Information consulting service system and method based on big data

Abstract

The present invention discloses an information consulting service system and method based on big data. First, the government affairs consulting question is processed by word segmentation and then passed through a word embedding layer to obtain a sequence of government affairs consulting word embedding vectors. Then, the sequence of government affairs consulting word embedding vectors is passed through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a government affairs consulting question semantic understanding feature vector. Then, the first alternative answer is processed by word segmentation and then passed through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain a first alternative answer semantic understanding feature vector. Finally, the first alternative answer semantic understanding feature vector and the government affairs consulting question semantic understanding feature vector are associated encoded and passed through a classifier to obtain a classification result indicating whether to return the first alternative answer. In this way, efficient and accurate government affairs consulting services can be achieved.

Description

Technical Field

[0001] The present application relates to the field of intelligent information consulting, and more specifically, to an information consulting service system and method based on big data. Background Art

[0002] With the rapid development of big data and artificial intelligence technology, the construction of government informationization has been continuously strengthened, and government consulting services have become an increasingly important part of government services. At present, with the growing demand for government consulting services, how to improve the efficiency and quality of government consulting services has become an urgent problem to be solved.

[0003] Traditional government consultation services often rely on manual consultation and answering, which has problems such as limited human resources and low efficiency. At the same time, the accuracy of consultation and answering is not well guaranteed. In addition, the existing rule-based and template-based question-and-answer systems often require manual writing of rules and templates, which cannot cope with the diversification and complexity of questions. As a result, traditional government consultation services can no longer meet the needs of citizens, enterprises and government departments.

[0004] Therefore, an optimized information consulting service system based on big data is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an information consulting service system and method based on big data. It first performs word segmentation processing on the government consultation question and then passes it through the word embedding layer to obtain a sequence of government consultation word embedding vectors. Then, the sequence of government consultation word embedding vectors is passed through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of the government consultation question. Then, the first alternative answer is processed through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain the first alternative answer semantic understanding feature vector. Finally, the first alternative answer semantic understanding feature vector and the government consultation question semantic understanding feature vector are associated encoded and then passed through a classifier to obtain a classification result indicating whether to return the first alternative answer. In this way, efficient and accurate government consultation services can be achieved.

[0006] According to one aspect of the present application, there is provided an information consulting service system based on big data, which includes: a consulting question collection module for acquiring government consulting questions; a word segmentation module for performing word segmentation processing on the government consulting questions to obtain a sequence of government consulting words; a word embedding module for passing the sequence of government consulting words through a word embedding layer to obtain a sequence of government consulting word embedding vectors; a consulting question semantic understanding module for passing the sequence of government consulting word embedding vectors through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a government consulting question semantic understanding feature vector; an alternative answer acquisition module for obtaining a first alternative answer; an alternative answer semantic understanding module for performing word segmentation processing on the first alternative answer and then passing it through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain a first alternative answer semantic understanding feature vector; a semantic association module for performing association encoding on the first alternative answer semantic understanding feature vector and the government consulting question semantic understanding feature vector to obtain a semantic matching feature matrix; and an alternative answer return detection module for passing the semantic matching feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether to return the first alternative answer.

[0007] In the above-mentioned information consulting service system based on big data, the consulting question semantic understanding module is used to: input the sequence of the government affairs consulting word embedding vectors into the semantic encoder to obtain multiple government affairs consulting question semantic feature vectors; and cascade the multiple government affairs consulting question semantic feature vectors to obtain the government affairs consulting question semantic understanding feature vector.

[0008] In the above-mentioned information consulting service system based on big data, the alternative answer semantic understanding module is used to: perform word segmentation processing on the first alternative answer to obtain a sequence of first alternative answer words; pass the sequence of the first alternative answer words through the word embedding layer to obtain a sequence of first alternative answer word embedding vectors; input the sequence of the first alternative answer word embedding vectors into the semantic encoder to obtain multiple first alternative answer semantic feature vectors; and cascade the multiple first alternative answer semantic feature vectors to obtain the first alternative answer semantic understanding feature vector.

[0009] In the above-mentioned information consulting service system based on big data, the semantic association module is used to: associate and encode the semantic understanding feature vector of the first candidate answer and the semantic understanding feature vector of the government affairs consulting question with the following association coding formula to obtain the semantic matching feature matrix; wherein the association coding formula is: ,in, represents the semantic understanding feature vector of the first candidate answer, represents the transposed vector of the semantic understanding feature vector of the first candidate answer, represents the semantic understanding feature vector of the government affairs consultation question, represents the semantic matching feature matrix, Represents vector multiplication.

