Vocational skills assessment method and system based on big data
By performing word segmentation processing and part-of-speech annotation on the answered text and corpus, and correcting the TF-IDF value with word meaning similarity and distance similarity, the problem of low accuracy of TF-IDF value in the existing technology is solved, and a more accurate career has been achieved Skills assessment improves the accuracy and relevance of the assessment.
Patent Information
- Application Number
- CN202410993592.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-07-24
AI Technical Summary
In the prior art, the calculation accuracy of TF-IDF values is low, which makes it impossible to accurately reflect the importance of words in the text, affecting the accuracy of professional skills assessment.
By performing word segmentation processing, part-of-speech annotation and clustering of the answered text and corpus, the part-of-speech influence value is calculated, and the initial TF-IDF value is corrected based on the word meaning similarity and distance similarity, more accurate text influence and materials are calculated. Influence degree, and ultimately obtain more accurate results for professional skills assessment.
It improves the accuracy of professional skills assessment, provides a more accurate evaluation basis, enhances the relevance and information weight of the assessment, and ensures the accuracy of professional skills assessment results of job seekers or employees.
Smart Images

Figure CN118627972B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing, and more specifically, to a method and system for evaluating professional skills based on big data. Background Art
[0002] As the society's demand for high-quality skilled personnel continues to grow, it has become particularly important to accurately and efficiently assess the professional skills of individuals. Due to the development and popularization of big data, more and more companies have begun to use big data technology to assess the professional skills of job seekers or employees, so as to better understand the professional skills of job seekers or employees, facilitate the subsequent training and management of job seekers or employees, and also promote the healthy development of the labor market and promote social and economic progress and innovation. Professional skills assessment mainly examines job seekers or employees in the form of written tests. Therefore, establishing a scientific and systematic professional skills assessment system and adopting modern assessment tools and methods are of great practical significance for improving the accuracy and efficiency of assessment.
[0003] The existing Chinese patent application document with publication number CN114004218A discloses an NLP recognition method for government work reports. This method divides the article into sentences according to commas, periods, exclamation marks, and question marks, and then extracts the top ten keywords based on the number of occurrences of the words in the sentences, thereby calculating the TF-IDF value.
[0004] However, the same word may express different meanings in a sentence, which will affect the TF-IDF value of the same word. This will cause the inverse document frequency of more important words to be lower, and the accuracy of the calculated TF-IDF value will be reduced, which will not accurately reflect the importance of the word in the text. Summary of the invention
[0005] In order to solve the problem of low accuracy of the calculated TF-IDF value, the present invention proposes a professional skill assessment method and system based on big data.
[0006] In a first aspect, the present invention discloses a method for assessing professional skills based on big data, comprising: performing word segmentation processing on an answer text and a corpus to obtain a number of words, performing part-of-speech tagging and clustering on all words, and taking the ratio of the maximum frequency of a word appearing in different parts of speech to the cumulative value of the frequency of the word appearing in each part of speech as the word's part-of-speech influence value; grouping the words, and in the answer text, taking the ratio of the part-of-speech influence value of a target word to the cumulative value of the part-of-speech influence value of each word in a target word group as a first influence degree, and taking the average of the first influence degrees of all target word groups as the text influence degree, wherein the target word is simultaneously For any word that appears in the answer text and the corpus, the target word group is the word group containing the target word; in the corpus, the ratio of the part-of-speech influence value of the target word to the cumulative value of the part-of-speech influence value of each word in the target word group is taken as the second influence, the average of the second influence of all target word groups is taken as the third influence of the text, and the average of the third influence of the target word in all texts is taken as the corpus influence; the semantic similarity of the text influence and the corpus influence of the target word is calculated, and the product of the semantic similarity and the initial TF-IDF value of the target word is taken as the TF-IDF correction value for professional skill assessment.
[0007] By segmenting and tagging the answer text and corpus, clustering words and calculating the part-of-speech influence value, the importance of each word in different parts of speech can be quantified. Then, by calculating the first influence and text influence, the significance and influence of the target word in a specific text are evaluated. Furthermore, by analyzing the second and third influences, the general importance of the target word in the entire corpus is evaluated. Finally, by calculating the word meaning similarity and adjusting the TF-IDF value, the meaning similarity and information weight of the word between the specific text and the broad corpus can be measured.
