Hybrid expert multi-classification method and system combined with embedded double-model organizational structure

Abstract

The invention discloses a hybrid expert multi-classification method and system combined with an embedded double-model organization structure, is applied to the technical field of data processing, and constructs a scientific value sentence recognition model by performing deep learning on a pre-training model based on a data distribution consistency threshold and a scientific value sentence training corpus. Based on a hybrid expert mechanism, a scientific value sentence multi-classification model is built, and the scientific value sentence multi-classification model comprises an academic value expert sub-model, an application value expert sub-model, an innovation value expert sub-model and a gating network. And packaging the scientific value sentence recognition model and the scientific value sentence multi-classification model into a scientific value sentence detection double model to obtain scientific and technical literatures. And inputting the scientific and technical literature into the scientific value sentence detection double model to obtain a scientific value sentence comprehensive classification result. The technical problems that in the prior art, when an existing scientific literature value sentence classification method is used for processing scientific literatures, the processing efficiency is low, and the accuracy is difficult to guarantee are solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a hybrid expert multi-classification method and system combined with an embedded dual-model organizational structure. Background Art

[0002] With the increasing number of scientific and technological documents, how to efficiently and accurately identify and classify scientific value sentences in documents, such as academic value, application value, and innovation value sentences, has become an important issue in scientific research and document management. Traditional document classification methods mostly rely on manual annotation and rule setting. Although they can partially meet the needs, they are often inefficient and difficult to guarantee accuracy when processing large-scale documents.

[0003] Therefore, the existing scientific literature value sentence classification methods in the prior art have technical problems of low processing efficiency and difficulty in ensuring accuracy when processing scientific and technological literature. Summary of the invention

[0004] This application solves the technical problems of low processing efficiency and difficulty in ensuring accuracy in the existing scientific literature value sentence classification methods in the prior art when processing scientific literature by providing a hybrid expert multi-classification method and system combined with an embedded dual-model organizational structure. By introducing a hybrid expert mechanism and optimizing data distribution consistency, the classification accuracy of scientific value sentences can be effectively improved, achieving the technical effect of efficient and accurate analysis of multi-dimensional academic, application, and innovation values.

[0005] The present application provides a hybrid expert multi-classification method combined with an embedded dual-model organizational structure, the method comprising: performing deep learning on a pre-trained model based on a data distribution consistency threshold and a scientific value sentence training corpus to construct a scientific value sentence recognition model. Based on a hybrid expert mechanism, a multi-classification model for scientific value sentences is constructed, wherein the multi-classification model for scientific value sentences includes an academic value expert sub-model, an application value expert sub-model, an innovation value expert sub-model, and a gating network. The scientific value sentence recognition model and the scientific value sentence multi-classification model are encapsulated into a scientific value sentence detection dual model. Scientific and technological literature is obtained. The scientific and technological literature is input into the scientific value sentence detection dual model to obtain a comprehensive classification result for scientific value sentences.

[0006] In the implementation method, the pre-trained model is deeply learned based on the data distribution consistency threshold and the scientific value sentence training corpus to construct a scientific value sentence recognition model, including: the scientific value sentence training corpus includes a scientific value sentence sample set and a non-scientific value sentence sample set. The scientific value sentence training corpus is divided according to a predetermined ratio to obtain a first corpus training set, a first corpus test set and a first corpus verification set. Based on the data distribution consistency threshold, the data distribution of the first corpus training set, the first corpus test set and the first corpus verification set is optimized to obtain a second corpus training set, a second corpus test set and a second corpus verification set. Based on the second corpus training set, the second corpus test set and the second corpus verification set, the pre-trained model is trained, tested and verified to generate the scientific value sentence recognition model.

[0007] In the implementation, based on the data distribution consistency threshold, the data distribution of the first corpus training set, the first corpus test set and the first corpus verification set is optimized to obtain the second corpus training set, the second corpus test set and the second corpus verification set, including: respectively calculating the positive and negative sample distribution of the first corpus training set, the first corpus test set and the first corpus verification set to obtain the first data distribution coefficient, the second data distribution coefficient and the third data distribution coefficient. Based on the first data distribution coefficient, the second data distribution coefficient and the third data distribution coefficient, the data distribution consistency of the first corpus training set, the first corpus test set and the first corpus verification set is evaluated to obtain the data distribution consistency coefficient. It is determined whether the data distribution consistency coefficient is greater than the data distribution consistency threshold. If the data distribution consistency coefficient is greater than the data distribution consistency threshold, the data distribution of the first corpus training set, the first corpus test set and the first corpus verification set is adjusted according to the data distribution consistency threshold to generate the second corpus training set, the second corpus test set and the second corpus verification set.

