updated phrase set is vectorized to obtain a vector representation of the character in the N phrase sets.
[0022] Optionally, extracting information from the vector representation of each of the characters to obtain representation information of each of the characters includes:
[0023] For each of the characters, concatenate the vector representations of the N phrase sets to obtain a concatenated vector;
[0024] The concatenated vector is dimensionally compressed according to a preset fixed dimension to obtain representation information of the character.
[0025] Optionally, after taking the named entity as the professional term, the method for extracting professional terminology in the intelligent interview further includes: storing the professional terminology in a blockchain network.
[0026] In order to solve the above technical problems, the embodiment of the present application further provides a device for extracting professional terms in an intelligent interview, including:
[0027] A phrase segmentation module is used to use the historical interview vocabulary to scan the answer sentences in the intelligent interview, and to segment each character in the answer sentence according to the preset N dimensions to obtain a set of N phrases corresponding to each character, where N is a positive integer;
[0028] A weight determination module, used to count the number of occurrences of each phrase in the historical interview vocabulary to obtain the word frequency of each phrase, and determine the weight information of the phrase set corresponding to each character according to the word frequency of each phrase;
[0029] A vector representation module, used for smoothing the weight information to obtain a vector representation of each character in the N phrase sets;
[0030] A character representation module, used for extracting information from the vector representation of each character to obtain representation information of each character;
[0031] A term determination module is used to combine the representation information of the characters, adopt a preset sequence labeling model, determine the named entities contained in the response sentence, and use the named entities as the professional terms.
[0032] Optionally, the phrase segmentation module includes: the value of N is 4, and among the preset 4 dimensions, the first dimension is the phrase starting with a character as a named entity, the second dimension is the phrase in the middle of a character as a named entity, the third dimension is the phrase ending with a character as a named entity, and the fourth dimension is the phrase with the character itself as a named entity.
[0033] Optionally, the weight determination module includes:
[0034] The weight calculation unit is used to determine the weight information of the phrase set corresponding to each character using the following formula:
[0035]
[0036]
[0037] Among them, v s For the phrase set The weight information, L is the length of the four phrases corresponding to the character ω, ω c represents the phrase that makes up the character ω, and z(ω) represents ω c The frequency of words appearing in the historical interview vocabulary, e ω (ω) represents ω c The embedding vector of , B, M, E, S are the phrase sets of four dimensions respectively.
[0038] Optionally, the vector representation module includes:
[0039] A word frequency sorting unit, for sorting each word in the N word group sets corresponding to each character in order of word frequency from low to high, to obtain a sorting result;
[0040] A phrase selection unit, used to obtain a preset ratio value, determine a selection quantity M according to the preset ratio value, and take the phrases corresponding to the first M word frequencies of the sorting result as the phrases to be processed;
[0041] A weight updating unit, configured to increase the weight of the phrase to be processed according to a preset weighting method, and to update the weight information of the phrase set corresponding to the character;
[0042] The weight vectorization unit is used to vectorize the weight information of the updated phrase set to obtain the vector representation of the character in the N phrase sets.
[0043] Optionally, the character representation module includes:
[0044] A vector concatenation unit, used for concatenating the vector representations of the N phrase sets for each of the characters to obtain a concatenated vector;
[0045] The vector compression unit is used to perform dimension compression on the series vector according to a preset fixed dimension to obtain representation information of the character.
[0046] Optionally, the device for extracting professional terms in the intelligent interview further includes:
[0047] A storage module is used to store the professional terminology in the blockchain network.
[0048] In order to solve the above technical problems, an embodiment of the present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method for extracting professional terms in the above intelligent interview when executing the computer program.
[0049] In order to solve the above technical problems, an embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the method for extracting professional terms in the above intelligent interview.
