An information classification method and system based on multiple semantic information

Abstract

The present invention proposes an information classification method and system based on multiple semantic information, which relates to the technical field of natural language processing and machine learning. The method comprises: collecting information data and preprocessing the information data; establishing an information classification model, inputting the preprocessed information data into the information classification model, mapping the information data to a high-dimensional space, reducing the spatial distance of the same semantic information, and expanding the spatial distance of the different semantic information, and outputting the mapped text data; using multiple convolution kernels to extract features of the mapped text data respectively, extracting the dependency relationship information between texts, and optimizing the information classification model according to the text information of the information data and the dependency relationship information between texts; setting a training set and a test set, and training and testing the optimized information classification model; using the trained information classification model to classify the information data to be classified to obtain a classification result.

Description

Technical Field

[0001] The present invention relates to the field of natural language processing and machine learning technology, and in particular to an information classification method and system based on multiple semantic information. Background Art

[0002] As an important customer channel, mobile banking APP plays an important role in the digital transformation of banks. In order to further improve customer experience, mobile banking has added information functions. Considering that various new information is generated every day, in order to better manage information, mobile banking needs to classify information according to its content; considering the amount of information data, relying solely on manual methods is costly and inefficient.

[0003] In summary, there is an urgent need for a technical solution that can overcome the above-mentioned defects and can classify information efficiently and accurately. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the present invention proposes an information classification method and system based on multiple semantic information. The present invention uses artificial intelligence technology to process the information title, extracts information from the title text, and classifies the information after processing it. The process does not require human operation, and is fast, labor-saving, efficient, and low-cost.

[0005] In a first aspect of an embodiment of the present invention, a method for information classification based on multiple semantic information is proposed, the method comprising:

[0006] Collecting information data and preprocessing the information data;

[0007] Establish an information classification model, input the preprocessed information data into the information classification model, map the information data to a high-dimensional space through the information classification model, reduce the spatial distance of the same semantic information, and expand the spatial distance of the different semantic information, and output the mapped text data;

[0008] Using multiple convolution kernels to perform feature extraction on the mapped text data, extract the dependency information between texts, and optimize the information classification model according to the text information of the information data and the dependency information between texts;

[0009] Set up training sets and test sets based on the preprocessed information data, train the optimized information classification model, and use the test set to test the training model;

[0010] The information data to be classified is obtained, and the information classification model that has been trained is used to classify the information data to be classified into categories to obtain the classification results.

[0011] Further, collecting information data and preprocessing the information data include:

[0012] Filter and annotate the collected information data, extract the title text of the information and mark the information category.

[0013] Furthermore, an information classification model is established, and the preprocessed information data is input into the information classification model. The information data is mapped to a high-dimensional space through the information classification model, the spatial distance of the same semantic information is reduced, and the spatial distance of the different semantic information is expanded, and the mapped text data is output, including:

[0014] Set the number of input texts to A and the text length limit to B;

[0015] According to the preprocessed information data, the data input to the model is E text data of length F, with a dimension of [A, B];

[0016] The information data is mapped to a high-dimensional space through the information classification model, and the dimension of the output mapped text data is [E, F, G]; where G represents the dimension corresponding to each character in the text after mapping.

[0017] Furthermore, multiple convolution kernels are used to perform feature extraction on the mapped text data, extract the dependency information between texts, and optimize the information classification model according to the text information of the information data and the dependency information between texts, including:

[0018] A convolution kernel of a first scale is used to extract the dependency relationship between texts outside a certain area corresponding to the texts, and a convolution kernel of a second scale is used to extract the dependency relationship between texts within a certain area corresponding to the texts; wherein the first scale is larger than the second scale;

[0019] The features extracted by multiple convolution kernels are spliced ​​to obtain the final features, and the final features are input into the classifier to obtain the final classification result, and the information classification model is optimized according to the final classification result.