[0008] In the implementation method, based on the hybrid expert mechanism, a multi-classification model of scientific value sentences is constructed, including: obtaining multi-classification indicators of scientific value sentences, wherein the multi-classification indicators of scientific value sentences include academic value, application value and innovation value. Based on the multi-classification indicators of scientific value sentences, load the academic value sentence scoring corpus, the application value sentence scoring corpus and the innovation value sentence scoring corpus. Based on the academic value sentence scoring corpus, train the academic value expert sub-model. Based on the application value sentence scoring corpus, train the application value expert sub-model. Based on the innovation value expert sub-model, train the innovation value expert sub-model. Based on the gated loss function, construct the gated network according to the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model. The academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model are used as multi-classification parallel nodes. Connect the multi-classification parallel nodes and the gated network to generate the multi-classification model of scientific value sentences.

[0009] In the implementation method, based on the gating loss function, the gating network is constructed according to the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model, including: collecting the output samples of the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model to obtain an expert output sample set. Collecting the discrimination probability parameters corresponding to the expert output sample set to obtain the discrimination probability distribution. Minimizing the gating loss is the training goal of the gating network. Based on the gating loss function and the gating network training goal, the expert output sample set and the discrimination probability distribution are unsupervisedly trained to generate the gating network.

[0010] In an implementation, the gating loss function is:

[0011]

[0012] in, Characterize the gated loss function, represents the unsupervised loss, λ represents the predetermined balance weight, Characterize the minimized regularization term.

[0013] In the implementation method, the scientific and technological literature is input into the scientific value sentence detection dual model to obtain a comprehensive classification result of the scientific value sentences, including: locating the scientific value sentence enrichment area based on the scientific and technological literature to obtain the value sentence enrichment area. The value sentence enrichment area is input into the scientific value sentence recognition model to obtain the scientific value sentence recognition result. The scientific value sentence recognition result is input into the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model to obtain multi-type value discrimination results. The multi-type value discrimination results are input into the gated network to generate the comprehensive classification result of the scientific value sentences.

[0014] The present application also provides a hybrid expert multi-classification system combined with an embedded dual-model organizational architecture, characterized in that the system includes:

[0015] The scientific value sentence recognition model construction module is used to perform deep learning on the pre-trained model based on the data distribution consistency threshold and the scientific value sentence training corpus to build a scientific value sentence recognition model.

[0016] The multi-classification model building module of scientific value sentences is used to build a multi-classification model of scientific value sentences based on a hybrid expert mechanism, wherein the multi-classification model of scientific value sentences includes an academic value expert sub-model, an application value expert sub-model, an innovation value expert sub-model and a gating network.

[0017] The model encapsulation module is used to encapsulate the scientific value sentence recognition model and the scientific value sentence multi-classification model into a scientific value sentence detection dual model.

[0018] The scientific and technological literature acquisition module is used to obtain scientific and technological literature.

[0019] The comprehensive classification result acquisition module is used to input the scientific literature into the scientific value sentence detection dual model to obtain the comprehensive classification result of the scientific value sentence.

[0020] It is proposed to use the hybrid expert multi-classification method and system proposed in this application in combination with an embedded dual-model organizational structure, and to construct a scientific value sentence recognition model by performing deep learning on the pre-trained model based on the data distribution consistency threshold and the scientific value sentence training corpus. Based on the hybrid expert mechanism, a scientific value sentence multi-classification model is built, wherein the scientific value sentence multi-classification model includes an academic value expert sub-model, an application value expert sub-model, an innovation value expert sub-model and a gating network. The scientific value sentence recognition model and the scientific value sentence multi-classification model are encapsulated into a scientific value sentence detection dual model. Obtain scientific and technological literature. The scientific and technological literature is input into the scientific value sentence detection dual model to obtain a comprehensive classification result of the scientific value sentence. The technical problem of low processing efficiency and difficulty in ensuring accuracy in the existing scientific literature value sentence classification method in the prior art when processing scientific and technological literature is solved. By introducing a hybrid expert mechanism and data distribution consistency optimization, the classification accuracy of scientific value sentences can be effectively improved, and the technical effect of efficient and accurate analysis of multi-dimensional academic, application, and innovation values ​​can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings of the embodiments of the present disclosure. Obviously, the drawings described below only relate to some embodiments of the present disclosure, and are not intended to limit the present disclosure.

[0022] Figure 1 It is a flow chart of the hybrid expert multi-classification method combined with the embedded dual-model organizational architecture of the present invention;

[0023] Figure 2 A schematic diagram of the structure of a hybrid expert multi-classification system combined with an embedded dual-model organizational architecture provided in an embodiment of the present application;

[0024] Explanation of the accompanying drawings: scientific value sentence recognition model construction module 11, scientific value sentence multi-classification model construction module 12, model encapsulation module 13, scientific and technological literature acquisition module 14, comprehensive classification result acquisition module 15. DETAILED DESCRIPTION

[0025] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0026] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0027] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0028] The present application embodiment provides a hybrid expert multi-classification method and system combined with an embedded dual-model organizational structure, such as Figure 1 As shown, the method includes:

[0029] Based on the data distribution consistency threshold and scientific value sentence training corpus, the pre-trained model is deeply learned to construct a scientific value sentence recognition model.

