Recommendation system for innovation and entrepreneurship education based on semantic analysis

Abstract

The invention discloses an innovation and entrepreneurship education recommendation system based on semantic analysis, and belongs to the technical field of semantic processing. The method comprises the following steps of: acquiring an entrepreneurship education background height, a professional height, a background height and a professional noun set of a to-be-entrepreneurship direction through an evaluation unit, and respectively calculating an education background integrating degree, a professional integrating degree and a background integrating degree in combination with user education background, professional information and family employment information; the related text is mapped into a vector by using a word embedding unit, and semantic analysis processing is performed by means of a multilayer LSTM neural network in combination with the height of each dimension and the integrating degree to obtain the fitness of the to-be-started direction. And finally, the recommendation unit selects the direction corresponding to the maximum fitness as the optimal entrepreneurial direction to be recommended to the user, so that the problem of low entrepreneurial direction recommendation accuracy in the prior art is effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of semantic processing, and in particular to a recommendation system for innovative entrepreneurship education based on semantic analysis. Background Art

[0002] In the current field of innovation and entrepreneurship education, colleges and universities and entrepreneurship guidance institutions generally face the challenge of personalized entrepreneurship guidance. Traditional entrepreneurship guidance methods mainly rely on empirical manual evaluation, usually by mentors or experts to make subjective judgments based on user resumes and entrepreneurial intentions. This method has significant limitations: the lack of a systematic evaluation framework makes it difficult to comprehensively analyze the user background, professional ability and entrepreneurial direction of entrepreneurs, making it impossible to accurately recommend personal entrepreneurial directions.

[0003] With the development of technology, existing entrepreneurial direction recommendations have also begun to adopt semantic analysis models, which analyze industry reports, social media data and other information to mine existing popular entrepreneurial directions and recommend them to individuals. However, this method has obvious flaws: first, it only focuses on popular trends and does not consider individual user preferences. The recommended direction may not match the user's actual interests; second, there is a lack of a quantitative evaluation mechanism for the adaptability of users and entrepreneurial directions, and it is impossible to measure the adaptability of individuals in specific entrepreneurial directions, which ultimately leads to a low accuracy rate in entrepreneurial direction recommendations. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a recommendation system for innovation and entrepreneurship education based on semantic analysis, which solves the problem of low accuracy in recommending entrepreneurship directions in the prior art.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a recommendation system for innovation and entrepreneurship education based on semantic analysis, comprising: an input unit, an evaluation unit, an academic qualification fit acquisition unit, a professional fit acquisition unit, a background fit acquisition unit, a word embedding unit, a semantic analysis unit and a recommendation unit;

[0006] The input unit is used to input multiple texts of the user's business directions to be started;

[0007] The evaluation unit is used to evaluate each entrepreneurial direction to be started, and obtain the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level and entrepreneurial professional terminology set of each entrepreneurial direction to be started;

[0008] The educational qualification matching acquisition unit is used to obtain educational qualification matching according to the difference between the educational qualification of the user and the educational qualification of the entrepreneur;

[0009] The professional fit acquisition unit is used to acquire professional fit according to user professional information, entrepreneurial professional term set and entrepreneurial professional height;

[0010] The background compatibility acquisition unit is used to acquire the background compatibility according to the user's family employment information, the business scope of the company to be started and the height of the entrepreneurial background;

[0011] The word embedding unit is used to map the text corresponding to the user's education level, the user's professional information, and the user's family's employment information into vectors, thereby obtaining the user's education level vector, the user's professional information vector, and the user's family's employment information vector;

[0012] The semantic analysis unit is used to use a multi-layer LSTM neural network to combine the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level, educational level fit, professional fit and background fit, and perform semantic analysis on the user's educational level vector, user professional information vector and user's family employment information vector to obtain the adaptability of the entrepreneurial direction;

[0013] The recommendation unit is used to select the entrepreneurial direction corresponding to the maximum fitness from the fitness of multiple entrepreneurial directions as the optimal entrepreneurial direction for the user, and recommend the optimal entrepreneurial direction to the user.

[0014] Furthermore, the process of obtaining the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level and entrepreneurial professional terminology set of the entrepreneurial direction to be started includes:

[0015] Collect companies in the same industry as the business you want to start, and obtain the educational background and professional information of the company's shareholders and the employment information of their family members;

[0016] According to the educational background information of shareholders of each company, obtain the educational background of the entrepreneurs to be started;

[0017] According to the professional text information of shareholders of each company, obtain the set of entrepreneurial professional terms and the entrepreneurial professional level of the direction to be started;

[0018] Based on the employment information of shareholders and family members of each company, the entrepreneurial background of the proposed business direction is obtained.