[0008] Preferably, the word segmentation processing of the answer text and the corpus to obtain a number of words includes: generating multiple texts from internal enterprise materials according to a text generation model as a corpus for professional skills assessment; cleaning the generated text data, removing punctuation marks and blank characters in the text and retaining numbers and special characters, using the UTF-8 encoding format to uniformly encode the text data, and normalizing the text data; using a forward maximum matching algorithm to perform word segmentation processing on the answer text and each text in the corpus to obtain a number of words.
[0009] By applying the text generation model, the internal data of the enterprise is effectively expanded, and a diverse text is created as a corpus for professional skills assessment, which enhances the richness and coverage of the assessment materials. The quality and consistency of the text data are ensured by thoroughly cleaning the generated text data, removing unnecessary punctuation and blank characters, while retaining key numbers and special characters.
[0010] Preferably, the part-of-speech tagging and clustering of all words includes: using natural language processing to tag the part-of-speech of each word in combination with the context in which the word is located; clustering all words according to the part-of-speech of the word using a preset clustering algorithm to obtain noun class, verb class and adjective class.
[0011] Using part-of-speech tagging to identify the grammatical function of each word helps to understand the text structure and semantic content more accurately. Then, based on the word's part of speech, a preset clustering algorithm is applied to classify the word into categories such as nouns, verbs, and adjectives. Classification not only helps to reveal the semantic composition of the text, but also highlights the role and importance of different parts of speech in expression.
[0012] Preferably, the word meaning similarity satisfies the relationship:
[0013] ,in, Representing words The similarity of the meaning of Representing words The text influence, Representing words The influence of the corpus, represents the normalization function, the word This is the target word.
[0014] Preferably, the TF-IDF correction value also includes:
[0015] ,in, Representing words The TF-IDF correction value of Representing words The similarity of the meaning of Representing words The distance similarity of Representing words The initial TF-IDF value of the word This is the target word.
[0016] By correcting the distance similarity and word meaning similarity, the importance of words in the current text can be more accurately reflected.
[0017] Preferably, the difference between the position information of the target word in the answer text and the position information of the target word in the corpus is calculated, and the result of negative correlation mapping of the difference is used as the distance similarity of the target word.
[0018] Preferably, the position information of the target word in the answer text includes: using a word embedding model to convert each word into a word vector; taking the first word in each word group as the starting word; the position information of the target word in the answer text satisfies the relationship:
[0019] ,in, Representing words Position information in the answer text, Representing words With The Euclidean distance between the starting words in the word group, Indicates The total number of words in the word group, Indicates the total number of target word groups, word This is the target word.
[0020] A method that comprehensively considers word semantics and contextual relationships is used to analyze text content. First, a word embedding model is used to convert words into word vectors, which can capture the rich semantic information and contextual relationships of words. Then, the first word in the word group is selected as the reference point, and the Euclidean distance between the target word and the reference point is calculated to quantify the position information of the target word in the text.
[0021] Preferably, the position information of the target word in the corpus is obtained in the same way.
[0022] In a second aspect, the present invention discloses a professional skills assessment system based on big data, comprising: a processor; and a memory, wherein the memory stores computer instructions, and when the computer instructions are executed by the processor, the device executes the above-mentioned professional skills assessment method based on big data.
[0023] Beneficial effects of the present invention:
[0024] The present invention can identify more important words in a specific part of speech by comparing the occurrence frequencies of words in different parts of speech, calculate the part-of-speech influence of important words, evaluate the importance of words in the text, and use the calculated word meaning similarity and distance similarity as weights to correct the initial TF-IDF value, so as to obtain a more accurate evaluation basis. The more accurate TF-IDF value can be used to reflect the correlation between the answer text and the corpus, thereby providing a more accurate evaluation reference for enterprises and enabling job seekers or employees to obtain more accurate professional skills evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:
[0026] Figure 1 It is a flow chart of a method for evaluating professional skills based on big data according to an embodiment of the present invention.