[0030] Based on the hybrid expert mechanism, a multi-classification model of scientific value sentences is constructed, wherein the multi-classification model of scientific value sentences includes an academic value expert sub-model, an application value expert sub-model, an innovation value expert sub-model and a gating network.

[0031] In scientific and technological literature, knowledge is not evenly distributed, but shows a certain concentration and regularity. In order to accurately classify scientific and technological literature into academic value sentences, application value sentences, and innovative value sentences, the pre-trained model is deeply learned based on the data distribution consistency threshold and the scientific value sentence training corpus to construct a scientific value sentence recognition model, wherein the data distribution consistency threshold is a preset value used to evaluate whether the sample distribution of the training, test, and verification data sets is balanced. When the category distribution of the data set is inconsistent with the category distribution of the overall data set, the data set is adjusted to meet the threshold. The scientific value sentence training corpus is a text collection specially prepared for training the scientific value sentence recognition model, including sentence samples marked as having or not having scientific value. Subsequently, based on the hybrid expert mechanism, a multi-classification model of scientific value sentences is built, wherein the multi-classification model of scientific value sentences includes an academic value expert sub-model, an application value expert sub-model, an innovative value expert sub-model, and a gated network. The hybrid expert mechanism is an integrated learning framework in which multiple sub-models (experts) evaluate and classify different categories of scientific value sentences according to their expertise. By combining multiple expert models and learning how to adaptively assign tasks to different experts based on the input, adaptive decomposition and modeling of tasks can be achieved. Through the collaboration of multiple expert models, input representation learning and decision output can be performed from different perspectives.

[0032] The method provided in the embodiment of the present application also includes: the scientific value sentence training corpus includes a scientific value sentence sample set and a non-scientific value sentence sample set. The scientific value sentence training corpus is divided according to a predetermined ratio to obtain a first corpus training set, a first corpus test set and a first corpus verification set. Based on the data distribution consistency threshold, the data distribution of the first corpus training set, the first corpus test set and the first corpus verification set is optimized to obtain a second corpus training set, a second corpus test set and a second corpus verification set. Based on the second corpus training set, the second corpus test set and the second corpus verification set, the pre-trained model is trained, tested and verified to generate the scientific value sentence recognition model.

[0033] The scientific value sentence training corpus is a sample sentence selected from scientific and technological literature. These sentences are marked as sentences with or without scientific value, that is, a sample set of scientific value sentences and a sample set of non-scientific value sentences. Subsequently, the scientific value sentence training corpus is divided according to a predetermined ratio to obtain a first corpus training set, a first corpus test set, and a first corpus verification set. The first corpus training set, the first corpus test set, and the first corpus verification set are used for model training, testing, and verification, respectively. Among them, the predetermined ratio is a preset data division ratio, which can be divided according to a ratio of 70% training, 15% testing, and 15% verification. Further, based on the data distribution consistency threshold, the data distribution of the first corpus training set, the first corpus test set, and the first corpus verification set is optimized to ensure that each category is evenly distributed in the training, testing, and verification sets, and the second corpus training set, the second corpus test set, and the second corpus verification set are obtained. Based on the second corpus training set, the second corpus test set and the second corpus verification set, the pre-trained model is trained, tested and verified. The pre-trained model is an untrained neural network model until the second corpus verification set is used for model verification, and the verification is passed when the output meets the preset accuracy, and a scientific value sentence recognition model is obtained. The scientific value sentence recognition model is mainly used to determine whether the input sentence type is a scientific value sentence. The main function of this layer model is to filter out a large number of irrelevant sentences and provide high-quality corpus for subsequent classification tasks.

[0034] The positive and negative sample distributions are calculated for the first corpus training set, the first corpus test set and the first corpus verification set respectively to obtain a first data distribution coefficient, a second data distribution coefficient and a third data distribution coefficient. Based on the first data distribution coefficient, the second data distribution coefficient and the third data distribution coefficient, the data distribution consistency of the first corpus training set, the first corpus test set and the first corpus verification set is evaluated to obtain a data distribution consistency coefficient. It is determined whether the data distribution consistency coefficient is greater than the data distribution consistency threshold. If the data distribution consistency coefficient is greater than the data distribution consistency threshold, the data distribution of the first corpus training set, the first corpus test set and the first corpus verification set is adjusted according to the data distribution consistency threshold to generate the second corpus training set, the second corpus test set and the second corpus verification set.