[0019] Furthermore, the process of obtaining the entrepreneurial academic qualifications of the direction to be started includes:

[0020] Different academic scores are assigned to each degree in ascending order of educational attainment;

[0021] Under each educational background, count the number of shareholders of all companies with the corresponding educational background to get the number of people with the educational background;

[0022] Multiply the number of people with academic qualifications by the corresponding academic qualification score to obtain the weighted value, sum up the weighted values, and then take the average of the number of shareholders of all companies to obtain the entrepreneurial academic qualification level.

[0023] Furthermore, the process of obtaining the set of entrepreneurial professional terms and the entrepreneurial professional height of the entrepreneurial direction to be started includes:

[0024] Merge the professional text information of shareholders belonging to the same company to obtain the professional text information of the company;

[0025] Count the frequency of the same professional term appearing in the professional text information of each company in all the professional text information of all companies;

[0026] According to the frequency, various professional terms are sorted in descending order, and the first N professional terms are extracted to form a set of entrepreneurial professional terms, where N is a positive integer greater than 2;

[0027] Take the intersection of each professional term in each company's professional text information and the entrepreneurial professional term set;

[0028] The ratio of the number of companies whose intersection is not empty to the total number of companies is used as the entrepreneurial professional level of the entrepreneurial direction to be started.

[0029] Furthermore, the process of obtaining the entrepreneurial background height of the entrepreneurial direction is as follows:

[0030] Merge the employment information of the family members of the shareholders of the same company to obtain the company's family employment information;

[0031] For the same company, take the intersection of the keywords of the company's business scope and the keywords of the company's family employment information;

[0032] The ratio of the number of companies whose intersection is not empty to the total number of companies is used as the entrepreneurial background height of the entrepreneurial direction to be started.

[0033] Furthermore, the process of obtaining academic qualifications includes:

[0034] Assign an academic score to the user's academic qualifications to obtain the user's academic level;

[0035] The absolute value of the difference between the user's educational level and the entrepreneur's educational level is taken as the educational level gap:

[0036] Calculate the degree of academic compatibility based on the academic gap: , where r h is the academic qualification fit, h p is the user's educational level, h I is the entrepreneurial education level, | | is the absolute value operation, w max is the maximum academic score, w min Score the maximum degree.

[0037] Further, the process of obtaining professional fit includes:

[0038] When the entrepreneurial professional height is less than the professional height threshold, the professional fit is assigned a value of 0;

[0039] When the entrepreneurial professional height is greater than or equal to the professional height threshold, the intersection of each professional term in the user's professional information and the entrepreneurial professional term set is taken as the user's professional intersection;

[0040] The frequency of each professional term in the intersection of user professions in all company professional text information is added up and normalized to obtain the professional fit.

[0041] Furthermore, the process of obtaining background fit includes:

[0042] When the entrepreneurial background height is less than the background height threshold, the background fit is assigned a value of 0;

[0043] When the entrepreneurial background height is greater than or equal to the background height threshold, the intersection of the keywords of the user's family's employment information and the keywords of all company's business scopes is taken as the user background intersection;

[0044] The number of keywords in the intersection of user backgrounds is counted and normalized to obtain the background fit.

[0045] Further, the multi-layer LSTM neural network includes: a first LSTM unit, a second LSTM unit, a third LSTM unit, a feature enhancement layer A1, a feature enhancement layer A2, a feature enhancement layer A3, a feature enhancement layer A4, a feature enhancement layer A5, a feature enhancement layer A6, a first Concat layer, a second Concat layer, a first BiLSTM unit, a second BiLSTM unit and a fully connected layer;

[0046] The input end of the first LSTM unit is used to input the user's education vector, and its output end is connected to the first input end of the feature enhancement layer A1 and the first input end of the feature enhancement layer A2 respectively; the input end of the second LSTM unit is used to input the user's professional information vector, and its output end is connected to the first input end of the feature enhancement layer A3 and the first input end of the feature enhancement layer A4 respectively; the input end of the third LSTM unit is used to input the user's family occupation information vector, and its output end is connected to the first input end of the feature enhancement layer A5 and the first input end of the feature enhancement layer A6 respectively;