[0010] The above-mentioned information consulting service system based on big data also includes a training module for training the semantic encoder based on the Transformer module and the BiLSTM model and the classifier.

[0011] In the above-mentioned information consulting service system based on big data, the training module includes: a training data acquisition unit, which is used to obtain training data, and the training data includes training government affairs consulting questions, training first alternative answers, and whether to return the true value of the first alternative answer; a training word segmentation unit, which is used to perform word segmentation processing on the training government affairs consulting questions to obtain a sequence of training government affairs consulting words; a training word embedding unit, which is used to pass the sequence of training government affairs consulting words through the word embedding layer to obtain a sequence of training government affairs consulting word embedding vectors; a training consulting question semantic understanding unit, which is used to pass the sequence of training government affairs consulting word embedding vectors through the semantic encoder based on the Transformer module and the BiLSTM model to obtain a training government affairs consulting question semantic understanding feature vector; a training alternative answer semantic understanding unit, which is used to perform word segmentation processing on the training first alternative answer The word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model are used to obtain the semantic understanding feature vector of the training first alternative answer; a training semantic association unit is used to associate and encode the semantic understanding feature vector of the training first alternative answer and the semantic understanding feature vector of the training government affairs consultation question to obtain a training semantic matching feature matrix; an iterative optimization unit is used to iteratively optimize the training semantic matching feature vector after the training semantic matching feature matrix is ​​expanded to obtain an optimized training semantic matching feature vector; a classification loss unit is used to pass the optimized training semantic matching feature vector through the classifier to obtain a classification loss function value; and a training unit is used to train the semantic encoder based on the Transformer module and the BiLSTM model and the classifier through back propagation of gradient descent with the classification loss function value.

[0012] In the above-mentioned information consulting service system based on big data, the iterative optimization unit is used to iteratively optimize the training semantic matching feature vector after the training semantic matching feature matrix is ​​expanded using the following iterative optimization formula to obtain the optimized training semantic matching feature vector; wherein the iterative optimization formula is: ,in, is the first eigenvalues, represents the probabilistic function of the eigenvalue, is the class probability value obtained by the classifier, is the length of the semantic matching feature vector, and is the weight hyperparameter, is the first feature vector of the optimized training semantic matching feature values.

[0013] In the above-mentioned information consulting service system based on big data, the classification loss unit is used to: use the classifier to process the optimized training semantic matching feature vector with the following classification loss formula to obtain a training classification result, and the classification loss formula is: ,in The optimization training semantic matching feature vector, to is the weight matrix of each fully connected layer, to Representing the bias matrix of each fully connected layer; and calculating the cross entropy value between the training classification result and the true value as the classification loss function value.

[0014] In the above-mentioned information consulting service system based on big data, according to another aspect of the present application, a method for information consulting service based on big data is provided, which includes: obtaining government affairs consulting questions; performing word segmentation processing on the government affairs consulting questions to obtain a sequence of government affairs consulting words; passing the sequence of government affairs consulting words through a word embedding layer to obtain a sequence of government affairs consulting word embedding vectors; passing the sequence of government affairs consulting word embedding vectors through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of government affairs consulting questions; obtaining a first alternative answer; performing word segmentation processing on the first alternative answer and passing it through the word embedding layer and the semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of the first alternative answer; performing associative coding on the semantic understanding feature vector of the first alternative answer and the semantic understanding feature vector of the government affairs consulting question to obtain a semantic matching feature matrix; performing soft matching on the semantic matching feature matrix with a rank-ordered distribution using a feature scale as an imitation mask to obtain a semantic matching feature matrix; and passing the semantic matching feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether to return the first alternative answer.

[0015] In the above-mentioned information consulting service method based on big data, the sequence of the government affairs consulting word embedding vectors is passed through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of the government affairs consulting question, including: inputting the sequence of the government affairs consulting word embedding vectors into the semantic encoder to obtain multiple government affairs consulting question semantic feature vectors; and cascading the multiple government affairs consulting question semantic feature vectors to obtain the government affairs consulting question semantic understanding feature vector.