[0050] The method, device, equipment and medium for extracting professional terms in intelligent interviews provided by the embodiments of the present invention use a historical interview vocabulary to scan the response sentences in the intelligent interview, and segment each character in the response sentence according to preset N dimensions to obtain a set of N phrases corresponding to each character, so as to realize that all phrases corresponding to the character are used as candidate phrases, thereby avoiding the existing word segmentation model from screening candidate phrases only according to the highest probability in the training corpus and over-reliance on the training corpus, which is conducive to improving the accuracy of professional term extraction. At the same time, the number of occurrences of each phrase in the historical interview vocabulary is counted to obtain the word frequency of each phrase, and the word frequency of each phrase is obtained according to the probability of each phrase. Word frequency, determine the weight information of the phrase set corresponding to each character, smooth the weight information, and obtain the vector representation of each character in the N phrase set, and combine the phrase information of historical interviews to give candidate phrases a weight that is more in line with the interview scenario, which is conducive to improving the accuracy of professional terminology extraction, and then extract information from the vector representation of each character to obtain the representation information of each character. Combined with the character representation information, a simple and general sequence labeling model is used to determine the named entities contained in the response sentence, and the named entities are used as professional terms to avoid determining named entities through complex named entity extraction models, thereby improving the efficiency of determining named entities. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.
[0052] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;
[0053] Figure 2 It is a flowchart of an embodiment of a method for extracting professional terms in an intelligent interview of the present application;
[0054] Figure 3 It is a structural schematic diagram of an embodiment of a device for extracting professional terms in an intelligent interview according to the present application;
[0055] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION
[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.
[0057] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0059] See also Figure 1 ,like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.
[0060] Users can use terminal devices 101 , 102 , 103 to interact with server 105 via network 104 to receive or send messages, etc.
[0061] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture E interface display perts Group Audio Layer III, Moving Picture Experts Compression Standard Audio Layer 3), MP4 (Moving Picture E interface display perts Group Audio Layer IV, Moving Picture Experts Compression Standard Audio Layer 4) players, laptop computers and desktop computers, etc.
[0062] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .
[0063] It should be noted that the method for extracting professional terms in the intelligent interview provided in the embodiment of the present application is executed by a server, and accordingly, the device for extracting professional terms in the intelligent interview is set in the server.
[0064] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. According to the implementation requirements, there may be any number of terminal devices, networks and servers. The terminal devices 101, 102, 103 in the embodiment of the present application may specifically correspond to application systems in actual production.
[0065] See also Figure 2 , Figure 2 A method for extracting professional terms in an intelligent interview provided by an embodiment of the present invention is shown. The method is applied in Figure 1 The server in is used as an example to illustrate the details as follows:
[0066] S201: Use the historical interview vocabulary to scan the response sentences in the intelligent interview, and segment each character in the response sentence according to the preset N dimensions to obtain a set of N phrases corresponding to each character, where N is a positive integer.
[0067] Specifically, the historical interview vocabulary is used to scan and identify the response sentences in the intelligent interview, and the synonyms and antonyms of each response sentence are identified. Each character in the response sentence is segmented through the preset N dimensions. Each time the segmentation is performed, a phrase set corresponding to the character is obtained, and a total of N phrase sets corresponding to the character are obtained, where N is a positive integer.
[0068] Among them, the historical interview vocabulary refers to the word segmentation vocabulary obtained after analyzing and processing the question sentences and answer sentences in the historical interview.
[0069] A phrase set is a set of phrases obtained by segmentation through one dimension.
[0070] Preferably, in this embodiment, the value of N is 4, and the preset 4 dimensions are respectively: the first dimension is the phrase starting with the character as the named entity, the second dimension is the phrase in the middle of the named entity, the third dimension is the phrase ending with the character as the named entity, and the fourth dimension is the phrase with the character itself as the named entity.
[0071] For example, in a specific embodiment, a character in sentence s is c, and four segmentation labels B(c), M(c), E(c) and S(c) are respectively generated, where B(c) is the set of all phrases in sentence s that start with the character c as a named entity, M(c) is the set of all phrases in sentence s that use the character c as the middle part of a named entity, E(c) is the set of all phrases in sentence s that end with the character c as a named entity, and S(c) refers to the phrase with the character c as a named entity.