[0035] Based on the data distribution consistency threshold, data distribution optimization is performed on the first corpus training set, the first corpus test set and the first corpus verification set to obtain the second corpus training set, the second corpus test set and the second corpus verification set, including: respectively calculating the positive and negative sample distribution of the first corpus training set, the first corpus test set and the first corpus verification set to obtain the first data distribution coefficient, the second data distribution coefficient and the third data distribution coefficient. Among them, the first data distribution coefficient = the sample size of the scientific value sentences in the first corpus training set ÷ the sample size of the non-scientific value sentences in the first corpus training set, and the calculation process of the second data distribution coefficient and the third data distribution coefficient is the same as the calculation process of the first data distribution coefficient. Based on the first data distribution coefficient, the second data distribution coefficient and the third data distribution coefficient, the data distribution consistency evaluation is performed on the first corpus training set, the first corpus test set and the first corpus verification set. When evaluating the data distribution consistency, the difference between the first data distribution coefficient and the second data distribution coefficient and the third data distribution coefficient is calculated to obtain the data distribution consistency coefficient. Subsequently, it is determined whether the data distribution consistency coefficient is greater than the data distribution consistency threshold. The data distribution consistency threshold is a preset maximum judgment threshold for the consistency of the data distribution difference. When it is less than or equal to the threshold, the data distribution consistency of the two groups of data is high. When it is greater than the threshold, the data distribution consistency of the two groups of data is low, and data distribution optimization is required. If the data distribution consistency coefficient is greater than the data distribution consistency threshold, the first corpus training set, the first corpus test set, and the first corpus verification set are adjusted for distribution compensation according to the data distribution consistency coefficient according to the data distribution consistency threshold, and the existing difference ratio is supplemented with data until it is less than the data distribution consistency threshold. Generate the second corpus training set, the second corpus test set, and the second corpus verification set.

[0036] The method provided in the embodiment of the present application also includes: obtaining a multi-classification index of scientific value sentences, wherein the multi-classification index of scientific value sentences includes academic value, application value and innovation value. Based on the multi-classification index of scientific value sentences, load the academic value sentence scoring corpus, the application value sentence scoring corpus and the innovation value sentence scoring corpus. Based on the academic value sentence scoring corpus, train the academic value expert sub-model. Based on the application value sentence scoring corpus, train the application value expert sub-model. Based on the innovation value expert sub-model, train the innovation value expert sub-model. Based on the gating loss function, construct the gating network according to the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model. The academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model are multi-classification parallel nodes. Connect the multi-classification parallel nodes and the gating network to generate the multi-classification model of scientific value sentences.

[0037] Based on the hybrid expert mechanism, a multi-classification model of scientific value sentences is constructed, including: obtaining multi-classification indicators of scientific value sentences, wherein the multi-classification indicators of scientific value sentences include academic value, application value and innovation value. The academic value reflects the contribution of the sentence to the academic community, such as the importance of the proposed theory or experimental results. The application value reflects the potential or practical benefits of the research or discovery described in the sentence in practical applications. The innovation value is to evaluate the degree of innovation and originality of the research, methods or results mentioned in the sentence.

[0038] Subsequently, based on the multi-classification index of the scientific value sentences, the academic value sentence scoring corpus, the application value sentence scoring corpus and the innovation value sentence scoring corpus are loaded. The academic value sentence scoring corpus includes sentences marked as high academic value and corresponding value scoring identifiers, such as sentences that deeply explore specific scientific issues. The application value sentence scoring corpus includes sentences marked as high application value and corresponding value scoring identifiers, such as sentences that describe research results with practical application prospects. The innovation value sentence scoring corpus includes sentences marked as high innovation value and corresponding value scoring identifiers, such as sentences that propose new theories or new methods. Further, based on the academic value sentence scoring corpus, the academic value expert sub-model is trained. Based on the application value sentence scoring corpus, the application value expert sub-model is trained. Based on the innovation value expert sub-model, the innovation value expert sub-model is trained. The academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model are all obtained after supervised training by a neural network model. Based on the gating loss function, the gating network is constructed according to the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model. The academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model are multi-classification parallel nodes. The multi-classification parallel nodes and the gating network are connected to generate the multi-classification model of scientific value sentences. Given an input x, the output y of the multi-classification model of scientific value sentences can be expressed as:

[0039]

[0040] Among them, f j (x) represents the output of the jth expert model for input x. g(x) j is the corresponding gating function. The gating mechanism plays the role of soft routing, adaptively assigning inputs to different experts based on the characteristics of the input, and performing weighted combinations of their outputs. g(x) is usually implemented using the Softmax function:

[0041]

[0042] Among them, e j It represents the gated logit value of the jth expert, which can be obtained by linear transformation or feedforward network mapping of x. In the multi-classification task of scientific value, this project sets up three expert models for academic value, application value, and innovation value.