[0016] Compared with the prior art, the information consulting service system and method based on big data provided by the present application first performs word segmentation processing on the government affairs consulting questions and then passes through the word embedding layer to obtain a sequence of government affairs consulting word embedding vectors. Next, the sequence of government affairs consulting word embedding vectors is passed through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of the government affairs consulting question. Then, the first alternative answer is subjected to word segmentation processing and then passed through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain a semantic understanding feature vector of the first alternative answer. Finally, the semantic understanding feature vector of the first alternative answer and the semantic understanding feature vector of the government affairs consulting question are associatedly encoded and then passed through a classifier to obtain a classification result indicating whether to return the first alternative answer. In this way, efficient and accurate government affairs consulting services can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. The following drawings are not deliberately scaled according to the actual size, and the focus is on illustrating the main purpose of the present application.

[0018] Figure 1 This is an application scenario diagram of the big data-based information consulting service system according to an embodiment of the present application.

[0019] Figure 2 It is a block diagram of an information consulting service system based on big data according to an embodiment of the present application.

[0020] Figure 3 It is a block diagram of a training module further included in the big data-based information consulting service system according to an embodiment of the present application.

[0021] Figure 4 The present invention is a flowchart of an information consulting service method based on big data according to an embodiment of the present application.

[0022] Figure 5 It is a schematic diagram of the system architecture of the information consulting service method based on big data according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0024] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0025] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0026] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0027] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0028] As mentioned above, traditional government consultation services often rely on manual consultation and answering, which has problems such as limited human resources and low efficiency. At the same time, the accuracy of consultation service questions and answers cannot be well guaranteed. In addition, the existing rule-based and template-based question-answering systems often require manual writing of rules and templates, which cannot cope with the diversification and complexity of questions, resulting in the inability of traditional government consultation services to meet the needs of citizens, enterprises and government departments. Therefore, an optimized information consultation service system based on big data is expected.

[0029] Accordingly, considering that in the actual process of providing government consultation services, in order to improve the efficiency and accuracy of the consultation services, it is expected to use big data and artificial intelligence technologies to realize intelligent consultation question and answer services for government information through semantic feature matching analysis of government consultation questions and alternative answers. In other words, the core of the government information intelligent question and answer system is semantic feature matching, and the key lies in the accurate and full semantic understanding of government consultation questions and each alternative answer, so as to perform semantic association matching between the two. Therefore, in this process, the difficulty and key lies in how to fully express the contextual semantic association feature information between the semantic understanding features of the government consultation questions and the semantic understanding features of the alternative answers, so as to accurately perform intelligent matching of government consultation questions and answers, thereby realizing efficient and accurate government consultation services.

[0030] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. The development of deep learning and neural networks provides new solutions and solutions for mining the contextual semantic association feature information between the semantic understanding features of the government consultation questions and the semantic understanding features of the alternative answers.

[0031] Figure 1 FIG. 1 is an application scenario diagram of the information consulting service system based on big data according to an embodiment of the present application. Figure 1 As shown, in this application scenario, first, obtain the government consultation questions (for example, Figure 1 D1 as indicated in the figure) and the first alternative answer (e.g. Figure 1 Then, the government affairs consultation question and the first candidate answer are input into a server that is deployed with an information consultation service algorithm based on big data (for example, Figure 1 S as shown in ), wherein the server is capable of using the big data-based information consulting service algorithm to process the government affairs consulting question and the first alternative answer to obtain a classification result indicating whether to return the first alternative answer.

[0032] Figure 2 FIG. 1 is a block diagram of an information consulting service system based on big data according to an embodiment of the present application. Figure 2As shown, according to the embodiment of the present application, the information consulting service system 100 based on big data includes: a consulting question acquisition module 110, which is used to obtain government consulting questions; a word segmentation module 120, which is used to perform word segmentation processing on the government consulting questions to obtain a sequence of government consulting words; a word embedding module 130, which is used to pass the sequence of government consulting words through a word embedding layer to obtain a sequence of government consulting word embedding vectors; a consulting question semantic understanding module 140, which is used to pass the sequence of government consulting word embedding vectors through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of government consulting questions; an alternative answer acquisition module 150, which is used to obtain a semantic understanding feature vector of government consulting questions; for obtaining the first alternative answer; an alternative answer semantic understanding module 160, for performing word segmentation on the first alternative answer and then passing it through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain a first alternative answer semantic understanding feature vector; a semantic association module 170, for performing association encoding on the first alternative answer semantic understanding feature vector and the government affairs consultation question semantic understanding feature vector to obtain a semantic matching feature matrix; and an alternative answer return detection module 180, for passing the semantic matching feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether to return the first alternative answer.