[0047] The second input end of the feature enhancement layer A1 is used to input the degree of academic qualification; the second input end of the feature enhancement layer A3 is used to input the degree of professional qualification; the second input end of the feature enhancement layer A5 is used to input the degree of background qualification; the second input end of the feature enhancement layer A2 is used to input the degree of entrepreneurial academic qualification; the second input end of the feature enhancement layer A4 is used to input the degree of entrepreneurial professional qualification; the second input end of the feature enhancement layer A6 is used to input the degree of entrepreneurial background;

[0048] The input end of the first Concat layer is respectively connected to the output end of feature enhancement layer A1, the output end of feature enhancement layer A3 and the output end of feature enhancement layer A5, and its output end is connected to the input end of the first BiLSTM unit; the input end of the second Concat layer is respectively connected to the output end of feature enhancement layer A2, the output end of feature enhancement layer A4 and the output end of feature enhancement layer A6, and its output end is connected to the input end of the second BiLSTM unit; the input end of the fully connected layer is respectively connected to the output end of the first BiLSTM unit and the output end of the second BiLSTM unit.

[0049] Furthermore, the expressions of the feature enhancement layers are: , where x out is the output of the feature enhancement layer, x 1,in is the input of the first input terminal of the feature enhancement layer, x 2,in is the input of the second input terminal of the feature enhancement layer, ω is the weight, b is the bias, and f is the activation function.

[0050] The beneficial effects of the present invention are:

[0051] 1. The present invention inputs multiple texts of the user's desired entrepreneurial directions through an input unit, so that the recommendation of the optimal entrepreneurial direction is carried out within the user's expected range, thereby improving the accuracy of the recommendation.

[0052] 2. The present invention obtains the entrepreneurial academic qualifications, entrepreneurial professional qualifications and entrepreneurial backgrounds, thereby evaluating the educational qualifications, professional levels and background correlation levels of the entrepreneurial group in the entrepreneurial direction, and then obtains the educational qualifications, professional qualifications and background correlation levels based on the gap between the user's educational qualifications and the entrepreneurial educational qualifications, the correlation between the user's professional information and the entrepreneurial professional terminology set and professional height, and the relationship between the user's family employment information and the business scope and background height of the company in the entrepreneurial direction, to comprehensively analyze the matching degree between the entrepreneur's educational qualifications, professional abilities, background and entrepreneurial direction.

[0053] 3. The present invention uses a multi-layer LSTM neural network to extract semantic features from user education, user professional information, and user family employment information, and combines the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level, educational level fit, professional fit, and background fit to strengthen the semantic features, thereby improving the accuracy of entrepreneurial direction evaluation. The present invention fully considers the individual preferences and individual circumstances of users, selects the entrepreneurial direction corresponding to the maximum fitness within the user's expected range and recommends it to the user, thereby improving the accuracy of entrepreneurial direction recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flowchart of a recommendation system for innovative entrepreneurship education based on semantic analysis;

[0055] Figure 2 Schematic diagram of the structure of a multi-layer LSTM neural network. DETAILED DESCRIPTION

[0056] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0057] like Figure 1 As shown, a recommendation system for innovative entrepreneurship education based on semantic analysis includes: an input unit, an evaluation unit, an academic qualification matching acquisition unit, a professional matching acquisition unit, a background matching acquisition unit, a word embedding unit, a semantic analysis unit and a recommendation unit;

[0058] The input unit is used to input multiple texts of the user's business directions to be started;

[0059] The evaluation unit is used to evaluate each entrepreneurial direction to be started, and obtain the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level and entrepreneurial professional terminology set of each entrepreneurial direction to be started;

[0060] The educational qualification matching acquisition unit is used to obtain educational qualification matching according to the difference between the educational qualification of the user and the educational qualification of the entrepreneur;

[0061] The professional fit acquisition unit is used to acquire professional fit according to user professional information, entrepreneurial professional term set and entrepreneurial professional height;

[0062] The background compatibility acquisition unit is used to acquire the background compatibility according to the user's family employment information, the business scope of the company to be started and the height of the entrepreneurial background;

[0063] The word embedding unit is used to map the text corresponding to the user's education level, the user's professional information, and the user's family's employment information into vectors, thereby obtaining the user's education level vector, the user's professional information vector, and the user's family's employment information vector;

[0064] The semantic analysis unit is used to use a multi-layer LSTM neural network to combine the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level, educational level fit, professional fit and background fit, and perform semantic analysis on the user's educational level vector, user professional information vector and user's family employment information vector to obtain the adaptability of the entrepreneurial direction;

[0065] The recommendation unit is used to select the entrepreneurial direction corresponding to the maximum fitness from the fitness of multiple entrepreneurial directions as the optimal entrepreneurial direction for the user, and recommend the optimal entrepreneurial direction to the user.