[0020] Furthermore, multiple convolution kernels are used to perform feature extraction on the mapped text data, extract the dependency information between texts, and optimize the information classification model according to the text information of the information data and the dependency information between texts, including:

[0021] After convolving the mapped text data with each convolution kernel, the output data dimension is [E, F, G];

[0022] Calculate the maximum and mean values ​​of the output data according to the dimension G corresponding to each character in the text after mapping, and concatenate the maximum and mean values ​​to obtain the final features corresponding to each convolution kernel, with a dimension of [E, F, 2];

[0023] The final features of all convolution kernels are concatenated to obtain the final features, with a dimension of [E, F, 2*N]; N is the number of convolution kernels;

[0024] The concatenated features are input into the classifier to obtain the final classification result.

[0025] Furthermore, a training set and a test set are set according to the preprocessed information data, the optimized information classification model is trained, and the training model is tested using the test set, including:

[0026] Use the training set to train the optimized information classification model;

[0027] The trained model is tested according to the information titles in the test set to determine whether the classification results are correct; if the accuracy reaches the preset value, the model training is completed.

[0028] In a second aspect of an embodiment of the present invention, an information classification system based on multiple semantic information is proposed, the system comprising:

[0029] A preprocessing module, used for collecting information data and preprocessing the information data;

[0030] A model building module is used to build an information classification model, input the preprocessed information data into the information classification model, map the information data to a high-dimensional space through the information classification model, reduce the spatial distance of the same semantic information, and expand the spatial distance of the different semantic information, and output the mapped text data;

[0031] An optimization module, used to use multiple convolution kernels to perform feature extraction on the mapped text data, extract dependency information between texts, and optimize the information classification model according to the text information of the information data and the dependency information between texts;

[0032] The training module is used to set a training set and a test set according to the preprocessed information data, train the optimized information classification model, and test the training model using the test set;

[0033] The classification module is used to obtain information data to be classified, and use the trained information classification model to classify the information data to be classified to obtain the classification result.

[0034] Furthermore, the preprocessing module is specifically used for:

[0035] Filter and annotate the collected information data, extract the title text of the information and mark the information category.

[0036] Furthermore, the model building module is specifically used for:

[0037] Set the number of input texts to A and the text length limit to B;

[0038] According to the preprocessed information data, the data input to the model is E text data of length F, with a dimension of [A, B];

[0039] The information data is mapped to a high-dimensional space through the information classification model, and the dimension of the output mapped text data is [E, F, G]; where G is the dimension corresponding to each character in the text after mapping.

[0040] Furthermore, the optimization module is specifically used for:

[0041] A convolution kernel of a first scale is used to extract the dependency relationship between texts outside a certain area corresponding to the texts, and a convolution kernel of a second scale is used to extract the dependency relationship between texts within a certain area corresponding to the texts; wherein the first scale is larger than the second scale;

[0042] The features extracted by multiple convolution kernels are spliced ​​to obtain the final features, and the final features are input into the classifier to obtain the final classification result, and the information classification model is optimized according to the final classification result.

[0043] Furthermore, the optimization module is specifically used for:

[0044] After convolving the mapped text data with each convolution kernel, the output data dimension is [E, F, G];

[0045] Calculate the maximum and mean values ​​of the output data according to the corresponding dimensions of each character in the text after mapping, and concatenate the maximum and mean values ​​to obtain the final features corresponding to each convolution kernel, with a dimension of [E, F, 2];

[0046] The final features of all convolution kernels are concatenated to obtain the final features, with a dimension of [E, F, 2*N]; N is the number of convolution kernels;

[0047] The concatenated features are input into the classifier to obtain the final classification result.

[0048] Furthermore, the training module is specifically used for:

[0049] Use the training set to train the optimized information classification model;

[0050] The trained model is tested according to the information titles in the test set to determine whether the classification results are correct; if the accuracy reaches the preset value, the model training is completed.

[0051] In a third aspect of an embodiment of the present invention, a computer device is proposed, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements an information classification method based on multiple semantic information when executing the computer program.

[0052] In a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, an information classification method based on multiple semantic information is implemented.

[0053] The information classification method and system based on multiple semantic information proposed in the present invention extract text features through convolution kernels. The method and system are simple to use, convenient to use, and efficient, and can save a lot of time and labor costs. In the process of information classification processing, an information-enhanced data processing mechanism is adopted, which can fully extract the features in the text, improve the information utilization rate, and improve the performance of the information classification model, thereby effectively improving the accuracy of information classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 It is a flowchart of an information classification method based on multiple semantic information according to an embodiment of the present invention.