[0033] More specifically, in the embodiment of the present application, the consulting question collection module 110 is used to obtain government consulting questions; the alternative answer acquisition module 150 is used to obtain the first alternative answer. The core of the government information intelligent question-answering system is semantic feature matching, and the key lies in accurately and fully understanding the semantics of government consulting questions and each alternative answer, so as to perform semantic association matching between the two.

[0034] More specifically, in an embodiment of the present application, the word segmentation module 120 is used to perform word segmentation processing on the government affairs consultation question to obtain a sequence of government affairs consultation words. Since the government affairs consultation question contains a large amount of semantic information, and the government affairs consultation question is composed of multiple words, these words are of great significance for the semantic understanding of the government affairs consultation question. Therefore, in order to be able to fully and accurately extract the contextual semantic association features between the words in the government affairs consultation question, in the technical solution of the present application, the government affairs consultation question is firstly subjected to word segmentation processing to avoid confusion in the word order during the subsequent semantic understanding, which may cause errors in the semantic understanding of the government affairs consultation question, thereby obtaining a sequence of government affairs consultation words.

[0035] More specifically, in an embodiment of the present application, the word embedding module 130 is used to pass the sequence of the government affairs consultation words through a word embedding layer to obtain a sequence of government affairs consultation word embedding vectors. Since the government affairs consultation questions are all government affairs professional terms, in order to improve the accuracy of the subsequent semantic understanding of the government affairs consultation questions, in the technical solution of the present application, the sequence of the government affairs consultation words is further passed through a word embedding layer, so as to use the word embedding layer to map the sequence of the government affairs consultation words to an embedding vector to obtain a sequence of government affairs consultation word embedding vectors. In particular, here, the word embedding layer can be constructed using a knowledge graph of semantic features of government affairs-related professional terms so that prior information on the semantic features of government affairs professional terms is introduced in the process of converting the sequence of government affairs consultation words into an embedding vector.

[0036] More specifically, in an embodiment of the present application, the consulting question semantic understanding module 140 is used to pass the sequence of the government consulting word embedding vectors through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of the government consulting question. Since the words in the government consulting question have context-based semantic association feature information, in order to fully capture the semantic association features between the words in the government consulting question, so as to accurately perform semantic understanding of the government consulting question and perform subsequent semantic feature matching, in the technical solution of the present application, the sequence of the government consulting word embedding vectors is further passed through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of the government consulting question.

[0037] Accordingly, in a specific example, the consulting question semantic understanding module 140 is used to: input the sequence of the government consultation word embedding vectors into the semantic encoder to obtain multiple government consultation question semantic feature vectors; and, cascade the multiple government consultation question semantic feature vectors to obtain the government consultation question semantic understanding feature vector.

[0038] More specifically, in the embodiment of the present application, the alternative answer semantic understanding module 160 is used to perform word segmentation on the first alternative answer and then pass it through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain the first alternative answer semantic understanding feature vector. In this way, the contextual semantic association fusion feature information based on long-distance dependency association and medium-short-distance dependency association in the first alternative answer is extracted, thereby obtaining the first alternative answer semantic understanding feature vector.

[0039] Accordingly, in a specific example, the alternative answer semantic understanding module 160 is used to: perform word segmentation processing on the first alternative answer to obtain a sequence of first alternative answer words; pass the sequence of the first alternative answer words through the word embedding layer to obtain a sequence of first alternative answer word embedding vectors; input the sequence of the first alternative answer word embedding vectors into the semantic encoder to obtain multiple first alternative answer semantic feature vectors; and cascade the multiple first alternative answer semantic feature vectors to obtain the first alternative answer semantic understanding feature vector.

[0040] More specifically, in the embodiment of the present application, the semantic association module 170 is used to associate and encode the semantic understanding feature vector of the first candidate answer and the semantic understanding feature vector of the government consultation question to obtain a semantic matching feature matrix. This is used to represent the contextual semantic association feature information between the semantic understanding feature of the government consultation question and the semantic understanding feature of the first candidate answer, that is, the semantic feature matching association information between the government consultation question and the first candidate answer, thereby obtaining a semantic matching feature matrix.