[0043] The method provided in the embodiment of the present application also includes: collecting output samples of the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model to obtain an expert output sample set. Collecting the discrimination probability parameters corresponding to the expert output sample set to obtain a discrimination probability distribution. Minimizing the gated loss is used as the training goal of the gated network. Based on the gated loss function and the gated network training goal, the expert output sample set and the discrimination probability distribution are unsupervisedly trained to generate the gated network.

[0044] Based on the gating loss function, the gating network is constructed according to the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model, including: collecting the output samples of the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model to obtain an expert output sample set. The expert output sample set is a collection of output samples obtained from each expert sub-model (academic value, application value, innovation value expert sub-model). Each sample set contains the scoring or classification results of a specific scientific value sentence, which reflects the attributes of the sentence in each value dimension. The discriminant probability parameters corresponding to the expert output sample set are collected to obtain the discriminant probability distribution. Taking minimizing the gating loss as the training goal of the gating network, based on the gating loss function and the gating network training goal, the expert output sample set and the discriminant probability distribution are unsupervisedly trained to generate the gating network.

[0045] The gating network plays a key routing role in the multi-classification model of scientific value sentences. Its goal is to adaptively assign value sentences to different experts based on their characteristics. Unlike the supervised training of expert models, it is difficult for the gating network to directly obtain supervision signals because the true expert distribution ratio of each value is unknown. Therefore, an unsupervised collaborative training method is adopted, using the output of the expert model as a soft label to guide the learning of the gating network. In each training batch, we first use three expert models to classify the input value sentences x1, x2, ..., x b Perform inferences separately to obtain their discriminant probabilities p in their respective value dimensions ij . p ij =Sigmoid(f j (x i )),i=1,2,...,b,j=1,2,3.

[0046] These probability outputs are then used as soft labels and matched with the outputs of the gated network by JS divergence to obtain the unsupervised loss function of the gated network:

[0047]

[0048] Among them, the JS divergence is defined as:

[0049]

[0050] By minimizing the loss The gated network learns to adaptively weight the experts according to the problem characteristics, so that their weighted output is as close as possible to the discrimination probability of each expert. This collaborative learning mechanism enables the gated network and the expert model to be mutually applicable and iteratively updated during the training process, ultimately achieving overall performance improvement of the model.

[0051] During training, a regularization term based on expert entropy is introduced to encourage the output of the gating network to have higher expert selectivity:

[0052]

[0053] Where H(·) represents the entropy function. Minimize the regularization term R g Encourage the gating network to select only a few experts for each problem as much as possible, avoid egalitarianism, and improve the efficiency of expert utilization.

[0054] Ultimately, the overall goal of the gating network is to minimize the unsupervised collaborative loss and the expert entropy regularization term, where the gating loss function is:

[0055]

[0056] in, Characterize the gated loss function, represents the unsupervised loss, λ represents the predetermined balance weight, Representation minimizes the regularization term. In summary, the expert model realizes the discrimination of different types of values ​​through supervised training of scientific value sentences. The gated network realizes the adaptive expert routing strategy through collaborative learning and entropy regularization. The expert model and the gated network cooperate with each other and optimize collaboratively to form a multi-classification method for scientific value sentences.

[0057] The scientific value sentence recognition model and the scientific value sentence multi-classification model are encapsulated into a scientific value sentence detection dual model. Scientific and technological literature is obtained. The scientific and technological literature is input into the scientific value sentence detection dual model to obtain a comprehensive classification result of scientific value sentences.

[0058] The scientific value sentence recognition model and the scientific value sentence multi-classification model are encapsulated into a scientific value sentence detection dual model. Subsequently, scientific and technological literature is obtained, for example, a research paper on biodiversity is uploaded. The scientific and technological literature is input into the scientific value sentence detection dual model, and the scientific value sentence detection dual model is used for analysis to obtain a comprehensive classification result of the scientific value sentences. The technical problems of low processing efficiency and difficulty in ensuring accuracy in the existing scientific literature value sentence classification methods in the prior art when processing scientific and technological literature are solved. By introducing a hybrid expert mechanism and data distribution consistency optimization, the classification accuracy of scientific value sentences can be effectively improved, and the technical effect of efficient and accurate analysis of multi-dimensional academic, application, and innovation values ​​can be achieved. For example, given an embedding representation x of a scientific problem value sentence, three expert models f 1,2,3 (x) outputs the discrimination scores of the three value dimensions respectively. The gated network g(x) calculates the weights of each expert according to the semantic features of the question and finally obtains the comprehensive classification result:

[0059] y=g(x)1f1(x)+g(x)2f2(x)+g(x)3f3(x)

[0060] Among them, f1(x), f2(x) and f3(x) represent the output of experts on academic value, application value and innovation value respectively.