[0072] It should be understood that using these four dimensions to perform word grouping effectively ensures that no named entity is missed, which is conducive to improving the accuracy of subsequent named entity recognition.
[0073] It should be noted that if the phrase set is empty, a special word "NONE" will be added to indicate this situation. In this way, pre-trained word embeddings are introduced, and at the same time, the corresponding matching results can be accurately restored from the phrase set of each character.
[0074] S202: Count the number of occurrences of each phrase in the historical interview vocabulary to obtain the word frequency of each phrase, and determine the weight information of the phrase set corresponding to each character according to the word frequency of each phrase.
[0075] Specifically, after obtaining a set of N phrases corresponding to each character to ensure that all phrases corresponding to the character are not missed, it is necessary to distinguish the importance of each phrase in combination with the actual scenario of the intelligent interview, so as to achieve rapid labeling in the subsequent named entity recognition and improve the efficiency of named entity recognition. In this embodiment, the frequency of each phrase is determined by counting the number of occurrences of each phrase in the historical interview vocabulary related to the intelligent interview scenario, and then the weight information of the phrase set corresponding to the character is determined according to the frequency of each phrase.
[0076] Among them, the word frequency of the phrase is a static value, which can be obtained through preliminary statistical calculation and stored in a data table. When needed, it can be obtained by looking up the table to improve the efficiency of determining the weight information. The word frequency of the phrase can also be updated according to actual needs.
[0077] It should be understood that in order to improve computational efficiency, this embodiment uses the frequency of occurrence of a phrase to represent the weight of the phrase, because the more times a specified word sequence appears in the historical interview vocabulary, the more likely the sequence is to be a phrase.
[0078] The specific implementation process of determining the weight information of the phrase set corresponding to each character according to the word frequency of each phrase may refer to the description of the subsequent embodiments, and will not be described again here to avoid repetition.
[0079] S203: Smoothing the weight information to obtain a vector representation of each character in the N phrase sets.
[0080] Specifically, through smoothing processing, the weights of phrases with smaller weights are optimized to avoid phrases with smaller weights being ignored in the subsequent labeling process due to large weight differences, which may lead to the loss of named entities corresponding to phrases with smaller weights. After smoothing, the weight information of the N phrase set is updated and vectorized to obtain the vector representation of each character in the N phrase set.
[0081] It should be noted that unsmoothed weight information will affect the data fitting during subsequent information extraction, which may easily lead to inaccurate information extraction. Therefore, the weight information needs to be smoothed. The smoothing process in this embodiment refers to optimizing the weights that do not meet the requirements through some mathematical models, so that the overall weight information is within a reasonable range, which is conducive to subsequent information extraction.
[0082] The specific process of smoothing the weight information may refer to the description of the subsequent embodiments, and will not be described again here to avoid repetition.
[0083] S204: extract information from the vector representation of each character to obtain representation information of each character.
[0084] Specifically, after obtaining the vector representation of each character, key information is extracted from the vector representation to obtain representation information of a fixed dimension, so that the representation information can be subsequently input into the sequence labeling model to determine the named entity.
[0085] The representation information is a vector of fixed dimension used to represent the association between characters and each phrase.
[0086] For the specific implementation process of extracting information from the vector representation of each character to obtain the representation information of each character, please refer to the description of the subsequent embodiments, which will not be described here to avoid repetition.
[0087] S205: In combination with the character representation information, a preset sequence labeling model is used to determine the named entities contained in the response sentence, and the named entities are used as professional terms.
[0088] Specifically, after obtaining the representation information of each character, the preset sequence labeling model is used to perform named entity recognition on the response sentence, so as to quickly and accurately obtain the named entities contained in the response sentence, that is, the professional terms mentioned in the candidate's response sentence can be quickly extracted.