[0066] In this embodiment, the process of obtaining the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level and entrepreneurial professional term set of the entrepreneurial direction to be started includes:

[0067] Collect companies in the same industry as the business you want to start, and obtain the educational background and professional information of the company's shareholders and the employment information of their family members;

[0068] According to the educational background information of shareholders of each company, obtain the educational background of the entrepreneurs to be started;

[0069] According to the professional text information of shareholders of each company, obtain the set of entrepreneurial professional terms and the level of entrepreneurial professionalism of the direction to be started;

[0070] Based on the employment information of shareholders and family members of each company, the entrepreneurial background of the proposed business direction is obtained.

[0071] The entrepreneurial directions to be explored include: artificial intelligence and big data, Internet of Things, software development, medical services, health management, pharmaceutical health care, film and television production, game development, etc.

[0072] Companies in the same industry as the business direction to be started refer to those companies that are highly similar or related to the business direction that the entrepreneur plans to start in terms of business areas, products or services provided, target customer groups, and industry norms and standards followed. For example: the business direction to be started is: healthy light food restaurant.

[0073] In this embodiment, the process of obtaining the entrepreneurial education level of the entrepreneurial direction to be started includes:

[0074] Different academic scores are assigned to each degree in ascending order of educational attainment;

[0075] Under each educational background, count the number of shareholders of all companies with the corresponding educational background to get the number of people with the educational background;

[0076] Multiply the number of people with academic qualifications by the corresponding academic qualification score to get the weighted value, sum up the weighted values, and then take the average of the number of shareholders of all companies to get the entrepreneurial academic qualifications: , where h I For entrepreneurial education, i is the academic score of the i-th academic degree, K i,p is the number of people with the ith academic degree, K sum is the total number of company shareholders, i is a positive integer, and L is the number of educational background types.

[0077] In this embodiment, the elementary school degree score is 0.1, the middle school degree score is 0.2, the high school degree score is 0.3, the university degree score is 0.4, the master's degree score is 0.5, and the doctoral degree score is 0.6.

[0078] In the present invention, the higher the entrepreneurial academic qualifications, the higher the academic qualifications required in the entrepreneurial direction, and the higher the knowledge reserves required for entrepreneurship in the industry.

[0079] In this embodiment, the process of obtaining the entrepreneurial professional term set and the entrepreneurial professional height of the entrepreneurial direction includes:

[0080] Merge the professional text information of shareholders belonging to the same company to obtain the professional text information of the company;

[0081] Count the frequency of the same professional term appearing in the professional text information of each company in all the professional text information of all companies;

[0082] According to the frequency, various professional terms are sorted in descending order, and the first N professional terms are extracted to form a set of entrepreneurial professional terms, where N is a positive integer greater than 2;

[0083] Take the intersection of each professional term in each company's professional text information and the entrepreneurial professional term set;

[0084] The ratio of the number of companies whose intersection is not empty to the total number of companies is used as the entrepreneurial professional level of the entrepreneurial direction to be started.

[0085] The professional text information is the text information (names of various majors) of the majors studied by the personnel.

[0086] In this embodiment, the professional terms are keywords in the professional text information.

[0087] The present invention merges the professional text information of shareholders belonging to the same company, presents the professional information of the same company, then counts the frequency of professional terms and extracts the first N to form a set of entrepreneurial professional terms, locks the core profession of the entrepreneurial direction, and then takes the intersection, which can intuitively reflect the concentration of professional fields of companies in the industry. The ratio of the number of companies whose intersection is not empty to the number of overall companies is taken as the entrepreneurial professional height of the entrepreneurial direction. This ratio can intuitively reflect how many companies' professional fields are associated with the entrepreneurial professional term set in the entire industry. If the ratio is high, it means that most companies are involved in these professional terms, indicating that the professional concentration of the entrepreneurial direction is high, and companies in the industry are more concentrated and convergent in profession.

