Method for matching business requirements and matching device for business requirements

Abstract

The application discloses a service demand matching method and device. The matching method comprises the following steps: obtaining service demand information; obtaining a preset service word library, wherein the preset service word library comprises a plurality of first-level matching items and a plurality of second-level matching items, each first-level matching item corresponds to at least one second-level matching item; performing a vector conversion operation on the service demand information, the first-level matching items and the second-level matching items to obtain corresponding vectorized demand, vectorized first-level matching items and vectorized second-level matching items; matching the vectorized demand with the vectorized first-level matching items and the vectorized second-level matching items to obtain specific first-level matching items and specific second-level matching items matched with the service demand information; and generating a demand matching result based on the specific first-level matching items and the specific second-level matching items.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of business matching, in particular to a business demand matching method, a business demand matching device, a processor, a machine readable storage medium and a computer program product. BACKGROUND

[0002] With the continuous development of domestic economy and the continuous improvement of economic level, people's life is getting richer and richer, which leads to the continuous development of bank business. In order to meet different needs, banks set up a large number of businesses. In the process of handling the above-mentioned businesses, the bank clerks need to determine a plurality of demand items according to the actual needs of the customers, and search for the corresponding three-level activities and four-level tasks of each demand item through fuzzy search, in order to handle the corresponding business for the user. However, there are thousands of three-level activities and four-level tasks in the existing bank system, so in the actual application process, the technical personnel found that the existing demand matching method at least has the following technical problems:

[0003] On the one hand, since each bank clerk needs to manually match the accurate three-level activities and four-level tasks for the user's demand, a large amount of business training needs to be carried out on the bank clerks in advance, and when the number of demand items is large, multiple bank clerks must be configured, thus greatly increasing the training cost and labor cost of the enterprise;

[0004] On the other hand, although the bank clerks can match multiple demand items through fuzzy search, the manual operation is low in efficiency, and for the three-level activities or four-level tasks with similar names, the correct matching items need to be further checked by viewing their detailed information, thus further reducing the work efficiency. SUMMARY

[0005] In order to overcome the above technical problems existing in the prior art, the embodiments of the present application provide a business demand matching method, which carries out vectorization processing on the existing demand items and business matching items, and calculates the business matching items matched with each demand item in a vectorization manner, thereby improving the matching efficiency and improving the matching accuracy.

[0006] To achieve the above object, the embodiment of the present application provides a service demand matching method, which comprises: obtaining service demand information; obtaining a preset service vocabulary library, wherein the preset service vocabulary library comprises a plurality of first-level matching items and a plurality of second-level matching items, each first-level matching item corresponds to at least one second-level matching item; performing a vector conversion operation on the service demand information, the first-level matching items and the second-level matching items to obtain corresponding vectorized demand, vectorized first-level matching items and vectorized second-level matching items; matching the vectorized demand with the vectorized first-level matching items and the vectorized second-level matching items to obtain specific first-level matching items and specific second-level matching items matched with the service demand information; and generating a demand matching result based on the specific first-level matching items and the specific second-level matching items.

[0007] Preferably, the obtaining of the preset service vocabulary library comprises: establishing an initial vocabulary library; encoding each word in the initial vocabulary library based on a preset encoding rule to obtain an encoded word; performing vector conversion on each encoded word to obtain a corresponding word vector; and establishing a preset service vocabulary library based on the word vector.

[0008] Preferably, the establishing of the initial vocabulary library comprises: obtaining historical demand information in a preset time period, extracting a first noun set and a first verb set of the historical demand information; obtaining a second noun set of all first-level matching items and a third noun set of all second-level matching items; obtaining a preset verb set; and generating the initial vocabulary library based on the first noun set, the second noun set, the third noun set, the first verb set and the preset verb set.

[0009] Preferably, the performing of the vector conversion on each encoded word to obtain a corresponding word vector comprises: determining a first initial matrix and a second initial matrix, wherein the number of rows of the first initial matrix is equal to the number of columns of the second initial matrix, and the number of columns of the first initial matrix is equal to the number of rows of the second initial matrix; performing a first processing on each encoded word based on the first initial matrix to obtain a first-processed vector corresponding to each encoded word; performing a second processing on each first-processed vector based on the second initial matrix to obtain a corresponding output matrix; and performing a weighted average processing on the output matrix to obtain a word vector corresponding to each encoded word.

[0010] Preferably, the second processing is performed on each first processed vector based on the second initial matrix to obtain a corresponding output matrix, comprising: multiplying the second initial matrix with each first processed vector to obtain an intermediate matrix; processing the intermediate matrix based on a preset activation function to obtain an output vector; obtaining a preset loss calculation rule, and processing the output vector based on the preset loss calculation rule to generate a corresponding output matrix, wherein the preset loss calculation rule is associated with the encoded word.

[0011] Preferably, the matching of the vectorized demand with the vectorized first-level matching item and the vectorized second-level matching item to obtain a specific first-level matching item and a specific second-level matching item matched with the business demand information comprises: obtaining a preset matching calculation rule; calculating a matching score between the vectorized demand and each vectorized first-level matching item and each vectorized second-level matching item based on the preset matching calculation rule; determining a maximum score value of the matching score; taking the first-level matching item corresponding to the maximum score value as the specific first-level matching item, and taking the second-level matching item corresponding to the maximum score value as the specific second-level matching item.