[0056] Figure 2 It is a schematic diagram of an information classification process according to an embodiment of the present invention.

[0057] Figure 3 It is a schematic diagram of the model architecture relationship of an embodiment of the present invention.

[0058] Figure 4 It is a schematic diagram of the architecture of an information classification system based on multiple semantic information according to an embodiment of the present invention.

[0059] Figure 5 It is a schematic diagram of the structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0061] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0062] According to an embodiment of the present invention, a method and system for information classification based on multiple semantic information is proposed, which relates to the field of natural language processing and machine learning technology. The present invention uses artificial intelligence technology to process information titles, extracts information from title texts, and processes it to classify the information.

[0063] The principle and spirit of the present invention are explained in detail below with reference to several representative embodiments of the present invention.

[0064] Figure 1 FIG. 1 is a flow chart of an information classification method based on multiple semantic information according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0065] Step S101, collecting information data and preprocessing the information data;

[0066] Step S102, establishing an information classification model, inputting the preprocessed information data into the information classification model, mapping the information data to a high-dimensional space through the information classification model, reducing the spatial distance of the same semantic information, and expanding the spatial distance of the different semantic information, and outputting the mapped text data;

[0067] Step S103, using a plurality of convolution kernels to perform feature extraction on the mapped text data, extracting dependency information between texts, and optimizing the information classification model according to the text information of the information data and the dependency information between texts;

[0068] Step S104, setting a training set and a test set according to the preprocessed information data, training the optimized information classification model, and testing the training model using the test set;

[0069] Step S105, obtaining information data to be classified, and using the trained information classification model to classify the information data to be classified to obtain a classification result.

[0070] In order to explain the above information classification method based on multiple semantic information more clearly, each step is described in detail below.

[0071] Step S101:

[0072] The preprocessing process includes: screening and labeling the collected information data, extracting the title text of the information and labeling the information category.

[0073] Step S102:

[0074] An information classification model is established, and the preprocessed information data is input into the information classification model. The information data is mapped to a high-dimensional space through the information classification model, thereby reducing the spatial distance of the same semantic information and expanding the spatial distance of the different semantic information, and the mapped text data is output.

[0075] The mapping process is as follows:

[0076] Set the number of input texts to A and the text length limit to B;

[0077] According to the preprocessed information data, the data input to the model is E text data of length F, with a dimension of [A, B];

[0078] The information data is mapped to a high-dimensional space through the information classification model, and the dimension of the output mapped text data is [E, F, G]; where G represents the dimension corresponding to each character in the text after mapping.

[0079] There are E texts in total, each of which contains F characters. The dimension of each character after mapping is G, so the output data dimension is [E, F, G].

[0080] Step S103:

[0081] A convolution kernel of a first scale (large scale) is used to extract the dependency between texts outside a certain area corresponding to the texts, and a convolution kernel of a second scale (small scale) is used to extract the dependency between texts within a certain area corresponding to the texts; wherein the first scale is larger than the second scale;

[0082] The features extracted by multiple convolution kernels are spliced ​​to obtain the final features, and the final features are input into the classifier to obtain the final classification result, and the information classification model is optimized according to the final classification result.

[0083] The specific process is as follows:

[0084] After convolving the mapped text data with each convolution kernel, the output data dimension is [E, F, G];

[0085] Calculate the maximum and mean values ​​of the output data according to the dimension G corresponding to each character in the text after mapping, and concatenate the maximum and mean values ​​to obtain the final features corresponding to each convolution kernel, with a dimension of [E, F, 2];

[0086] The final features of all convolution kernels are concatenated to obtain the final features, with a dimension of [E, F, 2*N]; N is the number of convolution kernels;

[0087] The concatenated features are input into the classifier to obtain the final classification result.

[0088] Step S104:

[0089] Use the training set to train the optimized information classification model;

[0090] The trained model is tested according to the information titles in the test set to determine whether the classification results are correct; if the accuracy reaches the preset value, the model training is completed.

[0091] Step S105, information classification:

[0092] Obtain information data to be classified, use the trained information classification model to classify the information data to be classified, and obtain classification results.