[0088] In this embodiment, the process of obtaining the entrepreneurial background height of the entrepreneurial direction to be started is:

[0089] Merge the employment information of the family members of the shareholders of the same company to obtain the company's family employment information;

[0090] For the same company, take the intersection of the keywords of the company's business scope and the keywords of the company's family employment information;

[0091] The ratio of the number of companies whose intersection is not empty to the total number of companies is used as the entrepreneurial background height of the entrepreneurial direction to be started.

[0092] The keywords of the company's business scope reflect the actual business areas of the company, while the employment information of shareholders' family members reflects, to a certain extent, the industry resources, personal connections, and potential business inheritance that shareholders may have. By calculating the intersection of the two keywords, we can understand how many companies' business scopes are related to the shareholders' family's employment background. The higher the ratio, the higher the degree of fit between the company's business development and the shareholders' family's employment background in this entrepreneurial direction. The entrepreneurial backgrounds of companies in the industry are relatively concentrated, and there may be certain family inheritance, resource sharing, or industry barriers.

[0093] A higher entrepreneurial background means that the entrepreneurial direction may be highly dependent on the family background of the entrepreneur. For example, some traditional manufacturing industries or highly professional fields may require entrepreneurs to have a family background in related industries in order to obtain technology, channels, experience and other resources. On the contrary, a lower entrepreneurial background indicates that the entrepreneurial direction is relatively more open.

[0094] The employment information of family members and the user's family includes: occupation category, professional information, business scope of the work unit, employment history, etc.

[0095] In this embodiment, the process of obtaining the degree of academic qualification compatibility includes:

[0096] Assign an academic score to the user's academic qualifications to obtain the user's academic level;

[0097] The absolute value of the difference between the user's educational level and the entrepreneur's educational level is taken as the educational level gap:

[0098] Calculate the degree of academic compatibility based on the academic gap: , where r h is the academic qualification fit, h p is the user's educational level, h I is the entrepreneurial education level, | | is the absolute value operation, w max is the maximum academic score, w min Score the maximum degree.

[0099] In the present invention, if there are multiple academic qualifications, the highest academic qualification is selected and the academic qualification compatibility is calculated.

[0100] The greater the degree of academic qualification fit, the more consistent the user's academic qualification is with the academic qualifications of the group of people in this entrepreneurial direction.

[0101] In this embodiment, the process of obtaining professional compatibility includes:

[0102] When the entrepreneurial professional height is less than the professional height threshold, the professional fit is assigned a value of 0;

[0103] When the entrepreneurial professional height is greater than or equal to the professional height threshold, the intersection of each professional term in the user's professional information and the entrepreneurial professional term set is taken as the user's professional intersection;

[0104] The frequency of each professional term in the user's professional intersection in all company professional text information is added up and normalized to obtain the professional fit: , where r s For professional fit, s i is the frequency of the i-th professional noun in the user's professional intersection in all company professional text information, i is a positive integer, Z is the sum of the frequencies of all professional nouns in the entrepreneurial professional noun set in all company professional text information, and M is the number of professional nouns in the user's professional intersection.

[0105] In this embodiment, the professional height threshold and the background height threshold can be set to 0.5, and the specific values ​​can be adjusted according to experiments or needs.

[0106] When the entrepreneurial professional height is less than the professional height threshold, it means that the entrepreneurial direction has low requirements for professionalism. Therefore, the professional fit is assigned a value of 0. When the entrepreneurial professional height is greater than or equal to the professional height threshold, it means that the entrepreneurial direction has high requirements for professionalism. According to the situation of professional terms in the intersection of user majors, the professional fit is obtained to evaluate the fit between the user's major and the entrepreneurial direction.

[0107] In this embodiment, the process of obtaining the background consistency includes:

[0108] When the entrepreneurial background height is less than the background height threshold, the background fit is assigned a value of 0;

[0109] When the entrepreneurial background height is greater than or equal to the background height threshold, the intersection of the keywords of the user's family's employment information and the keywords of all companies' business scopes is taken as the user background intersection, where all companies here refer to companies in the same industry as the entrepreneurial direction during the statistical process;

[0110] The number of keywords in the intersection of user backgrounds is counted and normalized to obtain the background fit.

[0111] In this embodiment, the number of keywords is normalized by dividing the number of keywords by the number of keywords in the user's family occupation information.

[0112] When the entrepreneurial background height is less than the background height threshold, it means that the entrepreneurial direction has low requirements for background resources. Therefore, the background fit is assigned a value of 0. When the entrepreneurial background height is greater than or equal to the background height threshold, it means that the entrepreneurial direction has high requirements for background resources. According to the keywords in the intersection of the user background, the background fit is obtained to evaluate the fit between the user's background and the entrepreneurial direction.

