A text analysis method and device based on big data

Abstract

The application discloses a text analysis method and device based on big data, collects text big data, and obtains target text by partially annotating the text big data; data preprocessing is performed on the target text, the data preprocessing comprises data cleaning, data transformation and data reduction, and model data is obtained; the model data is divided into a training set, a test set and a verification set, and a deep belief network model is constructed by using the training set; the deep belief network model is verified according to the verification set, and is used for analyzing the text big data. The deep belief network model learns the inherent characteristics of data, obtains initial weights close to global optimization, effectively avoids the problem that a network model converges to local optimization due to random initialization of weight parameters, and makes the model show excellent performance in classification and regression.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data processing, and particularly relates to a text analysis method and device based on big data. BACKGROUND

[0002] In the current Internet information industry, unstructured data accounts for more than 70% of all data, more than half of which is text data. These text data may include web pages, user documents, real-time network information, news, micro blogs, etc. How to efficiently utilize these resources is one of the common technical difficulties in the field of Internet data mining.

[0003] Text similarity analysis is a common technique included in text data mining, and its purpose is to analyze and compare different long and short texts through algorithms to calculate the similarity of the same text with other text data. This technology can be used in information recommendation, similar article deduplication, library retrieval and other Internet application scenarios. Text data in the Internet industry is generally obtained from user manual upload, periodic collection by web crawlers, etc., and often has the characteristics of large data volume, complex format, more redundant content and frequent updates.

[0004] Deep learning, as a new deep neural network learning algorithm, simulates the working mode of the human brain, constructs an artificial neural network model containing multiple hidden layers, and extracts features from high-dimensional data layer by layer to form abstract high-level representations, showing strong learning ability. Deep belief network, as an excellent deep learning model, has strong feature extraction capability. It is a deep probabilistic generative model that uses layer-by-layer unsupervised training to solve the problem of multi-hidden layer neural network training difficulty, and shows excellent performance in prediction and classification. It has been successfully applied to many fields. The method is based on the structure of deep belief network model, and proposes a method for intelligent evaluation and extraction of key content of text. SUMMARY

[0005] The present application provides a text analysis method based on big data, which aims to solve the above problems.

[0006] To achieve the above purpose, the present application adopts the following technical scheme:

[0007] The text analysis method based on big data provided by the present application comprises the following steps:

[0008] Collecting text big data, and annotating part of the text big data to obtain target text;

[0009] Data preprocessing is performed on the target text, including data cleaning, data transformation and data reduction, to obtain model data;

[0010] The model data is divided into a training set, a test set and a validation set, and a deep belief network model is constructed using the training set;

[0011] The deep belief network model is verified according to the validation set, and is used to analyze the text big data.

[0012] Preferably, the data preprocessing step includes:

[0013] Step 1: Establish a target function:

[0014] Where ζ(x, θ) represents the optimization goal of training the neural network, θ represents the network parameters, x represents a single data sample, and X represents a data set of N data points, {x i} N i＝1 ;

[0015] Step 2: Further optimize formula (1) by first-order optimization, where the parameter θ on the target data set B is decreased along the negative gradientт k Whereт k represents the step size of the descending gradient, k is an integer greater than 1, and the formula is represented as:

[0016]

[0017] Step 3: Remove the influence of small batch random gradient descent on the sampled data points to eliminate the error caused by random sampling data and improve the accuracy of the model. The update equation is:

[0018]

[0019] Preferably, the step of constructing the deep belief network model includes:

[0020] (1) Starting from the input layer, each adjacent two layers of nodes form an RBM; first, train the first layer of RBM using the CD algorithm to obtain the parameters of the first layer of RBM; then, fix the parameters of the RBM, take the output of the current RBM as the input of the next RBM, and continue to train the next layer of RBM until all RBMs are trained, to obtain the initial parameters of the deep belief network;

[0021] (2) using the initial parameters obtained in step (1) to construct an artificial neural network model of the same structure, and then using labeled data to fine-tune the entire model in reverse, to obtain final parameters of the deep belief network model, wherein the labeled data represents pre-set data for penetrating layer by layer during model training to obtain a deep belief network model meeting certain requirements.