[0093] The above-mentioned information classification method extracts text features through convolution kernels. It is simple to use, convenient to use, and efficient, which can save a lot of time and manpower costs. In the process of information classification, the information enhanced data processing mechanism is adopted, which can fully extract the features in the text, improve the information utilization rate, and improve the performance of the information classification model, thereby effectively improving the accuracy of information classification.

[0094] It should be noted that, although the operations of the method of the present invention are described in a specific order in the above embodiments and the accompanying drawings, this does not require or imply that the operations must be performed in the specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0095] In order to explain the above-mentioned information classification method based on multiple semantic information more clearly, a specific embodiment is described below. However, it should be noted that this embodiment is only for better illustrating the present invention and does not constitute an improper limitation to the present invention.

[0096] refer to Figure 2 , is a schematic diagram of the information classification process of a specific embodiment of the present invention. Figure 2 As shown, the specific process is:

[0097] Step S201, training data preprocessing:

[0098] The collected information data is manually screened and labeled.

[0099] Step S202, building a neural network model:

[0100] Build a model for information classification.

[0101] Models such as Figure 3 As shown, in Figure 3 In the figure, Text, Text 1, Text 2, Text 3: text; Embedding: converting discrete variables into continuous vectors; Conv 1, Conv 2, Conv 3: convolution; Max Pooling: maximum pooling, taking the maximum of feature points in the neighborhood; Average Pooling: global average pooling; Concat: Concat layer realizes the splicing of input data; Class: category.

[0102] Suppose the number of input texts is A and the text length limit is B, then the input data is E text data of length F, and its dimension is [A, B].

[0103] After the title text is input into the model, the model preprocesses the input text data and maps the input text information to a high-dimensional space in the form of word vectors. After mapping, the spatial distance of text data with the same semantic information will become smaller, while the spatial distance of text data with large semantic information differences will become larger, which will facilitate subsequent feature extraction and classification. The output data dimension is [E, F, G]; G is the dimension corresponding to each character in the text after mapping.

[0104] Use multiple convolution kernels to process the text separately to extract text features.

[0105] Text data is a sequence data, and there is a dependency relationship between the previous and next texts. In order to fully extract this dependency between text data, a multi-convolution method is used to extract text features.

[0106] In this embodiment, there are 3 convolution kernels with a scale of 3*1.

[0107] The process is explained based on a 3*1 convolution kernel. The convolution kernel kernel_1 has a scale of 3*1 and contains a total of 3 elements, each of which is adjacent to each other.

[0108] The output data x1 after kernel_1 convolution has a dimension of [E, F, G];

[0109] The convolution kernel kernel_2 has a scale of 3*1 and contains 3 elements. However, each adjacent element is separated by k1 pixels, which means that the actual scale of kernel_2 should be (3+2k1)*1.

[0110] The output data x2 dimension after kernel_3 convolution is [E, F, G];

[0111] The convolution kernel kernel_3 has a scale of 3*1 and contains 3 elements. However, each adjacent element is separated by k2 pixels, which means that the actual scale of kernel_3 should be (3+2k2)*1.

[0112] The output data x3 after kernel_3 convolution has a dimension of [E, F, G].

[0113] Among them, x1, x2, and x3 refer to the output data obtained after Text is subjected to three Conv operations by kernel_1, kernel_2, and kernel_3 respectively. They are the results obtained after extracting features from Text using specially designed convolution kernels by kernel_1, kernel_2, and kernel_3 respectively.

[0114] In this extraction method, large-scale convolution kernels are used to extract the dependency between text and text far away from its vicinity, and small-scale convolution kernels are used to extract the dependency between text and text in its vicinity.

[0115] The text features extracted from x1 are calculated according to the corresponding dimension G after each character in the text is mapped, and x1 is obtained. max and x1 avg , and x1 max and x1 avg After concatenation, the final feature corresponding to kernel_1 is obtained, and its dimension is [E, F, 2];

[0116] Follow the above operations for x2 and x3 to get the maximum value x2 max , x3 max , mean x2 avg , x3 avg ; x2 max and x2 avg , x3 max and x3 avg After concatenation, the final features corresponding to kernel_2 and kernel_3 are obtained, with a dimension of [E, F, 2];

[0117] The features of all scale convolution kernels are concatenated to obtain the final features, whose dimension is [E, F, 6];

[0118] The concatenated features are input into the classifier to obtain the final classification result. The core here is to use multiple convolution kernels to extract features from the same sentence, extract the dependencies between texts as much as possible, and then process the features extracted by each scale convolution kernel separately. The model can extract the most important information in the text and fully consider all the information in the text; finally, the results of all scale convolution kernels are concatenated and classified. Through this feature extraction method, the model can simultaneously use all the information in the text and the dependency information between texts, and use this dependency information to assist classification, which further improves performance.