[0012] Preferably, the matching method further comprises: determining whether there is unmatched business demand information with a maximum score value less than a preset score threshold; in the case that there is the unmatched business demand information, obtaining an information quantity of the unmatched business demand information; determining whether the information quantity is greater than a preset quantity threshold; in the case that the information quantity is greater than the preset quantity threshold, performing a re-inspection operation on the demand matching result of all business demand information; in the case that the information quantity is less than or equal to the preset quantity threshold, randomly obtaining a preset number of business demand information for re-inspection operation.

[0013] Preferably, the matching method further comprises: after performing the re-inspection operation, obtaining a re-inspection result; in the case that the re-inspection result is a false match, obtaining a false match quantity; in the case that the false match quantity is greater than a preset error quantity value, optimizing the preset business word library based on the re-inspection result to obtain an optimized preset business word library.

[0014] Correspondingly, the application also provides a service demand matching device, which comprises: a first acquisition unit configured to acquire service demand information; a second acquisition unit configured to acquire a preset service vocabulary, wherein the preset service vocabulary comprises a plurality of first-level matching items and a plurality of second-level matching items, each first-level matching item corresponds to at least one second-level matching item; a vector conversion unit configured to perform a vector conversion operation on the service demand information, the first-level matching items and the second-level matching items to obtain corresponding vectorized demand, vectorized first-level matching items and vectorized second-level matching items; a matching unit configured to match the vectorized demand with the vectorized first-level matching items and the vectorized second-level matching items to obtain specific first-level matching items and specific second-level matching items matched with the service demand information; and a matching result determination unit configured to generate a demand matching result based on the specific first-level matching items and the specific second-level matching items.

[0015] Preferably, the second acquisition unit comprises: an initial establishment module configured to establish an initial vocabulary; an encoding module configured to encode each word in the initial vocabulary based on a preset encoding rule to obtain an encoded word; a conversion module configured to perform vector conversion on each encoded word to obtain a corresponding word vector; and a vocabulary establishment module configured to establish a preset service vocabulary based on the word vector.

[0016] Preferably, the initial establishment module is specifically configured to: acquire historical demand information in a preset time period, extract a first noun set and a first verb set of the historical demand information; acquire a second noun set of all first-level matching items and a third noun set of all second-level matching items; acquire a preset verb set; and generate the initial vocabulary based on the first noun set, the second noun set, the third noun set, the first verb set and the preset verb set.

[0017] Preferably, the conversion module is specifically configured to: determine a first initial matrix and a second initial matrix, wherein the number of rows of the first initial matrix is equal to the number of columns of the second initial matrix, and the number of columns of the first initial matrix is equal to the number of rows of the second initial matrix; perform a first processing on each encoded word based on the first initial matrix to obtain a first processed vector corresponding to each encoded word; perform a second processing on each first processed vector based on the second initial matrix to obtain a corresponding output matrix; and perform a weighted average processing on the output matrix to obtain a word vector corresponding to each encoded word.

[0018] Preferably, the second processing is performed on each first processed vector based on the second initial matrix to obtain a corresponding output matrix, comprising: multiplying the second initial matrix with each first processed vector to obtain an intermediate matrix; processing the intermediate matrix based on a preset activation function to obtain an output vector; obtaining a preset loss calculation rule, and processing the output vector based on the preset loss calculation rule to generate a corresponding output matrix, wherein the preset loss calculation rule is associated with the encoded word.

[0019] Preferably, the matching unit comprises: a rule obtaining module, configured to obtain a preset matching calculation rule; a calculation module, configured to calculate a matching score between the vectorized demand and each vectorized first-level matching item and each vectorized second-level matching item based on the preset matching calculation rule; a maximum value determining module, configured to determine a maximum score value of the matching scores; and a matching item determining module, configured to take the first-level matching item corresponding to the maximum score value as the specific first-level matching item, and take the second-level matching item corresponding to the maximum score value as the specific second-level matching item.

[0020] Preferably, the matching unit further comprises a re-inspection module, which is specifically configured to: determine whether there is unmatching business demand information with a maximum score value less than a preset score threshold; in a case where there is the unmatching business demand information, obtain an information quantity of the unmatching business demand information; determine whether the information quantity is greater than a preset quantity threshold; in a case where the information quantity is greater than the preset quantity threshold, perform a re-inspection operation on demand matching results of all business demand information; and in a case where the information quantity is less than or equal to the preset quantity threshold, randomly obtain a preset number of business demand information, and perform a re-inspection operation on the randomly obtained business demand information.

[0021] Preferably, the matching unit further comprises a vocabulary optimization module, which is specifically configured to: after performing the re-inspection operation, obtain a re-inspection result; in a case where the re-inspection result is a false match, obtain a false match quantity; and in a case where the false match quantity is greater than a preset error quantity value, optimize the preset business vocabulary based on the re-inspection result to obtain an optimized preset business vocabulary.