[0041] Accordingly, in a specific example, the semantic association module 170 is used to: perform association coding on the first candidate answer semantic understanding feature vector and the government affairs consultation question semantic understanding feature vector using the following association coding formula to obtain the semantic matching feature matrix; wherein the association coding formula is: ,in, represents the semantic understanding feature vector of the first candidate answer, represents the transposed vector of the semantic understanding feature vector of the first candidate answer, represents the semantic understanding feature vector of the government affairs consultation question, represents the semantic matching feature matrix, Represents vector multiplication.

[0042] More specifically, in the embodiment of the present application, the alternative answer return detection module 180 is used to pass the semantic matching feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the first alternative answer is returned. In other words, the classification is performed based on the semantic feature matching degree between the government affairs consultation question and the first alternative answer, so as to determine whether the first alternative answer is returned.

[0043] It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression, SVM, etc. are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification, but this is prone to errors and inefficient. Commonly used multi-classification methods include the Softmax classification function.

[0044] Accordingly, in a specific example, the information consulting service system based on big data further includes a training module for training the semantic encoder based on the Transformer module and the BiLSTM model and the classifier. Figure 3 As shown, the training module 200 includes: a training data acquisition unit 210, which is used to obtain training data, and the training data includes training government consultation questions, training first alternative answers, and whether to return the true value of the first alternative answer; a training word segmentation unit 220, which is used to perform word segmentation processing on the training government consultation questions to obtain a sequence of training government consultation words; a training word embedding unit 230, which is used to pass the sequence of training government consultation words through the word embedding layer to obtain a sequence of training government consultation word embedding vectors; a training consultation question semantic understanding unit 240, which is used to pass the sequence of training government consultation word embedding vectors through the semantic encoder based on the Transformer module and the BiLSTM model to obtain a training government consultation question semantic understanding feature vector; a training alternative answer semantic understanding unit 250, which is used to perform word segmentation processing on the training first alternative answer and pass it through the word embedding layer The input layer and the semantic encoder based on the Transformer module and the BiLSTM model are used to obtain the training first alternative answer semantic understanding feature vector; the training semantic association unit 260 is used to associate the training first alternative answer semantic understanding feature vector and the training government affairs consultation question semantic understanding feature vector to obtain a training semantic matching feature matrix; the iterative optimization unit 270 is used to iteratively optimize the training semantic matching feature vector after the training semantic matching feature matrix is ​​expanded to obtain an optimized training semantic matching feature vector; the classification loss unit 280 is used to pass the optimized training semantic matching feature vector through the classifier to obtain a classification loss function value; and the training unit 290 is used to train the semantic encoder based on the Transformer module and the BiLSTM model and the classifier through back propagation of gradient descent with the classification loss function value.

[0045] In the technical solution of the present application, considering that the semantic understanding feature vector of the training first alternative answer and the semantic understanding feature vector of the training government affairs consultation question respectively express the contextual association coding features of the word-level text semantics of the training first alternative answer and the training government affairs consultation word and the bidirectional process-remote association coding features of the semantic feature granularity, the difference in the text semantics of their source data will be amplified by the text semantic encoding of the semantic encoder in the expression of association features at different scales, resulting in a relatively significant distribution imbalance between the overall feature distributions of the semantic understanding feature vector of the training first alternative answer and the semantic understanding feature vector of the training government affairs consultation question, which will affect the association coding effect of position-by-position association coding of the semantic understanding feature vector of the training first alternative answer and the semantic understanding feature vector of the training government affairs consultation question, thereby resulting in the difference in the eigenvalue feature distribution of the training semantic matching feature matrix due to the spatial heterogeneous distribution of high-dimensional features, so that when the training semantic matching feature matrix is ​​classified by a classifier, the inconsistency of the local feature distribution based on the eigenvalue numerical distribution affects the class regression iteration convergence speed.

[0046] Therefore, the present application iteratively optimizes the training semantic matching feature vector each time the training semantic matching feature vector after the training semantic matching feature matrix is ​​expanded passes through the classifier for iterative classification regression.