[0119] In this embodiment, the main improvements to the optimization process of the model are:

[0120] Using three specially designed convolution kernels, kernel_1, kernel_2, and kernel_3. These three convolution kernels have the same elements but different sizes. This design allows the model to learn the potential correlation between adjacent texts;

[0121] After the convolution is completed, the maximum and mean values ​​of x1, x2, and x3 are calculated according to the third dimension (G), that is, x1 max , x1 avg , x2 max , x2 avg , x3 max , x3 avg . The dimension of the output data x1, x2, and x3 is [E, F, G], which is three-dimensional data. The first dimension represents the number of texts, and the value here is E; the second dimension represents the number of characters in a text, and the value here is F; the third dimension represents the dimension corresponding to a single character after mapping, and the value here is G. Here, the third dimension represents the dimension corresponding to a single character after mapping. For the G elements corresponding to each character, it is necessary to calculate the maximum value and mean value of each character in the G elements of the high-dimensional space.

[0122] x1 max and x1 avg , x2 max and x2 avg , x3 max and x3 avg After splicing, the final features corresponding to kernel_1, kernel_2, and kernel_3 are obtained;

[0123] The features of all scale convolution kernels are concatenated to obtain the final feature, whose dimension is [E, F, 6].

[0124] Step S203, model training module:

[0125] The model established in step S202 is trained using the data obtained in step S201.

[0126] Step S204, information classification module:

[0127] The model trained in step S203 is used to judge the information data used for testing.

[0128] The information classification method based on multiple semantic information proposed in the present invention designs a special model for information classification; uses multiple convolution kernels to process the text separately to extract text features, which allows the model to obtain the information between texts at different scales; by processing the text features extracted by each scale convolution kernel separately, the model can use both the most important information in the text and the overall information contained in the text; finally, the features extracted by all scale convolution kernels are integrated as the final features, which finally integrates the information of the text itself and the dependency relationship between texts, and the two complement each other. Through this feature extraction method, the model can simultaneously use all the information in the text and the dependency information between texts, and use this dependency information to assist classification, which further improves the performance.

[0129] After introducing the method of the exemplary embodiment of the present invention, next, refer to Figure 4 An information classification system based on multiple semantic information according to an exemplary embodiment of the present invention is introduced.

[0130] The implementation of the information classification system based on multiple semantic information can refer to the implementation of the above method, and the repeated parts will not be repeated. The terms "module" or "unit" used below can be a combination of software and / or hardware that implements the predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0131] Based on the same inventive concept, the present invention also proposes an information classification system based on multiple semantic information, such as Figure 4 As shown, the system includes:

[0132] A preprocessing module 110, used to collect information data and preprocess the information data;

[0133] The model building module 120 is used to build an information classification model, input the preprocessed information data into the information classification model, map the information data to a high-dimensional space through the information classification model, reduce the spatial distance of the same semantic information, and expand the spatial distance of the different semantic information, and output the mapped text data;

[0134] An optimization module 130, for performing feature extraction on the mapped text data using a plurality of convolution kernels, extracting dependency information between texts, and optimizing the information classification model according to the text information of the information data and the dependency information between texts;

[0135] The training module 140 is used to set a training set and a test set according to the preprocessed information data, train the optimized information classification model, and test the training model using the test set;

[0136] The classification module 150 is used to obtain information data to be classified, and classify the information data to be classified using the trained information classification model to obtain a classification result.

[0137] In this embodiment, the preprocessing module 110 is specifically used for:

[0138] Filter and annotate the collected information data, extract the title text of the information and mark the information category.