[0061] The method provided in the embodiment of the present application also includes: locating the scientific value sentence enrichment area based on the scientific and technological literature to obtain the value sentence enrichment area. Inputting the value sentence enrichment area into the scientific value sentence recognition model to obtain the scientific value sentence recognition result. Inputting the scientific value sentence recognition result into the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model to obtain multi-type value discrimination results. Inputting the multi-type value discrimination results into the gating network to generate the comprehensive classification result of the scientific value sentence.

[0062] The scientific and technological literature is input into the scientific value sentence detection dual model to obtain the comprehensive classification results of scientific value sentences, including: in scientific and technological literature, knowledge is not evenly distributed, but presents a certain concentration and regularity. Certain specific chapters or positions often contain a large amount of key information and core knowledge, with the characteristics of high knowledge density, large amount of information, and important content. These areas are called knowledge-rich areas. Based on the scientific and technological literature, the scientific value sentence enrichment area is located to obtain the value sentence enrichment area. The value sentence enrichment area is input into the scientific value sentence recognition model to obtain the scientific value sentence recognition result, and the sentences with scientific value are obtained. The scientific value sentence recognition result is input into the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model to obtain multi-type value discrimination results. Finally, the multi-type value discrimination results are input into the gated network to generate the comprehensive classification results of the scientific value sentences, and the comprehensive classification results of the scientific value sentences are the comprehensive classification results of the scientific value sentences belonging to the academic value sentences, application value sentences, and innovation value sentences.

[0063] In the above, refer to Figure 1 The hybrid expert multi-classification method combined with the embedded dual-model organizational architecture according to an embodiment of the present invention is described in detail. Figure 2 A hybrid expert multi-classification system incorporating an embedded dual-model organizational architecture according to an embodiment of the present invention is described.

[0064] The hybrid expert multi-classification system combined with an embedded dual-model organizational structure according to the embodiment of the present invention solves the technical problems of low processing efficiency and difficulty in ensuring accuracy in the existing scientific literature value sentence classification methods in the prior art when processing scientific literature. By introducing a hybrid expert mechanism and optimizing data distribution consistency, the classification accuracy of scientific value sentences can be effectively improved, and the technical effect of efficient and accurate analysis of multi-dimensional academic, application, and innovation values ​​can be achieved. The hybrid expert multi-classification system combined with an embedded dual-model organizational structure includes: a scientific value sentence recognition model construction module 11, a scientific value sentence multi-classification model construction module 12, a model encapsulation module 13, a scientific literature acquisition module 14, and a comprehensive classification result acquisition module 15.

[0065] The scientific value sentence recognition model construction module 11 is used to perform deep learning on the pre-trained model based on the data distribution consistency threshold and the scientific value sentence training corpus to construct a scientific value sentence recognition model.

[0066] The scientific value sentence multi-classification model construction module 12 is used to build a scientific value sentence multi-classification model based on a hybrid expert mechanism, wherein the scientific value sentence multi-classification model includes an academic value expert sub-model, an application value expert sub-model, an innovation value expert sub-model and a gating network.

[0067] The model encapsulation module 13 is used to encapsulate the scientific value sentence recognition model and the scientific value sentence multi-classification model into a scientific value sentence detection dual model.

[0068] The scientific and technological literature acquisition module 14 is used to obtain scientific and technological literature.

[0069] The comprehensive classification result acquisition module 15 is used to input the scientific literature into the scientific value sentence detection dual model to obtain the comprehensive classification result of the scientific value sentence.

[0070] The specific configuration of the scientific value sentence recognition model construction module 11 will be described in detail below. The scientific value sentence recognition model construction module 11 may further include: based on the data distribution consistency threshold and the scientific value sentence training corpus, deep learning is performed on the pre-trained model to construct a scientific value sentence recognition model, including: the scientific value sentence training corpus includes a scientific value sentence sample set and a non-scientific value sentence sample set. The scientific value sentence training corpus is divided according to a predetermined ratio to obtain a first corpus training set, a first corpus test set and a first corpus verification set. Based on the data distribution consistency threshold, the data distribution of the first corpus training set, the first corpus test set and the first corpus verification set is optimized to obtain a second corpus training set, a second corpus test set and a second corpus verification set. Based on the second corpus training set, the second corpus test set and the second corpus verification set, the pre-trained model is trained, tested and verified to generate the scientific value sentence recognition model.