[0113] like Figure 2 As shown, the multi-layer LSTM neural network includes: a first LSTM unit, a second LSTM unit, a third LSTM unit, a feature reinforcement layer A1, a feature reinforcement layer A2, a feature reinforcement layer A3, a feature reinforcement layer A4, a feature reinforcement layer A5, a feature reinforcement layer A6, a first Concat layer, a second Concat layer, a first BiLSTM unit, a second BiLSTM unit and a fully connected layer;

[0114] The input end of the first LSTM unit is used to input the user's education vector, and its output end is connected to the first input end of the feature enhancement layer A1 and the first input end of the feature enhancement layer A2 respectively; the input end of the second LSTM unit is used to input the user's professional information vector, and its output end is connected to the first input end of the feature enhancement layer A3 and the first input end of the feature enhancement layer A4 respectively; the input end of the third LSTM unit is used to input the user's family occupation information vector, and its output end is connected to the first input end of the feature enhancement layer A5 and the first input end of the feature enhancement layer A6 respectively;

[0115] The second input end of the feature enhancement layer A1 is used to input the degree of academic qualification; the second input end of the feature enhancement layer A3 is used to input the degree of professional qualification; the second input end of the feature enhancement layer A5 is used to input the degree of background qualification; the second input end of the feature enhancement layer A2 is used to input the degree of entrepreneurial academic qualification; the second input end of the feature enhancement layer A4 is used to input the degree of entrepreneurial professional qualification; the second input end of the feature enhancement layer A6 is used to input the degree of entrepreneurial background;

[0116] The input end of the first Concat layer is respectively connected to the output end of feature enhancement layer A1, the output end of feature enhancement layer A3 and the output end of feature enhancement layer A5, and its output end is connected to the input end of the first BiLSTM unit; the input end of the second Concat layer is respectively connected to the output end of feature enhancement layer A2, the output end of feature enhancement layer A4 and the output end of feature enhancement layer A6, and its output end is connected to the input end of the second BiLSTM unit; the input end of the fully connected layer is respectively connected to the output end of the first BiLSTM unit and the output end of the second BiLSTM unit.

[0117] In this embodiment, the expressions of the feature enhancement layers are: , where x out is the output of the feature enhancement layer, x 1,in is the input of the first input terminal of the feature enhancement layer, x 2,in is the input of the second input terminal of the feature enhancement layer, ω is the weight, b is the bias, f is the activation function, and the activation function can be the Sigmoid function.

[0118] The present invention first processes the text vectors of user education, user professional information and user family employment information through three LSTM units to realize preliminary feature extraction, and then uses two feature enhancement layers after each LSTM unit to enhance the expressiveness of key features, and then splices the outputs of feature enhancement layers A1, A3 and A5 respectively, and inputs them into the first BiLSTM unit, and splices the outputs of feature enhancement layers A2, A4 and A6, and inputs them into the second BiLSTM unit, so as to realize comprehensive feature extraction of features under two different influencing factors (fit and height), capture the relationship between semantic features, and then synthesize the outputs of the first BiLSTM unit and the second BiLSTM unit through a fully connected layer to obtain fitness.

[0119] The weights and biases in the multi-layer LSTM neural network are trained using the existing gradient descent method.

[0120] The present invention inputs multiple texts of entrepreneurial directions to be proposed by the user through an input unit, so that the recommendation of the optimal entrepreneurial direction is carried out within the expected range of the user, thereby improving the accuracy of the recommendation.

[0121] In this embodiment, each encoding value in the user education vector, the user professional information vector and the family employment information vector is input into a cell unit of the LSTM unit.

[0122] The present invention obtains the entrepreneurial academic qualifications, entrepreneurial professional qualifications and entrepreneurial background qualifications, thereby evaluating the educational qualifications, professional levels and background correlation levels of the entrepreneurial group in the entrepreneurial direction, and then obtains the educational qualifications, professional qualifications and background correlation levels based on the gap between the user's educational qualifications and the entrepreneurial educational qualifications, the correlation between the user's professional information and the entrepreneurial professional terminology set and professional height, and the relationship between the user's family employment information and the business scope and background height of the company in the entrepreneurial direction, to comprehensively analyze the matching degree between the entrepreneur's educational qualifications, professional ability, background and entrepreneurial direction.