[0089] Named entity recognition refers to the process of identifying the names or symbols of a specific type of thing in a document collection. Named entity recognition consists of three problems: identifying the named entity in the text; determining the type of the entity; and when multiple entities represent the same thing, selecting one of them as the representative of the group of entities.
[0090] In this embodiment, named entity recognition mainly refers to identifying professional terms from the candidate's response statements, and subsequently evaluating the candidate's interview situation based on the identified professional terms, for example, determining whether the experience in the resume is consistent with the response, that is, verifying the credibility of the candidate's response, or testing the candidate's professional ability.
[0091] The sequence labeling model includes, but is not limited to, a conditional random field model (CRF), a hidden Markov model (HMM), and a maximum entropy Markov model (MEMM). The preset sequence labeling model in this embodiment can be any one of the existing sequence labeling models, and no further limitation is made here.
[0092] It should be noted that in step S204, character representation information is obtained. In the sequence labeling model, the character is weighted using the representation information, which helps the sequence labeling model to quickly label key features and improve the accuracy and efficiency of labeling.
[0093] In this embodiment, by using the historical interview vocabulary, the answer sentences in the intelligent interview are scanned, and each character in the answer sentence is segmented according to the preset N dimensions to obtain a set of N phrases corresponding to each character, so that all phrases corresponding to the character are used as candidate phrases, avoiding the existing word segmentation model to screen candidate phrases only according to the highest probability in the training corpus and over-reliance on the training corpus, which is conducive to improving the accuracy of professional terminology extraction. At the same time, the number of occurrences of each phrase in the historical interview vocabulary is counted to obtain the word frequency of each phrase, and the word frequency of each phrase is determined according to the word frequency of each phrase. The weight information of the group set is smoothed to obtain the vector representation of each character in the N phrase set, and the phrase information of historical interviews is combined to assign weights that are more in line with the interview scenario to the candidate phrases, which is conducive to improving the accuracy of professional terminology extraction, and then the vector representation of each character is extracted to obtain the representation information of each character. Combined with the character representation information, a simple and general sequence labeling model is used to determine the named entities contained in the response sentence, and the named entities are used as professional terms to avoid determining the named entities through complex named entity extraction models, thereby improving the efficiency of determining named entities.
[0094] In one embodiment, the extracted professional terms can be saved on the blockchain network. Through blockchain storage, data information can be shared between different platforms and data can be prevented from being tampered with.
[0095] Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.
[0096] In some optional implementations of this embodiment, in step S202, determining the weight information of the phrase set corresponding to each character according to the word frequency of each phrase includes:
[0097] For each character, the following formula is used to determine the weight information of the phrase set corresponding to the character:
[0098]
[0099]
[0100] Among them, v s For phrase collection The weight information, L is the length of the four phrases corresponding to the character ω, ω c represents the phrase that makes up the character ω, and z(ω) represents ω c The frequency of words appearing in the historical interview vocabulary, e ω (ω) represents ω c The embedding vector of , B, M, E, S are the phrase sets of four dimensions respectively.
[0101] Specifically, the weight information of each phrase set is calculated through the above formula, so that a corresponding vector representation is subsequently generated according to the weight information of each phrase set.
[0102] Among them, word embedding is an important concept in natural language processing, which is used to convert a word into a vector representation of fixed length, so as to facilitate mathematical processing. The embedding vector can be generated by inputting characters into the open source tensorflow model, or the phrases corresponding to the characters can be represented in the form of a sequence, and then the sequence is converted into an embedding vector through the tf.nn.embedding_lookup function.
[0103] In this embodiment, the weight information of the phrase set corresponding to the character is obtained through calculation by a set formula, so that the representation information of the character can be obtained subsequently through the weight information. By assigning different weights to different phrases corresponding to the characters, the fit with the interview scenario is higher, which is conducive to improving the accuracy of subsequent named entity extraction.