[0027] Figure 2 It is a schematic diagram of a vocational skill assessment system based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION
[0028] 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 those skilled in the art without creative work are within the scope of protection of the present invention.
[0029] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.
[0030] The present invention provides a method for evaluating professional skills based on big data. Figure 1 As shown, the occupational skill assessment method based on big data includes steps S1 to S3, which are described in detail below.
[0031] S1, perform word segmentation on the answer text and corpus to obtain several words, perform part-of-speech tagging and clustering on all words, and calculate the part-of-speech influence value.
[0032] It should be noted that the professional skills assessment is mainly conducted in the form of a written test on job applicants or employees, which includes two types of relevant questions: objective questions and subjective questions. The professional skills level of job applicants or employees is assessed based on the final answer scores. The relevant questions are usually generated from internal company information. Therefore, the similarity between the answer text of the subjective questions answered by the job applicant or employee and the corpus can be compared to evaluate the job applicant or employee.
[0033] The TF-IDF algorithm is a method to calculate the importance of words in a text and obtain text keywords. Its core is to calculate the TF-IDF value of a word based on the term frequency TF and the inverse document frequency IDF. The TF-IDF value of each word can reflect the importance of the word in the text, and the keywords in the document are obtained according to the importance of each word.
[0034] In one embodiment, multiple texts are generated from internal enterprise materials according to a text generation model as a corpus for professional skills assessment. Exemplarily, the text generation model includes a recurrent neural network, a gated recurrent unit, a generative adversarial network, and a generative pre-trained transformer, etc. The present invention adopts a generative adversarial network.
[0035] The generated text data is cleaned, punctuation marks and blank characters are removed from the text, numbers and special characters are retained, the text data is uniformly encoded using the UTF-8 encoding format, and the text data is normalized. Normalization is to remove common stop words that do not contain important information, such as auxiliary words, prepositions, conjunctions, etc.
[0036] Use the forward maximum matching algorithm to segment the answer text and each text in the corpus to obtain several words.
[0037] Use natural language processing to perform part-of-speech tagging on each word in combination with the context in which the word is located. It should be noted that a word is a noun, which is usually used to refer to people, things, and concepts. It is the most basic element that constitutes a sentence and is likely to appear frequently in a text. A word is a verb, which indicates an action or state and is the core component of the meaning of a sentence. Since a sentence usually only requires one main verb, a word that is a verb is less likely to appear frequently in a text. A word is an adjective, which is used to modify a noun and is not a necessary component of a sentence. Therefore, the frequency of adjectives in a text is lower than that of nouns, but is usually greater than or equal to the frequency of verbs.
[0038] All words are clustered according to their parts of speech using a preset clustering algorithm to obtain nouns, verbs, and adjectives. Exemplarily, the preset clustering algorithms include DBSCAN (density-based clustering), K-means (k-means clustering), hierarchical clustering, iterative self-organizing clustering, etc. The present invention uses K-means.
[0039] Since the word parts of speech are marked in combination with the context, the same word may have different parts of speech. Therefore, it is necessary to select the word with the most parts of speech and calculate the word's part of speech influence value based on it. The part of speech influence value satisfies the relationship:
[0040] ,in, Representing words The part-of-speech influence value of represents the normalization function, Representing words The maximum frequency of occurrence under different parts of speech, Indicates the frequency of occurrence, Indicates the part of speech, 1 refers to nouns, 2 refers to verbs, and 3 refers to adjectives.
[0041] Different parts of speech of words will have a certain impact on the number of times the words appear in the answer text and the corpus, and indirectly affect the word frequency in the corpus, that is, the inverse document frequency (IDF) of the word. Therefore, it is necessary to calculate the part-of-speech impact value of the word.
[0042] S2, group the words and calculate the text influence and corpus influence of the target words respectively.
[0043] It should be noted that in order to ensure that the divided word groups can reflect the word-part influence relationship between the words and other words in the word group, it is necessary to ensure that the word group contains basic sentence structure as much as possible, that is, the subject (noun), predicate (verb), and object (noun) appear in sequence, and there may be modifiers (adjectives). If there are modifiers, they will generally come before the object.