[0022] In another aspect, an embodiment of the present application further provides a processor configured to execute the matching method provided by the embodiment of the present application.

[0023] In another aspect, an embodiment of the present application further provides a machine readable storage medium, which stores instructions, and the instructions, when executed by a processor, cause the processor to be configured to execute the matching method provided by the embodiment of the present application.

[0024] In another aspect, the embodiment of the present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the matching method provided by the embodiment of the present application.

[0025] Through the technical solution provided by the present application, the present application has at least the following technical effects:

[0026] By vectorizing the demand item and the business term in the process of matching the demand item and the business term, and determining the accurate word vector of each term in all application sentences by using the weighted average method, and by using the vector-based matching method in the subsequent matching process, the accurate matching of the demand item and the business term is realized, and compared with the traditional natural language direct matching or fuzzy matching based on the literal semantics, the matching accuracy and the matching efficiency are improved.

[0027] Other features and advantages of the embodiment of the present application will be described in detail in the following specific implementation part. BRIEF DESCRIPTION OF DRAWINGS

[0028] The accompanying drawings are included to provide a further understanding of the embodiment of the present application, and constitute a part of the specification, and are used together with the following specific implementation to explain the embodiment of the present application, but do not constitute a limitation on the embodiment of the present application. In the drawings:

[0029] Figure 1 is a specific implementation flowchart of the matching method of the business demand provided by the embodiment of the present application;

[0030] Figure 2 is a specific implementation flowchart of obtaining a preset business term library in the matching method of the business demand provided by the embodiment of the present application;

[0031] Figure 3 is a specific implementation flowchart of matching to obtain a specific first-level matching item and a specific second-level matching item in the matching method of the business demand provided by the embodiment of the present application;

[0032] Figure 4 is a structural schematic diagram of the matching device of the business demand provided by the embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be clearly and completely described below in combination with the drawings of the embodiments of the present application, and it should be understood that the specific implementation described here is only used to illustrate and explain the embodiments of the present application, and is not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0034] It should be noted that if the application embodiments have directionality indication (such as up, down, left, right, front, back, …), the directionality indication is only used to explain the relative position relationship, motion condition, etc. between components in a certain posture (as shown in the drawings), if the certain posture changes, the directionality indication also changes accordingly.

[0035] In addition, if the application embodiments have descriptions such as "first", "second", etc., the "first", "second", etc. descriptions are only for description purposes, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of indicated technical features. Therefore, the features limited by "first", "second" can explicitly or implicitly include at least one of the features. In addition, the technical solutions of each embodiment can be combined with each other, but it must be based on the realization of ordinary skilled in the art, when the combination of technical solutions appears contradictory or cannot be realized, it should be considered that the combination of technical solutions does not exist, nor within the protection scope required by the application.

[0036] Please refer to Figure 1 The embodiment of the application provides a business demand matching method, the matching method comprises:

[0037] S10) obtaining business demand information;

[0038] S20) obtaining a preset business word library, the preset business word library comprises a plurality of first-level matching items and a plurality of second-level matching items, each first-level matching item corresponds to at least one second-level matching item;

[0039] S30) performing vector conversion operation on the business demand information, the first-level matching item and the second-level matching item to obtain corresponding vectorized demand, vectorized first-level matching item and vectorized second-level matching item;

[0040] S40) matching the vectorized demand with the vectorized first-level matching item and the vectorized second-level matching item to obtain specific first-level matching item and specific second-level matching item matched with the business demand information;

[0041] S50) generating demand matching result based on the specific first-level matching item and the specific second-level matching item.

[0042] In a possible implementation, the business requirement information is first acquired, for example, the business requirement information is an excel document uploaded by a user, the excel document can include multiple requirement items, each requirement item includes a name of the requirement item and a content description, the content description is composed of multiple sentences, please refer to Table 1 for an example of a requirement item provided by an embodiment of the present application:

[0043]

[0044] Table 1: Requirement item example

[0045] After acquiring the business requirement information, a preset business vocabulary is further acquired, the preset business vocabulary includes multiple first-level matching items and multiple second-level matching items, each first-level matching item corresponds to at least one second-level matching item. For example, please refer to Figure 2 In an embodiment of the present application, the acquiring of the preset business vocabulary includes:

[0046] S21) establishing an initial vocabulary;

[0047] S22) encoding each word in the initial vocabulary based on a preset encoding rule to obtain an encoded word;

[0048] S23) performing vector conversion on each encoded word to obtain a corresponding word vector;

[0049] S24) establishing a preset business vocabulary based on the word vector.