[0071] The specific configuration of the scientific value sentence recognition model construction module 11 will be described in detail below. The scientific value sentence recognition model construction module 11 further includes: based on the data distribution consistency threshold, the data distribution optimization is performed on the first corpus training set, the first corpus test set and the first corpus verification set to obtain the second corpus training set, the second corpus test set and the second corpus verification set, including: respectively performing positive and negative sample distribution calculations on the first corpus training set, the first corpus test set and the first corpus verification set to obtain the first data distribution coefficient, the second data distribution coefficient and the third data distribution coefficient. Based on the first data distribution coefficient, the second data distribution coefficient and the third data distribution coefficient, the data distribution consistency of the first corpus training set, the first corpus test set and the first corpus verification set is evaluated to obtain the data distribution consistency coefficient. It is determined whether the data distribution consistency coefficient is greater than the data distribution consistency threshold. If the data distribution consistency coefficient is greater than the data distribution consistency threshold, the data distribution of the first corpus training set, the first corpus test set and the first corpus verification set is adjusted according to the data distribution consistency threshold to generate the second corpus training set, the second corpus test set and the second corpus verification set.

[0072] The specific configuration of the scientific value sentence multi-classification model construction module 12 will be described in detail below. The scientific value sentence multi-classification model construction module 12 may further include: building a scientific value sentence multi-classification model based on a hybrid expert mechanism, including: obtaining a scientific value sentence multi-classification index, wherein the scientific value sentence multi-classification index includes academic value, application value and innovation value. Based on the scientific value sentence multi-classification index, load the academic value sentence scoring corpus, the application value sentence scoring corpus and the innovation value sentence scoring corpus. Based on the academic value sentence scoring corpus, train the academic value expert sub-model. Based on the application value sentence scoring corpus, train the application value expert sub-model. Based on the innovation value expert sub-model, train the innovation value expert sub-model. Based on the gated loss function, construct the gated network according to the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model. The academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model are multi-classification parallel nodes. Connect the multi-classification parallel nodes and the gated network to generate the scientific value sentence multi-classification model.

[0073] Below, the specific configuration of the scientific value sentence multi-classification model construction module 12 will be described in detail. The scientific value sentence multi-classification model construction module 12 further includes: based on the gating loss function, according to the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model, constructing the gating network, including: collecting the output samples of the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model to obtain the expert output sample set. Collect the discriminant probability parameters corresponding to the expert output sample set to obtain the discriminant probability distribution. Minimizing the gating loss is the training goal of the gated network. Based on the gating loss function and the gated network training goal, the expert output sample set and the discriminant probability distribution are unsupervisedly trained to generate the gated network.

[0074] The specific configuration of the scientific value sentence multi-classification model construction module 12 will be described in detail below. The scientific value sentence multi-classification model construction module 12 further includes: the gate loss function is:

[0075]

[0076] in, Characterize the gated loss function, represents the unsupervised loss, λ represents the predetermined balance weight, Characterize the minimized regularization term.

[0077] Below, the specific configuration of the scientific value sentence comprehensive classification result acquisition module 15 will be described in detail. The comprehensive classification result acquisition module 15 further includes: inputting the scientific and technological literature into the scientific value sentence detection dual model to obtain the scientific value sentence comprehensive classification result, including: locating the scientific value sentence enrichment area based on the scientific and technological literature to obtain the value sentence enrichment area. Inputting the value sentence enrichment area into the scientific value sentence recognition model to obtain the scientific value sentence recognition result. Inputting the scientific value sentence recognition result into the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model to obtain multi-type value discrimination results. Inputting the multi-type value discrimination results into the gated network to generate the scientific value sentence comprehensive classification result.

[0078] The hybrid expert multi-classification system combined with an embedded dual-model organizational structure provided in an embodiment of the present invention can execute the hybrid expert multi-classification method combined with an embedded dual-model organizational structure provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0079] 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 the server, and the various units and modules included are only divided according to functional logic, but are not limited to the above divisions, as long as the corresponding functions can be achieved. In addition, the specific names of the functional units are only for the convenience of distinguishing from each other, and are not used to limit the scope of protection of the present invention.

[0080] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A hybrid expert multi-classification method combined with an embedded dual-model organizational architecture, characterized by: The method comprises: Based on the data distribution consistency threshold and scientific value sentence training corpus, the pre-training model is deeply learned to build a scientific value sentence recognition model; Based on the hybrid expert mechanism, a multi-classification model of scientific value sentences is constructed, wherein the multi-classification model of scientific value sentences includes an academic value expert sub-model, an application value expert sub-model, an innovation value expert sub-model and a gating network; Encapsulating the scientific value sentence recognition model and the scientific value sentence multi-classification model into a scientific value sentence detection dual model; Access to scientific and technological literature; The scientific and technological literature is input into the dual model for detecting scientific value sentences to obtain a comprehensive classification result of scientific value sentences.