[0104] In some optional implementations of this embodiment, in step S203, the weight information is smoothed to obtain a vector representation of each character in the N phrase set, including:
[0105] For each character, sort each phrase in the set of N phrases corresponding to the character in order of word frequency from low to high to obtain a sorting result;
[0106] Obtain a preset ratio value, determine the selection quantity M according to the preset ratio value, and take the phrases corresponding to the first M word frequencies of the sorting results as the phrases to be processed;
[0107] According to a preset weighting method, the weight of the phrase to be processed is increased, and the weight information of the phrase set corresponding to the character is updated;
[0108] The weight information of the updated phrase set is vectorized to obtain the vector representation of the characters in the N phrase sets.
[0109] Specifically, the phrases in the set of N phrases corresponding to the characters are sorted by frequency, and the phrases that need to be smoothed are selected according to preset conditions for smoothing, and then the weight information is updated, and the updated weight information is vectorized to obtain the vector representation of the character in the set of N phrases.
[0110] The preset ratio value may be set according to actual needs. For example, in a specific implementation, the preset comparison value is 10%.
[0111] The value of M is determined according to the product of a preset comparison value and the number of phrases in the phrase set.
[0112] The preset weighting method can be selected according to actual needs. In this embodiment, weighting is performed by the following formula:
[0113]
[0114] Wherein, a is a constant, and its value is determined by the maximum word frequency in the phrase to be processed.
[0115] For example, in a specific implementation, the preset ratio is 10%, and the number of phrases in the phrase set is 180, then the corresponding value of M is 18. After sorting by word frequency, the phrases corresponding to the last 18 word frequencies in the sequence are taken as the phrases to be processed, and the maximum word frequency in the phrases to be processed is taken as the constant a.
[0116] In this embodiment, by appropriately smoothing the weights of phrases with lower word frequencies, it is prevented that the phrases with lower word frequencies lose their effects in the future, resulting in missing possible named entities.
[0117] In some optional implementations of this embodiment, in step S204, information is extracted from the vector representation of each character to obtain the representation information of each character, including:
[0118] For each character, concatenate the vector representations of the N phrase sets to obtain a concatenated vector;
[0119] According to the preset fixed dimension, the concatenated vector is dimensionally compressed to obtain the character representation information.
[0120] Specifically, after obtaining the vector representation of the character in N phrases, in order to more accurately determine the phrase combination corresponding to the character, the vector identifiers of the N phrases corresponding to each character are compressed in a preset manner to obtain a vector of fixed dimension.
[0121] In this embodiment, in order to retain as much information as possible, the preset method is to concatenate the four phrase sets and represent them as a whole to obtain the single word representation of the character, and the following formula is used for compression:
[0122]
[0123] Among them, e s represents the embedding vector corresponding to the character, represents vector concatenation, v s represents the function that maps the corresponding phrase set to a dense vector, x c It is the representation information corresponding to the character.
[0124] In this embodiment, the vector representations of the phrase set are concatenated to obtain a concatenated vector, and the concatenated vector is compressed to a fixed dimension so as to be subsequently input into a preset sequence labeling model to determine the named entity.
[0125] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
[0126] Figure 3 The principle block diagram of the device for extracting professional terms in the intelligent interview corresponding to the method for extracting professional terms in the intelligent interview in the above embodiment is shown. Figure 3 As shown, the device for extracting professional terms in the intelligent interview includes a phrase segmentation module 31, a weight determination module 32, a vector representation module 33, a character representation module 34 and a term determination module 35. The functional modules are described in detail as follows:
[0127] The phrase segmentation module 31 is used to use the historical interview vocabulary to scan the answer sentences in the intelligent interview, and segment each character in the answer sentence according to the preset N dimensions to obtain N phrase sets corresponding to each character, where N is a positive integer;
[0128] The weight determination module 32 is used to count the number of occurrences of each phrase in the historical interview word library to obtain the word frequency of each phrase, and determine the weight information of the phrase set corresponding to each character according to the word frequency of each phrase;
[0129] A vector representation module 33, used for smoothing the weight information to obtain a vector representation of each character in the N phrase set;
[0130] A character representation module 34 is used to extract information from the vector representation of each character to obtain representation information of each character;
[0131] The term determination module 35 is used to combine the character representation information, adopt a preset sequence labeling model, determine the named entities contained in the response sentence, and use the named entities as professional terms.