[0050] In an embodiment, the preset business vocabulary is created by the technician in advance before the demand matching, and the preset business vocabulary is directly called when the demand matching is performed. In the process of creating the preset business vocabulary, an initial vocabulary is first established. For example, in an embodiment of the present application, the matching method provided by the present application is applied to the demand matching field of bank business, in which field each demand item needs to obtain a corresponding three-level activity and four-level task in the bank, and the bank includes a plurality of three-level activities and four-level tasks, which are in a hierarchical relationship, and each three-level activity is associated with at least one four-level task. In the process of matching the vocabulary of bank business with demand items, the preset business vocabulary of the bank is created in advance, and an initial vocabulary is first established. For example, the initial vocabulary is first generated according to the historical demand information of a preset time period (for example, 1 year), and then all the verbs and nouns used in the historical demand information are extracted to generate a first verb set and a first noun set. Then, the nouns of all three-level activities in the bank are obtained to generate a second noun set, and the nouns of all four-level tasks are obtained to generate a third noun set. At this time, a preset verb set is further obtained, for example, the preset verb set is a verb set generated by the technician in advance according to the actual operation demand or business demand. Then, the first noun set, the second noun set, the third noun set, the first verb set and the preset verb set are integrated to generate the initial vocabulary, for example, the initial vocabulary includes N1 words.

[0051] Then, each word in the initial vocabulary is encoded according to a preset encoding rule, for example, the preset encoding rule is onehot encoding rule. According to the encoding rule, the onehot encoding of the nth word is obtained as [0, 0, 0, …, 1, …, 0, 0, 0]. T Thus, the encoded word corresponding to each word in the initial vocabulary is obtained, and then each encoded word is converted into a vector to obtain a corresponding word vector.

[0052] It is easy for those skilled in the art to know that in the process of matching the vocabulary of bank business with demand items, the essence is to match the natural language. However, because the natural language has various expressions for the same name or the same action, if the bank business vocabulary and the user demand item are simply matched directly, the matching accuracy will be greatly reduced.

[0053] To solve the above technical problems, the encoded words are converted into vectors, and the similarity between different words is determined and accurately matched through the vector. In an embodiment of the present application, the vector conversion is performed on each encoded word to obtain a corresponding word vector, which includes: determining a first initial matrix and a second initial matrix, the number of rows of the first initial matrix is equal to the number of columns of the second initial matrix, and the number of columns of the first initial matrix is equal to the number of rows of the second initial matrix; performing a first processing on each encoded word based on the first initial matrix to obtain a corresponding first processed vector; performing a second processing on each first processed vector based on the second initial matrix to obtain a corresponding output matrix; performing a weighted average processing on the output matrix to obtain a corresponding word vector for each encoded word.

[0054] In a possible implementation, first, a first initial matrix and a second initial matrix are determined, for example, the first initial matrix is a matrix W with a dimension of VxN1, and the second initial matrix is a matrix R with a dimension of N1xV, and the initial values in the above-mentioned matrices W and R are random values. At this time, a first processing operation is performed on each encoded word based on the above-mentioned matrix W to obtain a corresponding first processed vector for each encoded word, for example, in an embodiment, the word b can be processed in combination with the previous and subsequent words (words a and c) of the word b, and a corresponding first processed vector S is obtained, for example, the processing rule is represented as: wherein e a ,e b ,e c respectively represent the onehot encoding of words a, b, and c, and then a second processing is performed on the first processed vector based on the second initial matrix to obtain a corresponding output matrix.

[0055] In an embodiment of the present application, the second processing is performed on each first processed vector based on the second initial matrix to obtain a corresponding output matrix, which includes: multiplying the second initial matrix by each first processed vector to obtain an intermediate matrix; processing the intermediate matrix based on a preset activation function to obtain an output vector; obtaining a preset loss calculation rule, processing the output vector based on the preset loss calculation rule to generate a corresponding output matrix, and the preset loss calculation rule is associated with the encoded word.

[0056] For example, the second initial matrix is first multiplied with the first processed vector S to obtain an intermediate matrix, and then the intermediate matrix is processed by using a sigmoid activation function to obtain a corresponding output vector, which can be used to represent the word vector of the word b. However, at this time, the output vector may have a large deviation from the actual word vector of the word b. Therefore, in order to improve the generation accuracy of the word vector, a preset loss calculation rule is further obtained, and the output vector is optimized according to the preset loss calculation rule to generate a corresponding output matrix. The output matrix can more accurately represent the word vector of the word b. For example, the preset loss calculation rule is represented as: According to the preset loss calculation rule, the accurate output matrix can be derived.

[0057] However, in actual application, the word b may appear in multiple sentences, and the role of the word b in each sentence may be different. Therefore, in order to accurately represent the word vector of the word b as a whole, the output matrix is further subjected to weighted average processing to obtain a corresponding word vector for each encoded word. The word vector can reflect the meaning of the word as a whole. At this time, a preset business word library is established according to the above word vector.

[0058] In the embodiment of the present application, by performing vectorization conversion on the demand items based on natural language and the three-level activities and four-level tasks, and processing each word based on the principle of vector mapping, the word vector of each word is obtained. In the subsequent demand matching process, the vector matching based method can be used instead of the natural language direct matching or the text semantic based matching, thereby effectively improving the matching accuracy, reducing the false matching, and meeting the actual needs of users.