2. The method according to claim 1, characterized in that Based on the data distribution consistency threshold and scientific value sentence training corpus, the pre-trained model is deeply learned to build a scientific value sentence recognition model, including: The scientific value sentence training corpus includes a scientific value sentence sample set and a non-scientific value sentence sample set; Divide the scientific value sentence training corpus according to a predetermined ratio to obtain a first corpus training set, a first corpus test set, and a first corpus verification set; Based on the data distribution consistency threshold, optimizing the data distribution of the first corpus training set, the first corpus test set, and the first corpus verification set to obtain a second corpus training set, a second corpus test set, and a second corpus verification set; Based on the second corpus training set, the second corpus test set and the second corpus verification set, the pre-trained model is trained, tested and verified to generate the scientific value sentence recognition model.

3. The method according to claim 2, characterized in that Based on the data distribution consistency threshold, data distribution optimization is performed on the first corpus training set, the first corpus test set, and the first corpus verification set to obtain a second corpus training set, a second corpus test set, and a second corpus verification set, including: Performing positive and negative sample distribution calculations on the first corpus training set, the first corpus test set, and the first corpus verification set, respectively, to obtain a first data distribution coefficient, a second data distribution coefficient, and a third data distribution coefficient; Based on the first data distribution coefficient, the second data distribution coefficient and the third data distribution coefficient, performing data distribution consistency evaluation on the first corpus training set, the first corpus test set and the first corpus verification set to obtain a data distribution consistency coefficient; Determining whether the data distribution consistency coefficient is greater than the data distribution consistency threshold; If the data distribution consistency coefficient is greater than the data distribution consistency threshold, data distribution adjustment is performed on the first corpus training set, the first corpus test set and the first corpus verification set according to the data distribution consistency threshold to generate the second corpus training set, the second corpus test set and the second corpus verification set.

4. The method according to claim 1, characterized in that Based on the hybrid expert mechanism, a multi-classification model of scientific value sentences is built, including: Obtaining multi-classification indicators of scientific value sentences, wherein the multi-classification indicators of scientific value sentences include academic value, application value and innovation value; Based on the multi-classification index of scientific value sentences, the scoring corpus of academic value sentences, the scoring corpus of application value sentences and the scoring corpus of innovative value sentences are loaded; Based on the academic value sentence scoring corpus, training the academic value expert sub-model; Based on the application value sentence scoring corpus, training the application value expert sub-model; Based on the innovation value expert sub-model, training the innovation value expert sub-model; Based on the gating loss function, constructing the gating network according to the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model; The academic value expert sub-model, the applied value expert sub-model and the innovative value expert sub-model are multi-classification parallel nodes; The multi-classification parallel nodes and the gating network are connected to generate the multi-classification model of scientific value sentences.

5. The method according to claim 4, characterized in that Based on the gating loss function, the gating network is constructed according to the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model, including: Collect output samples of the academic value expert sub-model, the applied value expert sub-model and the innovative value expert sub-model to obtain an expert output sample set; Collecting the discriminant probability parameters corresponding to the expert output sample set to obtain the discriminant probability distribution; The training objective of the gated network is to minimize the gated loss; Based on the gating loss function and the gating network training objective, unsupervised training is performed on the expert output sample set and the discriminant probability distribution to generate the gating network.

6. The method according to claim 4, characterized in that The gating loss function is: in, Characterize the gated loss function, represents the unsupervised loss, λ represents the predetermined balance weight, Characterize the minimized regularization term.

7. The method according to claim 1, characterized in that Input the scientific literature into the dual model for detecting scientific value sentences to obtain comprehensive classification results of scientific value sentences, including: Based on the scientific literature, locate the rich area of ​​scientifically valuable sentences to obtain the rich area of ​​valuable sentences; Inputting the value sentence enrichment region into the scientific value sentence recognition model to obtain a scientific value sentence recognition result; Inputting the scientific value sentence recognition results into the academic value expert sub-model, the application value expert sub-model and the innovation value expert sub-model to obtain multi-type value discrimination results; The multi-type value discrimination results are input into the gating network to generate the comprehensive classification results of the scientific value sentences.

8. A hybrid expert multi-classification system combined with an embedded dual-model organizational architecture, characterized by: The system comprises: The scientific value sentence recognition model construction module is used to conduct deep learning on the pre-trained model based on the data distribution consistency threshold and the scientific value sentence training corpus to build a scientific value sentence recognition model; A scientific value sentence multi-classification model construction module is used to build a scientific value sentence multi-classification model based on a hybrid expert mechanism, wherein the scientific value sentence multi-classification model includes an academic value expert sub-model, an application value expert sub-model, an innovation value expert sub-model and a gating network; A model encapsulation module, used for encapsulating the scientific value sentence recognition model and the scientific value sentence multi-classification model into a scientific value sentence detection dual model; A scientific and technological literature acquisition module is used to obtain scientific and technological literature; The comprehensive classification result acquisition module is used to input the scientific literature into the scientific value sentence detection dual model to obtain the comprehensive classification result of the scientific value sentence.

Images