[0047] Accordingly, in a specific example, the iterative optimization unit 270 is used to iteratively optimize the training semantic matching feature vector after the training semantic matching feature matrix is ​​expanded using the following iterative optimization formula to obtain the optimized training semantic matching feature vector; wherein the iterative optimization formula is: ,in, is the first eigenvalues, Represents the probabilistic function of the eigenvalue, that is, the eigenvalue Map to The probability function of the interval, is the class probability value obtained by the classifier, is the length of the semantic matching feature vector, and is the weight hyperparameter, is the first feature vector of the optimized training semantic matching feature values.

[0048] Here, the training semantic matching feature vector is performed by taking the Bernoulli probability likelihood distribution as the authoritative meta-entity Based on Bernoulli probability distribution The high-level class decision is made by introducing a probability density space mapping allocation relative to the entire feature set, thereby determining the eigenvalues ​​of the training semantic matching feature vector The key insight with respect to the class probability mapping is to implement the feature values ​​of the trained semantic matching feature vector Based on class probability distribution The correctness of the matching function with class convergence can promote sample iterative convergence through convergence performance superiority and improve classification training speed. In this way, it is possible to accurately match government consultation questions and answers, thereby achieving efficient and accurate government consultation services.

[0049] Accordingly, in a specific example, the classification loss unit 280 is used to: use the classifier to process the optimized training semantic matching feature vector using the following classification loss formula to obtain a training classification result, and the classification loss formula is: ,in represents the optimized training semantic matching feature vector, to is the weight matrix of each fully connected layer, to Representing the bias matrix of each fully connected layer; and calculating the cross entropy value between the training classification result and the true value as the classification loss function value.

[0050] In summary, the big data-based information consulting service system 100 based on the embodiment of the present application is explained, which first performs word segmentation processing on the government affairs consulting question and then passes it through the word embedding layer to obtain a sequence of government affairs consulting word embedding vectors, then passes the sequence of government affairs consulting word embedding vectors through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of the government affairs consulting question, then, performs word segmentation processing on the first alternative answer and passes it through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain the first alternative answer semantic understanding feature vector, finally, the first alternative answer semantic understanding feature vector and the government affairs consulting question semantic understanding feature vector are associated encoded and then passed through a classifier to obtain a classification result indicating whether to return the first alternative answer. In this way, efficient and accurate government affairs consulting services can be achieved.

[0051] As described above, according to the embodiment of the present application, the information consulting service system 100 based on big data based on the embodiment of the present application can be implemented in various terminal devices, such as a server having an information consulting service algorithm based on big data based on the embodiment of the present application. In one example, the information consulting service system 100 based on big data based on the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the information consulting service system 100 based on big data based on the embodiment of the present application can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the information consulting service system 100 based on big data based on the embodiment of the present application can also be one of the many hardware modules of the terminal device.

[0052] Alternatively, in another example, the big data-based information consulting service system 100 based on the embodiment of the present application and the terminal device may also be separate devices, and the big data-based information consulting service system 100 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0053] Figure 4 The present invention is a flowchart of an information consulting service method based on big data according to an embodiment of the present application. Figure 5 Schematic diagram of the system architecture of the information consulting service method based on big data according to the embodiment of the present application. Figure 4 and Figure 5 As shown, according to the information consulting service method based on big data of the embodiment of the present application, it includes: S110, obtaining government consultation questions; S120, performing word segmentation processing on the government consultation questions to obtain a sequence of government consultation words; S130, passing the sequence of government consultation words through a word embedding layer to obtain a sequence of government consultation word embedding vectors; S140, passing the sequence of government consultation word embedding vectors through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of government consultation questions; S150, obtaining a first alternative answer; S160, performing word segmentation processing on the first alternative answer and passing it through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain a semantic understanding feature vector of the first alternative answer; S170, performing association coding on the semantic understanding feature vector of the first alternative answer and the semantic understanding feature vector of the government consultation question to obtain a semantic matching feature matrix; and, S180, passing the semantic matching feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether to return the first alternative answer.

[0054] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned information consulting service method based on big data have been referred to above. Figures 1 to 3 The description of the big data-based information consulting service system 100 has been introduced in detail, and therefore, its repeated description will be omitted.

[0055] According to another aspect of the present application, a non-volatile computer-readable storage medium is provided, on which computer-readable instructions are stored. When the instructions are executed by a computer, the above-mentioned method can be executed.

[0056] The program part of the technology can be considered as a "product" or "manufactured product" in the form of executable code and / or related data, which is participated in or realized by computer-readable media. Tangible and permanent storage media can include any memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives or any similar devices that can provide storage functions for software.