[0044] In one embodiment, the answer texts and the texts in the corpus are grouped separately. Since the corpus is generated based on internal data of an enterprise, which usually contains more content, the number of generated texts is more than one.
[0045] Any word that appears in the answer text and the corpus at the same time is taken as the target word, and the word group containing the target word is taken as the target word group.
[0046] In the answer text, the first influence is calculated, and the first influence satisfies the relationship:
[0047] ,in, Indicates The first influence of the word phrase, Representing words The part-of-speech influence value of Representing words The part-of-speech influence value of Indicates The total number of words in a word group, the word This is the target word.
[0048] All target word groups in the answer text are traversed, and the average of the first influences of all target word groups is used as the text influence.
[0049] In the corpus, the second influence is calculated, and the second influence satisfies the relationship:
[0050] ,in, Indicates The second influence of the word phrase, Representing words The part-of-speech influence value of Representing words The part-of-speech influence value of Indicates The total number of words in a word group, the word This is the target word.
[0051] Traverse all target word groups in the corpus, take the average of the second influence of all target word groups as the third influence of the text, and then take the average of the third influence of the target word in all texts as the corpus influence.
[0052] S3, calculate the TF-IDF correction value for occupational skill assessment.
[0053] In one embodiment, the semantic similarity of the text influence and the corpus influence of the target word is calculated. For example, the similarity calculation method is a well-known technology and can be calculated using existing technology. The present invention exemplarily provides a calculation method, and the semantic similarity satisfies the relationship:
[0054] ,in, Representing words The similarity of the meaning of Representing words The text influence, Representing words The influence of the corpus, represents the normalization function, the word This is the target word.
[0055] The product of the semantic similarity and the initial TF-IDF value of the target word is taken as the TF-IDF correction value, and the TF-IDF correction value satisfies the relationship:
[0056] ,in, Representing words The TF-IDF correction value of Representing words The similarity of the meaning of Representing words The initial TF-IDF value of the word This is the target word.
[0057] Traverse and obtain the TF-IDF modified values of all words, and arrange them in descending order. The larger the value, the greater the probability of being a keyword in the text. For example, select the words with a TF-IDF modified value greater than The words are composed of the answer text keyword vector, and the TF-IDF correction value in the corpus is greater than The words of the corpus keyword vector are used to calculate the correlation between the answer text keyword vector and the corpus keyword vector. The method for calculating the correlation is well known to those skilled in the art and will not be elaborated on here. The greater the correlation, the closer the answer text of the job seeker or employee is to the corpus of the company, and the closer the answer text is to the correct answer of the company. Auxiliary scoring is performed based on the correlation, and the scoring value is set according to the specific actual subjective question score to complete the professional skills assessment.
[0058] In another embodiment, the semantic similarity of the text influence and the corpus influence of the target word is calculated.
[0059] Use the word embedding model to convert each word into a word vector; take the first word in each word group as the starting word, and calculate the position information of the target word in the answer text and the position information of the target word in the corpus respectively.
[0060] The position information of the target word in the answer text satisfies the relationship:
[0061] ,in, Representing words Position information in the answer text, Representing words With The Euclidean distance between the starting words in the word group, Indicates The total number of words in the word group, Indicates the total number of target word groups, word This is the target word.
[0062] Similarly, the position information of the target word in the corpus is obtained.
[0063] Calculate the difference between the position information of the target word in the answer text and the position information of the target word in the corpus, and use the result of negative correlation mapping of the difference as the distance similarity of the target word. The distance similarity satisfies the relationship:
[0064] ,in, Representing words The distance similarity of Representing words Position information in the answer text, Representing words Position information in the corpus, represents the normalization function, the word This is the target word.
[0065] The product of the semantic similarity and distance similarity and the initial TF-IDF value of the target word is taken as the TF-IDF correction value, and the TF-IDF correction value satisfies the relationship:
[0066] ,in, Representing words The TF-IDF correction value of Representing words The similarity of the meaning of Representing words The distance similarity of Representing words The initial TF-IDF value of the word This is the target word.