[0059] However, at this time, the business requirement information, the first-level matching item and the second-level matching item processed above still cannot meet the data format of direct vector processing, and therefore further vector mapping conversion needs to be performed on the business requirement information, the first-level matching item and the second-level matching item, for example, in an embodiment, the name of a certain requirement item is composed of m1 words, and the mapping vector H1 of the name of the requirement item is spliced by the word vectors of the m1 words, that is, after vectorization, the requirement can be represented as a matrix with a dimension of m1*N1, and the mapping vector H2 of the content description of the requirement item has a dimension of m2*N1, where m2 is the number of words of the content description of the requirement item. Similarly, according to the name, purpose, definition and range of the third-level activity and the fourth-level task (the name, purpose, definition and range above are predetermined in the bank), it can be determined that the mapping vector A1 of the name of each third-level activity is a matrix with a dimension of k1*N1, the mapping vector A2 of the purpose is a matrix with a dimension of k2*N1, the mapping vector A3 of the definition is a matrix with a dimension of k2*N1, and the mapping vector A4 of the range is a matrix with a dimension of k4*N1, and it can be determined that the mapping vector T1 of the name of each fourth-level task is a matrix with a dimension of l1*N1, the mapping vector T2 of the purpose is a matrix with a dimension of l2*N1, the mapping vector T3 of the definition is a matrix with a dimension of l3*N1, and the mapping vector T4 of the range is a matrix with a dimension of l4*N1, that is, the data of vectorization processing of the business requirement information, the first-level matching item and the second-level matching item is obtained, and at this time, a matching operation is performed.

[0060] Please refer to Figure 3 In the embodiment of the present application, the matching of the vectorized requirement with the vectorized first-level matching item and the vectorized second-level matching item to obtain a specific first-level matching item and a specific second-level matching item matched with the business requirement information comprises:

[0061] S41) obtaining a preset matching calculation rule;

[0062] S42) calculating a matching score between the vectorized requirement and each vectorized first-level matching item and each vectorized second-level matching item based on the preset matching calculation rule;

[0063] S43) determining a maximum score value of the matching score; taking the first-level matching item corresponding to the maximum score value as the specific first-level matching item, and taking the second-level matching item corresponding to the maximum score value as the specific second-level matching item.

[0064] In a possible implementation, first, a preset matching rule is obtained, and a matching score of each requirement item with each first-level matching item and second-level matching item is calculated according to the preset matching rule, and specifically, the preset matching rule can be represented as: Score = cor(H1, A1 T )+cor(H2, A2T )+cor(H2,A3 T )+cor(H2,A4 T )+cor(H1,T1 T )+cor(H2,T2 T )+cor(H2,T3 T )+cor(H2,T4 T ), thus, each demand item calculates multiple matching scores, at this time, the maximum score value in the multiple matching scores is determined, and the first-level matching item corresponding to the maximum score value is taken as a specific first-level matching item, and the second-level matching item corresponding to the maximum score value is taken as a specific second-level matching item, for example, in the demand matching process of bank business, a third-level activity and a fourth-level task matched with the user demand item are obtained.

[0065] In the embodiment of the application, the original demand item and the corresponding multiple matching items are vectorized, so that the demand item and the multiple matching items are matched in a vectorized manner, instead of direct matching based on natural language or matching based on literal semantics, thereby effectively improving the matching accuracy and the matching efficiency, and meeting the actual demand of the user.

[0066] However, in actual application, although the specific first-level matching item and the specific second-level matching item can be obtained through the maximum score value, when the demand item is singular or new, the maximum score value may still be low, that is, the third-level activity or the fourth-level task matched may still have deviation, and cannot meet the actual demand of the user.

[0067] Therefore, in order to further improve the matching accuracy, in the embodiment of the application, the matching method further includes: determining whether there is unmatched business demand information with a maximum score value less than a preset score threshold; in the case that there is the unmatched business demand information, obtaining the information quantity of the unmatched business demand information; determining whether the information quantity is greater than a preset quantity threshold; in the case that the information quantity is greater than the preset quantity threshold, performing a rechecking operation on the demand matching result of all business demand information; in the case that the information quantity is less than or equal to the preset quantity threshold, randomly obtaining a preset number of business demand information for rechecking, and performing a rechecking operation on the randomly obtained business demand information.

[0068] Further, in the embodiment of the application, the matching method further includes: after performing the rechecking operation, obtaining a rechecking result; in the case that the rechecking result is a false match, obtaining a false match quantity; in the case that the false match quantity is greater than a preset error quantity value, optimizing the preset business word library based on the rechecking result to obtain an optimized preset business word library.

[0069] In a possible implementation, in the process of matching a new demand item, it is determined that the maximum score value for the new matching item is less than a preset score threshold, and thus the new matching item is determined as non-matching service demand information, at this time, it is further determined that the information quantity of the non-matching service demand information, for example, in an embodiment, the information quantity of the non-matching service demand information is 3, and the preset quantity threshold is 5, thus it is determined that the above information quantity is less than or equal to the preset quantity threshold, thus it can be determined that the phenomenon of the above maximum score value being too low is caused by matching deviation, thus a preset sampling quantity of service demand information is randomly obtained, and rechecking operation is performed on the randomly obtained service demand information, for example, rechecking operation can be performed in the manner of manual rechecking.