[0057] The present application uses specific words to describe the embodiments of the present application. For example, "first / second embodiment", "one embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or multiple times in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0058] In addition, it will be appreciated by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0059] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

[0060] The above is an explanation of the present invention and should not be considered as a limitation thereof. Although several exemplary embodiments of the present invention have been described, it will be readily appreciated by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined in the claims. It should be understood that the above is an explanation of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. An information consulting service system based on big data, characterized in that: include: Consultation question collection module, used to obtain government consultation questions; A word segmentation module, used for performing word segmentation processing on the government affairs consultation question to obtain a sequence of government affairs consultation words; A word embedding module, used for passing the sequence of government affairs consultation words through a word embedding layer to obtain a sequence of government affairs consultation word embedding vectors; A consultation question semantic understanding module, used to pass the sequence of the government consultation word embedding vectors through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of the government consultation question; The alternative answer acquisition module is used to obtain the first alternative answer; An alternative answer semantic understanding module, used for performing word segmentation processing on the first alternative answer and then passing it through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain a semantic understanding feature vector of the first alternative answer; A semantic association module, used for associating and encoding the semantic understanding feature vector of the first candidate answer and the semantic understanding feature vector of the government affairs consultation question to obtain a semantic matching feature matrix; as well as An alternative answer return detection module is used to pass the semantic matching feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether to return the first alternative answer; The system also includes a training module for training the semantic encoder based on the Transformer module and the BiLSTM model and the classifier; The training module comprises: A training data acquisition unit, used to obtain training data, wherein the training data includes training government affairs consultation questions, training first candidate answers, and whether to return a true value of the first candidate answer; A training word segmentation unit, used for performing word segmentation processing on the training government affairs consultation question to obtain a sequence of training government affairs consultation words; A training word embedding unit, used for passing the sequence of training government affairs consultation words through the word embedding layer to obtain a sequence of training government affairs consultation word embedding vectors; A training consultation question semantic understanding unit, used for passing the sequence of the training government affairs consultation word embedding vectors through the semantic encoder based on the Transformer module and the BiLSTM model to obtain a training government affairs consultation question semantic understanding feature vector; A training candidate answer semantic understanding unit, used for performing word segmentation processing on the training first candidate answer and then passing it through the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model to obtain a semantic understanding feature vector of the training first candidate answer; A training semantic association unit, used for performing association coding on the training first candidate answer semantic understanding feature vector and the training government affairs consultation question semantic understanding feature vector to obtain a training semantic matching feature matrix; An iterative optimization unit, used for iteratively optimizing the training semantic matching feature vector after the training semantic matching feature matrix is ​​expanded to obtain an optimized training semantic matching feature vector; A classification loss unit, used for passing the optimized training semantic matching feature vector through the classifier to obtain a classification loss function value; and A training unit, used to train the semantic encoder based on the Transformer module and the BiLSTM model and the classifier by back propagation of gradient descent with the classification loss function value; The iterative optimization unit is used to: The training semantic matching feature vector after the training semantic matching feature matrix is ​​expanded is iteratively optimized using the following iterative optimization formula to obtain the optimized training semantic matching feature vector; wherein the iterative optimization formula is: Among them, v i is the i-th eigenvalue of the training semantic matching feature vector, φ(v i ) represents the probabilistic function of the eigenvalue, that is, the eigenvalue v i The probability function mapped to the interval [0,1], p is the class probability value obtained by the classifier, L is the length of the semantic matching feature vector, and α is the weight hyperparameter, v' i is the i-th eigenvalue of the optimized training semantic matching feature vector.

2. The information consulting service system based on big data according to claim 1 is characterized in that: The consulting question semantic understanding module is used to: Inputting the sequence of the government affairs consultation word embedding vectors into the semantic encoder to obtain a plurality of government affairs consultation question semantic feature vectors; as well as The multiple government affairs consultation question semantic feature vectors are cascaded to obtain the government affairs consultation question semantic understanding feature vector.

3. The information consulting service system based on big data according to claim 2 is characterized in that: The alternative answer semantic understanding module is used to: Performing word segmentation processing on the first candidate answer to obtain a sequence of first candidate answer words; Passing the sequence of the first candidate answer words through the word embedding layer to obtain a sequence of first candidate answer word embedding vectors; Inputting the sequence of the first candidate answer word embedding vectors into the semantic encoder to obtain a plurality of first candidate answer semantic feature vectors; as well as The plurality of first candidate answer semantic feature vectors are concatenated to obtain the first candidate answer semantic understanding feature vector.