[0139] In this embodiment, the model building module 120 is specifically used for:

[0140] Set the number of input texts to A and the text length limit to B;

[0141] According to the preprocessed information data, the data input to the model is E text data of length F, with a dimension of [A, B];

[0142] The information data is mapped to a high-dimensional space through the information classification model, and the dimension of the output mapped text data is [E, F, G]; where G is the dimension corresponding to each character in the text after mapping.

[0143] In this embodiment, the optimization module 130 is specifically used for:

[0144] A convolution kernel of a first scale is used to extract the dependency relationship between texts outside a certain area corresponding to the texts, and a convolution kernel of a second scale is used to extract the dependency relationship between texts within a certain area corresponding to the texts; wherein the first scale is larger than the second scale;

[0145] The features extracted by multiple convolution kernels are spliced ​​to obtain the final features, and the final features are input into the classifier to obtain the final classification result, and the information classification model is optimized according to the final classification result.

[0146] In this embodiment, the optimization module 130 is specifically used for:

[0147] After convolving the mapped text data with each convolution kernel, the output data dimension is [E, F, G];

[0148] Calculate the maximum and mean values ​​of the output data according to the dimension G corresponding to each character in the text after mapping, and concatenate the maximum and mean values ​​to obtain the final features corresponding to each convolution kernel, with a dimension of [E, F, 2];

[0149] The final features of all convolution kernels are concatenated to obtain the final features, with a dimension of [E, F, 2*N]; N is the number of convolution kernels;

[0150] The concatenated features are input into the classifier to obtain the final classification result.

[0151] In this embodiment, the training module 140 is specifically used for:

[0152] Use the training set to train the optimized information classification model;

[0153] The trained model is tested according to the information titles in the test set to determine whether the classification results are correct; if the accuracy reaches the preset value, the model training is completed.

[0154] It should be noted that although several modules of the information classification system based on multiple semantic information are mentioned in the above detailed description, this division is only exemplary and not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more modules described above can be concretized in one module. Conversely, the features and functions of one module described above can be further divided into multiple modules for concretization.

[0155] Based on the above invention concept, Figure 5 As shown, the present invention also proposes a computer device 500, including a memory 510, a processor 520, and a computer program 530 stored in the memory 510 and executable on the processor 520, wherein the processor 520 implements the aforementioned information classification method based on multiple semantic information when executing the computer program 530.

[0156] Based on the aforementioned inventive concept, the present invention proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the information classification method based on multiple semantic information is implemented.

[0157] The information classification method and system based on multiple semantic information proposed in the present invention extract text features through convolution kernels. The method and system are simple to use, convenient to use, and efficient, and can save a lot of time and labor costs. In the process of information classification processing, an information-enhanced data processing mechanism is adopted, which can fully extract the features in the text, improve the information utilization rate, and improve the performance of the information classification model, thereby effectively improving the accuracy of information classification.

[0158] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0159] The present invention is described with reference to flowcharts and / or block diagrams of methods and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0160] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0161] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0162] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for information classification based on multiple semantic information, characterized in that: The method includes: Collecting information data and preprocessing the information data; Establish an information classification model, input the preprocessed information data into the information classification model, map the information data to a high-dimensional space through the information classification model, reduce the spatial distance of the same semantic information, and expand the spatial distance of the different semantic information, and output the mapped text data; Using multiple convolution kernels to perform feature extraction on the mapped text data, extract the dependency information between texts, and optimize the information classification model according to the text information of the information data and the dependency information between texts; Set up training sets and test sets based on the preprocessed information data, train the optimized information classification model, and use the test set to test the training model; Obtain information data to be classified, and use the trained information classification model to classify the information data to be classified to obtain classification results; The collecting of information data and preprocessing of the information data include: Screen and annotate the collected information data, extract the title text of the information and annotate the information category; Among them, an information classification model is established, the preprocessed information data is input into the information classification model, the information data is mapped to a high-dimensional space through the information classification model, the spatial distance of the same semantic information is reduced, and the spatial distance of the different semantic information is expanded, and the mapped text data is output, including: Set the number of input texts to A and the text length limit to B; According to the preprocessed information data, the input data of the model is E text data of length F, with the dimension [A, B]; The information data is mapped to a high-dimensional space through the information classification model, and the output text data dimension after mapping is [E, F, G]; where G represents the dimension corresponding to each character in the text after mapping; Among them, using multiple convolution kernels to extract features of the mapped text data respectively, extracting the dependency information between texts, and optimizing the information classification model according to the text information of the information data and the dependency information between texts, including: A convolution kernel of a first scale is used to extract the dependency relationship between texts outside a certain area corresponding to the texts, and a convolution kernel of a second scale is used to extract the dependency relationship between texts within a certain area corresponding to the texts; wherein the first scale is larger than the second scale; The features extracted by multiple convolution kernels are concatenated to obtain final features, and the final features are input into the classifier to obtain final classification results, and the information classification model is optimized according to the final classification results; Among them, using multiple convolution kernels to extract features of the mapped text data respectively, extracting the dependency information between texts, and optimizing the information classification model according to the text information of the information data and the dependency information between texts, including: After convolving the mapped text data with each convolution kernel, the output data dimension is [E, F, G]; Calculate the maximum and mean values ​​of the output data according to the dimension G corresponding to each character in the text after mapping, and concatenate the maximum and mean values ​​to obtain the final features corresponding to each convolution kernel, with a dimension of [E, F, 2]; The final features of all convolution kernels are concatenated to obtain the final features, with a dimension of [E, F, 2*N]; N is the number of convolution kernels; The concatenated features are input into the classifier to obtain the final classification result.