[0070] In another embodiment, after the information quantity of non-matching service demand information in another service demand information is counted, it is determined that the information quantity is greater than a preset quantity threshold, thus it is determined that the generated demand matching result is greatly deviated, and does not meet the actual demand of the user, thus rechecking operation is performed on the demand matching result of all service demand information.

[0071] Further, after rechecking is performed, the corresponding rechecking result is obtained, and the preset service word library is optimized according to the rechecking result, for example, it is displayed according to the rechecking result that the current preset service word library lacks words for matching the new demand item, and does not meet the actual demand of the user, thus the preset service word library is optimized, for example, new words are added into the preset service word library, and the optimized preset service word library is generated, in subsequent use, the demand item of the user can be matched based on the optimized preset service word library, thus the matching capability is realized to be real-time increased, and the actual demand of the user is dynamically met.

[0072] In the embodiment of the application, after the demand item is matched, the matching result is further monitored and managed, and the matching result that does not meet the actual demand of the user is processed in time, thus the matching accuracy is further improved, the influence of mis-matching on the user is reduced, the matching efficiency is improved, and the use experience of the user is improved.

[0073] The service demand matching device provided by the embodiment of the application will be described below with reference to the accompanying drawings.

[0074] Please refer to Figure 4Based on the same inventive concept, the embodiment of the present application provides a service demand matching device, which comprises: a first acquisition unit configured to acquire service demand information; a second acquisition unit configured to acquire a preset service vocabulary library, wherein the preset service vocabulary library comprises a plurality of first-level matching items and a plurality of second-level matching items, each first-level matching item corresponds to at least one second-level matching item; a vector conversion unit configured to perform a vector conversion operation on the service demand information, the first-level matching items and the second-level matching items to obtain corresponding vectorized demand, vectorized first-level matching items and vectorized second-level matching items; a matching unit configured to match the vectorized demand with the vectorized first-level matching items and the vectorized second-level matching items to obtain specific first-level matching items and specific second-level matching items matched with the service demand information; and a matching result determination unit configured to generate a demand matching result based on the specific first-level matching items and the specific second-level matching items.

[0075] In the embodiment of the present application, the second acquisition unit comprises: an initial establishment module configured to establish an initial vocabulary library; an encoding module configured to encode each word in the initial vocabulary library based on a preset encoding rule to obtain an encoded word; a conversion module configured to perform vector conversion on each encoded word to obtain a corresponding word vector; and a vocabulary library establishment module configured to establish a preset service vocabulary library based on the word vector.

[0076] In the embodiment of the present application, the initial establishment module is specifically configured to: acquire historical demand information in a preset time period, extract a first noun set and a first verb set of the historical demand information; acquire a second noun set of all first-level matching items and a third noun set of all second-level matching items; acquire a preset verb set; and generate the initial vocabulary library based on the first noun set, the second noun set, the third noun set, the first verb set and the preset verb set.

[0077] In the embodiment of the present application, the conversion module is specifically configured to: determine a first initial matrix and a second initial matrix, wherein the number of rows of the first initial matrix is equal to the number of columns of the second initial matrix, and the number of columns of the first initial matrix is equal to the number of rows of the second initial matrix; perform a first processing on each encoded word based on the first initial matrix to obtain a first processed vector corresponding to each encoded word; perform a second processing on each first processed vector based on the second initial matrix to obtain a corresponding output matrix; and perform a weighted average processing on the output matrix to obtain a word vector corresponding to each encoded word.

[0078] In the embodiment of the present application, the second processing is performed on each first processed vector based on the second initial matrix to obtain a corresponding output matrix, comprising: multiplying the second initial matrix with each first processed vector to obtain an intermediate matrix; processing the intermediate matrix based on a preset activation function to obtain an output vector; obtaining a preset loss calculation rule, processing the output vector based on the preset loss calculation rule to generate a corresponding output matrix, and the preset loss calculation rule is associated with the encoded word.

[0079] In the embodiment of the present application, the matching unit comprises: a rule obtaining module, configured to obtain a preset matching calculation rule; a calculation module, configured to calculate the matching score between the vectorized demand and each vectorized first-level matching item and each vectorized second-level matching item based on the preset matching calculation rule; a maximum value determination module, configured to determine the maximum score value of the matching score; a matching item determination module, configured to take the first-level matching item corresponding to the maximum score value as the specific first-level matching item, and take the second-level matching item corresponding to the maximum score value as the specific second-level matching item.

[0080] In the embodiment of the present application, the matching unit further comprises a recheck module, which is specifically configured to: judge whether there is unmatched business demand information with a maximum score value less than a preset score threshold; in the case that there is the unmatched business demand information, obtain the information quantity of the unmatched business demand information; judge whether the information quantity is greater than a preset quantity threshold; in the case that the information quantity is greater than the preset quantity threshold, perform a recheck operation on the demand matching result of all business demand information; in the case that the information quantity is less than or equal to the preset quantity threshold, randomly obtain a preset number of business demand information, and perform a recheck operation on the randomly obtained business demand information.