4. The information consulting service system based on big data according to claim 3 is characterized in that: The semantic association module is used to: The first candidate answer semantic understanding feature vector and the government affairs consultation question semantic understanding feature vector are associated and encoded using the following association coding formula to obtain the semantic matching feature matrix; Wherein, the association coding formula is: Among them, V a represents the semantic understanding feature vector of the first candidate answer, represents the transposed vector of the semantic understanding feature vector of the first candidate answer, V b represents the semantic understanding feature vector of the government consultation question, M represents the semantic matching feature matrix, Represents vector multiplication.

5. The information consulting service system based on big data according to claim 4 is characterized in that: The classification loss unit is used to: The classifier is used to process the optimized training semantic matching feature vector using the following classification loss formula to obtain a training classification result, wherein the classification loss formula is: softmax{(W n ,B n ):…:(W1,B1)|V)}, where V represents the optimized training semantic matching feature vector, W1 to W n is the weight matrix of each fully connected layer, B1 to B n Represents the bias matrix of each fully connected layer; as well as A cross entropy value between the training classification result and the true value is calculated as the classification loss function value.

6. An information consulting service method based on big data, characterized in that: include: Obtain government consultation questions; Performing word segmentation processing on the government affairs consultation question to obtain a sequence of government affairs consultation words; Passing the sequence of government affairs consultation words through a word embedding layer to obtain a sequence of government affairs consultation word embedding vectors; The sequence of embedding vectors of the government consultation words is passed through a semantic encoder based on a Transformer module and a BiLSTM model to obtain a semantic understanding feature vector of the government consultation question; Get the first alternative answer; After word segmentation processing is performed on the first candidate answer, the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model are passed to obtain a semantic understanding feature vector of the first candidate answer; Performing association coding on the semantic understanding feature vector of the first candidate answer and the semantic understanding feature vector of the government affairs consultation question to obtain a semantic matching feature matrix; Performing soft matching of the rank arrangement distribution of the feature scale as the imitation mask on the semantic matching feature matrix to obtain a semantic matching feature matrix; as well as Passing the semantic matching feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether to return the first candidate answer; The method further comprises a training step of training the semantic encoder based on the Transformer module and the BiLSTM model and the classifier; The training step comprises: Acquire training data, the training data including training government affairs consultation questions, training first candidate answers, and whether to return a true value of the first candidate answer; Performing word segmentation processing on the training government affairs consultation question to obtain a sequence of training government affairs consultation words; Passing the sequence of training government affairs consultation words through the word embedding layer to obtain a sequence of training government affairs consultation word embedding vectors; Passing the sequence of embedding vectors of the training government affairs consultation words through the semantic encoder based on the Transformer module and the BiLSTM model to obtain a semantic understanding feature vector of the training government affairs consultation question; After word segmentation processing is performed on the first candidate answer for training, the word embedding layer and the semantic encoder based on the Transformer module and the BiLSTM model are passed through to obtain a semantic understanding feature vector of the first candidate answer for training; Performing association coding on the semantic understanding feature vector of the training first candidate answer and the semantic understanding feature vector of the training government affairs consultation question to obtain a training semantic matching feature matrix; Iteratively optimizing the training semantic matching feature vector after the training semantic matching feature matrix is ​​expanded to obtain an optimized training semantic matching feature vector; Passing the optimized training semantic matching feature vector through the classifier to obtain a classification loss function value; and Training the semantic encoder based on the Transformer module and the BiLSTM model and the classifier by back propagation of the classification loss function value through gradient descent; The training semantic matching feature vector after the training semantic matching feature matrix is ​​expanded is iteratively optimized using the following iterative optimization formula to obtain the optimized training semantic matching feature vector; wherein the iterative optimization formula is: Among them, v i is the i-th eigenvalue of the training semantic matching feature vector, φ(v i ) represents the probabilistic function of the eigenvalue, that is, the eigenvalue v i The probability function mapped to the interval [0,1], p is the class probability value obtained by the classifier, L is the length of the semantic matching feature vector, and α is the weight hyperparameter, v' i is the i-th eigenvalue of the optimized training semantic matching feature vector.

Images