[0067] Traverse and obtain the TF-IDF modified values of all words, and arrange them in descending order. The larger the value, the greater the probability of being a keyword in the text. For example, select the words with a TF-IDF modified value greater than The words are composed of the answer text keyword vector, and the TF-IDF correction value in the corpus is greater than The words form the corpus keyword vector, calculate the correlation between the answer text keyword vector and the corpus keyword vector, and complete the professional skills assessment.
[0068] The embodiment of the present invention also discloses a vocational skill evaluation system based on big data, referring to Figure 2 , including a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the big data-based professional skills assessment method according to the present invention is implemented.
[0069] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.
[0070] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module or both. Any such computer storage medium may be part of a device or accessible or connectable to a device.
[0071] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.
[0072] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A vocational skills assessment method based on big data, characterized in that: include: The answer text and corpus are segmented to obtain a number of words, all words are tagged and clustered, and the ratio of the maximum frequency of a word in different parts of speech to the cumulative value of the frequency of the word in each part of speech is used as the word's part of speech influence value; wherein, multiple texts are generated from the internal data of the enterprise according to the text generation model as the corpus for professional skills assessment; The words are grouped, and in the answer text, the ratio of the part-of-speech influence value of the target word to the cumulative value of the part-of-speech influence value of each word in the target word group is used as the first influence degree, and the average of the first influence degrees of all target words is used as the text influence degree, wherein the target word is any word that appears in the answer text and the corpus at the same time, and the target word group is the word group containing the target word; In the corpus, the ratio of the part-of-speech influence value of the target word to the cumulative value of the part-of-speech influence value of each word in the target word group is used as the second influence, the average of the second influences of all target words is used as the third influence of the text in the corpus, and the average of the third influences of the target word in all texts in the corpus is used as the corpus influence; Calculate the semantic similarity of the target word's text influence and corpus influence, satisfying the relationship: ,in, Representing words The similarity of the meaning of Representing words The text influence, Representing words The influence of the corpus, represents the normalization function, the word That is the target word; the product of the word semantic similarity and the initial TF-IDF value of the target word is taken as the TF-IDF correction value. The TF-IDF correction value can reflect the relevance between the answer text and the corpus for professional skills assessment.
2. The method for evaluating occupational skills based on big data according to claim 1, characterized in that: The word segmentation process of the answer text and the corpus to obtain a number of words includes: Clean the generated text data, remove punctuation marks and blank characters in the text and retain numbers and special characters, use the UTF-8 encoding format to uniformly encode the text data, and normalize the text data; Use the forward maximum matching algorithm to segment the answer text and each text in the corpus to obtain several words.
3. The occupational skill assessment method based on big data according to claim 1 is characterized in that: The part-of-speech tagging and clustering of all words includes: Use natural language processing to tag each word based on its context; All words are clustered according to their parts of speech using a preset clustering algorithm to obtain noun, verb and adjective classes.
4. The method for evaluating occupational skills based on big data according to claim 1, characterized in that: The calculation method of the TF-IDF correction value is replaced by: ,in, Representing words The TF-IDF correction value of Representing words The similarity of the meaning of Representing words The distance similarity of Representing words The initial TF-IDF value of the word is the target word; Calculate the difference between the position information of the target word in the answer text and the position information of the target word in the corpus, and use the result of negative correlation mapping of the difference as the distance similarity of the target word; The position information of the target word in the answer text includes: Use word embedding models to convert each word into a word vector; Take the first word in each word group as the starting word; The position information of the target word in the answer text satisfies the relationship: ,in, Representing words Position information in the answer text, Representing words With The Euclidean distance between the starting words in the word group, Indicates The total number of words in the word group, Indicates the total number of target word groups, word is the target word; According to the method for obtaining the position information of the target word in the answer text, the position information of the target word in the corpus is obtained.
5. The vocational skills assessment system based on big data is characterized by: include: Processor; and A memory storing computer instructions, which, when executed by a processor, enables the device to execute the big data-based professional skills assessment method according to any one of claims 1 to 4.