[0132] Optionally, the phrase segmentation module 31 includes: the value of N is 4, and among the four preset dimensions, the first dimension is the phrase that starts with a character as a named entity, the second dimension is the phrase that uses a character as a named entity in the middle, the third dimension is the phrase that ends with a character as a named entity, and the fourth dimension is the phrase that uses the character itself as a named entity.
[0133] Optionally, the weight determination module 32 includes:
[0134] The weight calculation unit is used to determine the weight information of the phrase set corresponding to each character using the following formula:
[0135]
[0136]
[0137] Among them, v s For phrase collection The weight information, L is the length of the four phrases corresponding to the character ω, ω c represents the phrase that makes up the character ω, and z(ω) represents ω c The frequency of words appearing in the historical interview vocabulary, e ω (ω) represents ω c The embedding vector of , B, M, E, S are the phrase sets of four dimensions respectively.
[0138] Optionally, the vector representation module 33 includes:
[0139] A word frequency sorting unit is used to sort each word in the N word group set corresponding to each character in the order of word frequency from low to high to obtain a sorting result;
[0140] A phrase selection unit is used to obtain a preset ratio value, determine the selection quantity M according to the preset ratio value, and take the phrases corresponding to the first M word frequencies in the sorting result as the phrases to be processed;
[0141] A weight updating unit, used to increase the weight of the phrase to be processed according to a preset weighting method, and update the weight information of the phrase set corresponding to the character;
[0142] The weight vectorization unit is used to vectorize the weight information of the updated phrase set to obtain the vector representation of the character in the N phrase sets.
[0143] Optionally, the character representation module 34 includes:
[0144] A vector concatenation unit, for concatenating the vector representations of the N phrase sets for each character to obtain a concatenated vector;
[0145] The vector compression unit is used to perform dimension compression on the concatenated vector according to a preset fixed dimension to obtain character representation information.
[0146] Optionally, the device for extracting professional terms in the intelligent interview also includes:
[0147] The storage module is used to store professional terms in the blockchain network.
[0148] For the specific definition of the device for extracting professional terms in intelligent interviews, please refer to the definition of the method for extracting professional terms in intelligent interviews in the above text, which will not be repeated here. Each module in the above-mentioned device for extracting professional terms in intelligent interviews can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0149] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.
[0150] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components connected to the memory 41, the processor 42, and the network interface 43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.
[0151] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.
[0152] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or D interface display memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk equipped on the computer device 4, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as the program code for controlling electronic files, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.
[0153] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the program code stored in the memory 41 or process data, such as running the program code for controlling the electronic file.
[0154] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.
[0155] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores an interface display program, and the interface display program can be executed by at least one processor so that the at least one processor performs the steps of the method for extracting professional terms in the intelligent interview as described above.
[0156] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.
[0157] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.
Claims
1. A method for extracting professional terms in intelligent interviews, characterized in that: The method for extracting professional terms in the intelligent interview includes: Use the historical interview vocabulary to scan the answer sentences in the intelligent interview, and segment each character in the answer sentence according to the preset N dimensions to obtain a set of N phrases corresponding to each character, where N is a positive integer; Counting the number of occurrences of each phrase in the historical interview vocabulary to obtain the word frequency of each phrase, and determining the weight information of the phrase set corresponding to each character according to the word frequency of each phrase; Smoothing the weight information to obtain a vector representation of each character in the N phrase sets; Extracting information from the vector representation of each of the characters to obtain representation information of each of the characters; In combination with the representation information of the characters, a preset sequence labeling model is adopted to determine the named entities contained in the response sentence, and the named entities are used as the professional terms.