[0123] The present invention uses a multi-layer LSTM neural network to extract semantic features from user education, user professional information, and user family employment information, and combines the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level, educational level fit, professional fit, and background fit to strengthen the semantic features, thereby improving the accuracy of entrepreneurial direction evaluation. The present invention fully considers the individual preferences and individual circumstances of users, selects the entrepreneurial direction corresponding to the maximum fitness within the user's expected range and recommends it to the user, thereby improving the accuracy of entrepreneurial direction recommendation.

[0124] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A recommendation system for innovative entrepreneurship education based on semantic analysis, characterized in that: include: Input unit, evaluation unit, academic qualification matching acquisition unit, professional matching acquisition unit, background matching acquisition unit, word embedding unit, semantic analysis unit and recommendation unit; The input unit is used to input multiple texts of the user's business directions to be started; The evaluation unit is used to evaluate each entrepreneurial direction to be started, and obtain the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level and entrepreneurial professional terminology set of each entrepreneurial direction to be started; The educational qualification matching acquisition unit is used to obtain educational qualification matching according to the difference between the educational qualification of the user and the educational qualification of the entrepreneur; The professional fit acquisition unit is used to acquire professional fit according to user professional information, entrepreneurial professional term set and entrepreneurial professional height; The background compatibility acquisition unit is used to acquire the background compatibility according to the user's family employment information, the business scope of the company to be started and the height of the entrepreneurial background; The word embedding unit is used to map the text corresponding to the user's education level, the user's professional information, and the user's family's employment information into vectors, thereby obtaining the user's education level vector, the user's professional information vector, and the user's family's employment information vector; The semantic analysis unit is used to use a multi-layer LSTM neural network to combine the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level, educational level fit, professional fit and background fit, and perform semantic analysis on the user's educational level vector, user professional information vector and user's family employment information vector to obtain the adaptability of the entrepreneurial direction; The recommendation unit is used to select the entrepreneurial direction corresponding to the maximum fitness from the fitness of multiple entrepreneurial directions as the optimal entrepreneurial direction for the user, and recommend the optimal entrepreneurial direction to the user.

2. The recommendation system for innovation and entrepreneurship education based on semantic analysis according to claim 1 is characterized in that: The process of obtaining the entrepreneurial education level, entrepreneurial professional level, entrepreneurial background level and entrepreneurial professional terminology set of the entrepreneurial direction to be started includes: Collect companies in the same industry as the business you want to start, and obtain the educational background and professional information of the company's shareholders and the employment information of their family members; According to the educational background information of shareholders of each company, obtain the educational background of the entrepreneurs to be started; According to the professional text information of shareholders of each company, obtain the set of entrepreneurial professional terms and the entrepreneurial professional level of the direction to be started; Based on the employment information of shareholders and family members of each company, the entrepreneurial background of the proposed business direction is obtained.

3. The recommendation system for innovation and entrepreneurship education based on semantic analysis according to claim 2 is characterized in that: The process of obtaining the degree of entrepreneurship required for the business to be started includes: Different academic scores are assigned to each degree in ascending order of educational attainment; Under each educational background, count the number of shareholders of all companies with the corresponding educational background to get the number of people with the educational background; Multiply the number of people with academic qualifications by the corresponding academic qualification score to obtain the weighted value, sum up the weighted values, and then take the average of the number of shareholders of all companies to obtain the entrepreneurial academic qualification level.

4. The recommendation system for innovation and entrepreneurship education based on semantic analysis according to claim 2 is characterized in that: The process of obtaining the set of entrepreneurial professional terms and the height of entrepreneurial professional knowledge for the entrepreneurial direction to be started includes: Merge the professional text information of shareholders belonging to the same company to obtain the professional text information of the company; Count the frequency of the same professional term appearing in the professional text information of each company in all the professional text information of all companies; According to the frequency, various professional terms are sorted in descending order, and the first N professional terms are extracted to form a set of entrepreneurial professional terms, where N is a positive integer greater than 2; Take the intersection of each professional term in each company's professional text information and the entrepreneurial professional term set; The ratio of the number of companies whose intersection is not empty to the total number of companies is used as the entrepreneurial professional level of the entrepreneurial direction to be started.