[0081] In the embodiment of the present application, the matching unit further comprises a vocabulary optimization module, which is specifically configured to: after performing the recheck operation, obtain a recheck result; in the case that the recheck result is a false match, obtain a false match quantity; in the case that the false match quantity is greater than a preset error quantity value, optimize the preset business vocabulary based on the recheck result to obtain an optimized preset business vocabulary.

[0082] Further, the embodiment of the present application also provides a processor configured to execute the matching method of the embodiment of the present application.

[0083] Further, the embodiment of the present application also provides a machine readable storage medium, which stores instructions, and the instructions make the processor configured to execute the matching method of the embodiment of the present application when the instructions are executed by the processor.

[0084] Further, an embodiment of the present application also provides a computer program product, comprising a computer program which, when executed by a processor, implements the matching method according to the embodiment of the present application.

[0085] Those skilled in the art will understand that the embodiments of the present application can be provided as methods, systems or computer program products. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage etc.) containing computer usable program code.

[0086] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for performing the function specified by one or more blocks

[0087] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce a product including instruction means, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for performing the function specified by one or more blocks

[0088] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are performed on the computer or other programmable data processing device to produce a computer implemented process, so that the instructions executed on the computer or other programmable device provide the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for performing the function specified by one or more blocks

[0089] In a typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces and memories.

[0090] Memory can include non-persistent memory, Random Access Memory (RAM), and / or non-volatile memory, such as read only memory (ROM) or flash memory, in a computer readable medium. Memory is an example of computer readable media.

[0091] Computer readable media includes permanent and non-permanent, removable and non-removable media, which can be implemented by any method or technology for information storage. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carriers.

[0092] It should also be noted that the terms "comprising", "containing", or any other variant thereof, are intended to encompass non-exclusive inclusion, such that a process, method, article or apparatus that comprises a list of elements does not only include those elements, but can also include other elements not expressly listed or inherent to such process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0093] The above only is an embodiment of the present application, and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A method of matching business requirements, characterized by, The matching method comprises: obtaining service demand information; obtaining a preset service vocabulary library, the preset service vocabulary library comprising a plurality of first-level matching items and a plurality of second-level matching items, each first-level matching item corresponding to at least one second-level matching item; performing vector conversion operations on the service demand information, the first-level matching items and the second-level matching items to obtain corresponding vectorized demand, vectorized first-level matching items and vectorized second-level matching items; matching the vectorized demand with the vectorized first-level matching items and the vectorized second-level matching items to obtain specific first-level matching items and specific second-level matching items that match the service demand information, comprising: obtaining a preset matching calculation rule; calculating matching scores between the vectorized demand and each vectorized first-level matching item and each vectorized second-level matching item based on the preset matching calculation rule; determining a maximum score value of the matching scores; taking the first-level matching item corresponding to the maximum score value as the specific first-level matching item and taking the second-level matching item corresponding to the maximum score value as the specific second-level matching item; and generating a demand matching result based on the specific first-level matching items and the specific second-level matching items.

2. The matching method of claim 1, wherein, The obtaining of the preset service vocabulary library comprises: establishing an initial vocabulary library; encoding each word in the initial vocabulary library based on a preset encoding rule to obtain encoded words; performing vector conversion on each encoded word to obtain a corresponding word vector; and establishing a preset service vocabulary library based on the word vector.

3. The matching method of claim 2, wherein, The establishment of the initial vocabulary library comprises: obtaining historical demand information in a preset time period, extracting a first set of nouns and a first set of verbs of the historical demand information; obtaining a second set of nouns of all first-level matching items and a third set of nouns of all second-level matching items; obtaining a preset set of verbs; generating the initial vocabulary library based on the first set of nouns, the second set of nouns, the third set of nouns, the first set of verbs and the preset set of verbs.

4. The matching method of claim 2, wherein, The vector conversion on each encoded word to obtain a corresponding word vector comprises: determining a first initial matrix and a second initial matrix, the number of rows of the first initial matrix being equal to the number of columns of the second initial matrix, and the number of columns of the first initial matrix being equal to the number of rows of the second initial matrix; performing first processing on each encoded word based on the first initial matrix to obtain a first-processed vector corresponding to each encoded word; performing second processing on each first-processed vector based on the second initial matrix to obtain a corresponding output matrix; and performing weighted average processing on the output matrix to obtain a word vector corresponding to each encoded word.

5. The matching method of claim 4, wherein, The second processing on each first-processed vector based on the second initial matrix to obtain a corresponding output matrix comprises: multiplying the second initial matrix by each first-processed vector to obtain an intermediate matrix; processing the intermediate matrix based on a preset activation function to obtain an output vector; and Obtain a preset loss calculation rule, process the output vector based on the preset loss calculation rule, and generate a corresponding output matrix, wherein the preset loss calculation rule is associated with the encoded words.

6. The matching method of claim 1, wherein, The matching method further includes: determining whether there is unmatched business requirement information with a maximum score value less than a preset score threshold; in the case where there is the unmatched business requirement information, obtaining an information quantity of the unmatched business requirement information; determining whether the information quantity is greater than a preset quantity threshold; in the case where the information quantity is greater than the preset quantity threshold, performing a re-inspection operation on the demand matching result of all business requirement information; in the case where the information quantity is less than or equal to the preset quantity threshold, randomly obtaining a preset number of business requirement information for re-inspection, and performing a re-inspection operation on the randomly obtained business requirement information.