2. The method for extracting professional terms in an intelligent interview as claimed in claim 1, characterized in that: The value of N is 4. Among the preset 4 dimensions, the first dimension is the phrase that starts with a character as a named entity, the second dimension is the phrase that uses a character as a named entity in the middle, the third dimension is the phrase that ends with a character as a named entity, and the fourth dimension is the phrase that uses the character itself as a named entity.
3. The method for extracting professional terms in an intelligent interview as claimed in claim 2, characterized in that: Determining the weight information of the phrase set corresponding to each character according to the word frequency of each phrase includes: For each character, the weight information of the phrase set corresponding to the character is determined using the following formula: Among them, v s For the phrase set The weight information, L is the length of the four phrases corresponding to the character ω, ω c represents the phrase that makes up the character ω, and z(ω) represents ω c The frequency of words appearing in the historical interview vocabulary, e ω (ω) represents ω c The embedding vector of , B, M, E, S are the phrase sets of four dimensions respectively.
4. The method for extracting professional terms in an intelligent interview as claimed in claim 3, characterized in that: The smoothing process of the weight information to obtain the vector representation of each character in the N phrase sets includes: For each character, sort each phrase in the N phrase sets corresponding to the character in order of word frequency from low to high to obtain a sorting result; Obtain a preset ratio value, determine the selection quantity M according to the preset ratio value, and take the phrases corresponding to the first M word frequencies of the sorting result as the phrases to be processed; According to a preset weighting method, the weight of the phrase to be processed is increased, and the weight information of the phrase set corresponding to the character is updated; The weight information of the updated phrase set is vectorized to obtain a vector representation of the character in the N phrase sets.
5. The method for extracting professional terms in an intelligent interview according to any one of claims 1 to 4, characterized in that: The extracting information from the vector representation of each character to obtain the representation information of each character includes: For each of the characters, concatenate the vector representations of the N phrase sets to obtain a concatenated vector; The concatenated vector is dimensionally compressed according to a preset fixed dimension to obtain representation information of the character.
6. The method for extracting professional terms in an intelligent interview as claimed in claim 1, characterized in that: After the named entity is used as the professional term, the method further includes: storing the professional term in a blockchain network.
7. A device for extracting professional terms in intelligent interviews, characterized in that: The device for extracting professional terms in the intelligent interview includes: A phrase segmentation module is used to use the historical interview vocabulary to scan the answer sentences in the intelligent interview, and to segment each character in the answer sentence according to the preset N dimensions to obtain a set of N phrases corresponding to each character, where N is a positive integer; A weight determination module, used to count the number of occurrences of each phrase in the historical interview vocabulary to obtain the word frequency of each phrase, and determine the weight information of the phrase set corresponding to each character according to the word frequency of each phrase; A vector representation module, used for smoothing the weight information to obtain a vector representation of each character in the N phrase sets; A character representation module, used for extracting information from the vector representation of each character to obtain representation information of each character; A term determination module is used to combine the representation information of the characters, adopt a preset sequence labeling model, determine the named entities contained in the response sentence, and use the named entities as the professional terms.
8. The device for extracting professional terms in intelligent interview according to claim 7, characterized in that: The vector representation module comprises: A word frequency sorting unit, for sorting each word in the N word group sets corresponding to each character in order of word frequency from low to high, to obtain a sorting result; A phrase selection unit, used to obtain a preset ratio value, determine a selection quantity M according to the preset ratio value, and take the phrases corresponding to the first M word frequencies of the sorting result as the phrases to be processed; A weight updating unit, configured to increase the weight of the phrase to be processed according to a preset weighting method, and to update the weight information of the phrase set corresponding to the character; The weight vectorization unit is used to vectorize the weight information of the updated phrase set to obtain the vector representation of the character in the N phrase sets.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for extracting professional terms in an intelligent interview as described in any one of claims 1 to 6 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for extracting professional terms in an intelligent interview as described in any one of claims 1 to 6 is implemented.