5. The recommendation system for innovation and entrepreneurship education based on semantic analysis according to claim 2 is characterized in that: The process of obtaining the entrepreneurial background height of the entrepreneurial direction to be started is: Merge the employment information of the family members of the shareholders of the same company to obtain the company's family employment information; For the same company, take the intersection of the keywords of the company's business scope and the keywords of the company's family employment information; The ratio of the number of companies whose intersection is not empty to the total number of companies is used as the entrepreneurial background height of the entrepreneurial direction to be started.

6. The recommendation system for innovation and entrepreneurship education based on semantic analysis according to claim 1, characterized in that: The process of obtaining academic fit includes: Assign an academic score to the user's academic qualifications to obtain the user's academic level; The absolute value of the difference between the user's educational level and the entrepreneur's educational level is taken as the educational level gap: Calculate the degree of academic compatibility based on the academic gap: , where r h is the academic qualification fit, h p is the user's educational level, h I is the entrepreneurial education level, | | is the absolute value operation, w max is the maximum academic score, w min Score the maximum degree.

7. The recommendation system for innovation and entrepreneurship education based on semantic analysis according to claim 1, characterized in that: The process of obtaining professional fit includes: When the entrepreneurial professional height is less than the professional height threshold, the professional fit is assigned a value of 0; When the entrepreneurial professional height is greater than or equal to the professional height threshold, the intersection of each professional term in the user's professional information and the entrepreneurial professional term set is taken as the user's professional intersection; The frequency of each professional term in the user's professional intersection in all company professional text information is added up and normalized to obtain the professional fit.

8. The recommendation system for innovation and entrepreneurship education based on semantic analysis according to claim 1, characterized in that: The process of obtaining background fit includes: When the entrepreneurial background height is less than the background height threshold, the background fit is assigned a value of 0; When the entrepreneurial background height is greater than or equal to the background height threshold, the intersection of the keywords of the user's family's employment information and the keywords of all company's business scopes is taken as the user background intersection; The number of keywords in the intersection of user backgrounds is counted and normalized to obtain the background fit.

9. The recommendation system for innovation and entrepreneurship education based on semantic analysis according to claim 1, characterized in that: The multi-layer LSTM neural network includes: a first LSTM unit, a second LSTM unit, a third LSTM unit, a feature reinforcement layer A1, a feature reinforcement layer A2, a feature reinforcement layer A3, a feature reinforcement layer A4, a feature reinforcement layer A5, a feature reinforcement layer A6, a first Concat layer, a second Concat layer, a first BiLSTM unit, a second BiLSTM unit and a fully connected layer; The input end of the first LSTM unit is used to input the user's education vector, and its output end is connected to the first input end of the feature enhancement layer A1 and the first input end of the feature enhancement layer A2 respectively; the input end of the second LSTM unit is used to input the user's professional information vector, and its output end is connected to the first input end of the feature enhancement layer A3 and the first input end of the feature enhancement layer A4 respectively; the input end of the third LSTM unit is used to input the user's family occupation information vector, and its output end is connected to the first input end of the feature enhancement layer A5 and the first input end of the feature enhancement layer A6 respectively; The second input end of the feature enhancement layer A1 is used to input the degree of academic qualification; the second input end of the feature enhancement layer A3 is used to input the degree of professional qualification; the second input end of the feature enhancement layer A5 is used to input the degree of background qualification; the second input end of the feature enhancement layer A2 is used to input the degree of entrepreneurial academic qualification; the second input end of the feature enhancement layer A4 is used to input the degree of entrepreneurial professional qualification; the second input end of the feature enhancement layer A6 is used to input the degree of entrepreneurial background; The input end of the first Concat layer is respectively connected to the output end of feature enhancement layer A1, the output end of feature enhancement layer A3 and the output end of feature enhancement layer A5, and its output end is connected to the input end of the first BiLSTM unit; the input end of the second Concat layer is respectively connected to the output end of feature enhancement layer A2, the output end of feature enhancement layer A4 and the output end of feature enhancement layer A6, and its output end is connected to the input end of the second BiLSTM unit; the input end of the fully connected layer is respectively connected to the output end of the first BiLSTM unit and the output end of the second BiLSTM unit.

10. The recommendation system for innovation and entrepreneurship education based on semantic analysis according to claim 9, characterized in that: The expressions of the feature enhancement layer are: , where x out is the output of the feature enhancement layer, x 1,in is the input of the first input terminal of the feature enhancement layer, x 2,in is the input of the second input terminal of the feature enhancement layer, ω is the weight, b is the bias, and f is the activation function.

Images