7. The matching method of claim 6, wherein, The matching method further includes: after performing the re-inspection operation, obtaining a re-inspection result; in the case where the re-inspection result is a mis-match, obtaining a mis-match quantity; in the case where the mis-match quantity is greater than a preset error quantity value, optimizing the preset business word library based on the re-inspection result to obtain an optimized preset business word library.

8. A service requirement matching apparatus characterized by comprising: The matching device includes: a first obtaining unit configured to obtain business requirement information; a second obtaining unit configured to obtain a preset business word library, wherein the preset business word library includes a plurality of first-level matching items and a plurality of second-level matching items, and each first-level matching item corresponds to at least one second-level matching item; a vector conversion unit configured to perform a vector conversion operation on the business requirement information, the first-level matching items, and the second-level matching items to obtain corresponding vectorized requirements, vectorized first-level matching items, and vectorized second-level matching items; a matching unit configured to match the vectorized requirements with the vectorized first-level matching items and the vectorized second-level matching items to obtain specific first-level matching items and specific second-level matching items that match the business requirement information; a matching result determination unit configured to generate a demand matching result based on the specific first-level matching items and the specific second-level matching items. The matching unit includes: a rule obtaining module configured to obtain a preset matching calculation rule; a calculation module configured to calculate a matching score between the vectorized requirements and each vectorized first-level matching item and each vectorized second-level matching item based on the preset matching calculation rule; a maximum value determination module configured to determine a maximum score value of the matching scores; a matching item determination module configured to take the first-level matching item corresponding to the maximum score value as the specific first-level matching item, and take the second-level matching item corresponding to the maximum score value as the specific second-level matching item.

9. The matching device of claim 8, wherein, The second obtaining unit includes: an initial establishment module configured to establish an initial word library; an encoding module configured to encode each word in the initial word library based on a preset encoding rule to obtain encoded words; a conversion module configured to perform vector conversion on each encoded word to obtain a corresponding word vector; a word library establishment module configured to establish a preset business word library based on the word vector.

10. The matching device of claim 9, wherein, The initial establishment module is specifically used for: obtaining historical demand information in a preset time period, extracting a first set of nouns and a first set of verbs of the historical demand information; obtaining a second set of nouns of all first-level matching items, and obtaining a third set of nouns of all second-level matching items; obtaining a preset set of verbs; generating the initial vocabulary based on the first set of nouns, the second set of nouns, the third set of nouns, the first set of verbs, and the preset set of verbs.

11. The matching device of claim 9, wherein, The conversion module is specifically used for: determining a first initial matrix and a second initial matrix, a number of rows of the first initial matrix being equal to a number of columns of the second initial matrix, and a number of columns of the first initial matrix being equal to a number of rows of the second initial matrix; performing first processing on each encoded word based on the first initial matrix to obtain a first processed vector corresponding to each encoded word; performing second processing on each first processed vector based on the second initial matrix to obtain a corresponding output matrix; performing weighted average processing on the output matrix to obtain a word vector corresponding to each encoded word.

12. The matching device of claim 11, wherein, The performing second processing on each first processed vector based on the second initial matrix to obtain a corresponding output matrix comprises: multiplying the second initial matrix and each first processed vector to obtain an intermediate matrix; processing the intermediate matrix based on a preset activation function to obtain an output vector; obtaining a preset loss calculation rule, processing the output vector based on the preset loss calculation rule to generate a corresponding output matrix, and the preset loss calculation rule being associated with the encoded word.

13. The matching apparatus according to claim 8, characterized by The matching unit further comprises a re-inspection module, and the re-inspection module is specifically used for: determining whether there is unmatching business demand information with a maximum score value less than a preset score threshold; in a case where there is the unmatching business demand information, obtaining an information quantity of the unmatching business demand information; determining whether the information quantity is greater than a preset quantity threshold; in a case where the information quantity is greater than the preset quantity threshold, performing a re-inspection operation on a demand matching result of all business demand information; in a case where the information quantity is less than or equal to the preset quantity threshold, randomly obtaining a preset number of business demand information for performing a re-inspection operation on the randomly obtained business demand information.

14. The matching device of claim 13, wherein, The matching unit further comprises a vocabulary optimization module, and the vocabulary optimization module is specifically used for: after performing the re-inspection operation, obtaining a re-inspection result; in a case where the re-inspection result is a false match, obtaining a false match quantity; in a case where the false match quantity is greater than a preset error quantity value, optimizing the preset business vocabulary based on the re-inspection result to obtain an optimized preset business vocabulary.

15. A processor, comprising: The computer program is configured to perform the matching method of any one of claims 1-7.

16. A machine-readable storage medium having stored thereon instructions, the instructions comprising: The computer program is configured to perform the matching method of any one of claims 1-7.

17. A computer program product comprising a computer program, characterized in that, The computer program is configured to perform the matching method of any one of claims 1-7.

Images