2. The information classification method based on multiple semantic information according to claim 1, characterized in that: According to the preprocessed information data, a training set and a test set are set to train the optimized information classification model, and the training model is tested using the test set, including: Use the training set to train the optimized information classification model; The trained model is tested according to the information titles in the test set to determine whether the classification results are correct; if the accuracy reaches the preset value, the model training is completed.

3. An information classification system based on multiple semantic information, characterized in that: The system includes: A preprocessing module, used for collecting information data and preprocessing the information data; A model building module is used to build an information classification model, input the preprocessed information data into the information classification model, map the information data to a high-dimensional space through the information classification model, reduce the spatial distance of the same semantic information, and expand the spatial distance of the different semantic information, and output the mapped text data; An optimization module, used to use multiple convolution kernels to perform feature extraction on the mapped text data, extract dependency information between texts, and optimize the information classification model according to the text information of the information data and the dependency information between texts; The training module is used to set a training set and a test set according to the preprocessed information data, train the optimized information classification model, and test the training model using the test set; The classification module is used to obtain information data to be classified, and classify the information data to be classified using the trained information classification model to obtain the classification result; Among them, the preprocessing module is specifically used for: Screen and annotate the collected information data, extract the title text of the information and annotate the information category; Among them, the model building module is specifically used for: Set the number of input texts to A and the text length limit to B; According to the preprocessed information data, the input data of the model is E text data of length F, with the dimension [A, B]; The information data is mapped to a high-dimensional space through the information classification model, and the dimension of the output mapped text data is [E, F, G]; where G is the dimension corresponding to each character in the text after mapping; The optimization module is specifically used for: A convolution kernel of a first scale is used to extract the dependency relationship between texts outside a certain area corresponding to the texts, and a convolution kernel of a second scale is used to extract the dependency relationship between texts within a certain area corresponding to the texts; wherein the first scale is larger than the second scale; The features extracted by multiple convolution kernels are concatenated to obtain final features, and the final features are input into the classifier to obtain final classification results, and the information classification model is optimized according to the final classification results; The optimization module is specifically used for: After convolving the mapped text data with each convolution kernel, the output data dimension is [E, F, G]; Calculate the maximum and mean values ​​of the output data according to the corresponding dimensions of each character in the text after mapping, and concatenate the maximum and mean values ​​to obtain the final features corresponding to each convolution kernel, with dimensions of [E, F, 2]; The final features of all convolution kernels are concatenated to obtain the final features, with a dimension of [E, F, 2*N]; N is the number of convolution kernels; The concatenated features are input into the classifier to obtain the final classification result.

4. The information classification system based on multiple semantic information according to claim 3, characterized in that: The training module is specifically used for: Use the training set to train the optimized information classification model; The trained model is tested according to the information titles in the test set to determine whether the classification results are correct; if the accuracy reaches the preset value, the model training is completed.

5. A computer device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method according to claim 1 or 2 is implemented.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to claim 1 or 2 is implemented.

Images