[0022] Preferably, the partial annotation of the text big data includes manual annotation or extraction of keywords from the text big data using a text keyword extraction algorithm, and filtering of incorrect keywords including stop words.

[0023] A text analysis device based on big data, comprising:

[0024] A collection module for collecting text big data and performing partial annotation on the text big data to obtain target text;

[0025] A data preprocessing module for preprocessing the target text, wherein the data preprocessing includes data cleaning, data transformation, and data reduction to obtain model data;

[0026] A training module for dividing the model data into a training set, a test set, and a validation set, and constructing a deep belief network model using the training set;

[0027] An analysis module for verifying the deep belief network model according to the validation set to analyze the text big data.

[0028] Preferably, the data preprocessing includes:

[0029] Step 1: Establish a target function:

[0030] wherein ζ(x, θ) represents the optimization target of the trained neural network, θ represents the network parameters, x represents a single data sample, and X = {x i} N i＝1 ;

[0031] Step 2: Further optimize formula (1) through first-order optimization, wherein the parameters θ on the target data set B are decreased along the negative gradient k , wherein k represents the step size of the descending gradient, k is an integer greater than 1, and the formula is represented as:

[0032]

[0033] Step three: remove the influence of small batch stochastic gradient descent corresponding to the sampling data points, to eliminate the error caused by random sampling data, improve the precision of the model, and the update equation is:

[0034]

[0035] As preferred, the training module comprises:

[0036] (1) from the input layer, every two adjacent nodes form an RBM; first, the first layer RBM is trained by using the CD algorithm to obtain the parameters of the first layer RBM; then, the parameters of the RBM are fixed, the output of the current RBM is taken as the input of the next RBM, and the next layer RBM is continuously trained until all RBMs are trained to obtain the initial parameters of the deep belief network;

[0037] (2) an artificial neural network model with the same structure is constructed by using the initial parameters obtained in step (1), and then the entire model is fine-tuned in reverse by using labeled data, to obtain the final parameters of the deep belief network model, wherein the labeled data represents pre-set data, and is used to penetrate layer by layer during training of the model to obtain a deep belief network model meeting certain requirements.

[0038] As preferred, the collection module comprises manual annotation or using a text keyword extraction algorithm to extract keywords from the text big data, and filtering error keywords including stop words.

[0039] An electronic device comprising a memory and a processor, the memory being configured to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the text analysis method based on big data according to any one of the above.

[0040] A computer readable storage medium storing a computer program, wherein the computer program enables a computer to implement the text analysis method based on big data according to any one of the above.

[0041] The present application has the following beneficial effects:

[0042] Data preprocessing, on the one hand, improves the quality of data, and on the other hand, makes the data better adapt to the construction of the deep belief network model. Then, through the layer-by-layer pre-training of the RBM, the deep belief network model learns the internal characteristics of the data, obtains the initial weight close to the global optimum, effectively avoids the problem that the network model converges to the local optimum due to the random initialization of the weight parameter, and makes the model show excellent performance in classification and regression.

[0043] The preprocessed data is divided into training, testing, and validation sets. A deep belief network model is then constructed, and the trained model is applied to analyze the collected large amount of text data, improving the efficiency of users in obtaining key content. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating a text analysis method based on big data, as implemented in an embodiment of the present invention.

[0045] Figure 2 This is a schematic diagram illustrating the training of a deep belief network model in a big data-based text analysis method according to an embodiment of the present invention.

[0046] Figure 3 This is a structural diagram of a big data-based text analysis system implemented according to an embodiment of the present invention;

[0047] Figure 4 This is a schematic diagram of an electronic device that implements a big data-based text analysis system according to an embodiment of the present invention. Detailed Implementation

[0048] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0049] The terms “first,” “second,” etc., used in the claims and description of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate. This is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion so that a process, method, system, product, or apparatus that comprises a series of units is not necessarily limited to those units, but may include other units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein in the specification of this application is for the purpose of describing particular embodiments only and is not intended to limit this application.

[0051] Example 1, as Figures 1-2 As shown, a text analysis method based on big data includes the following steps:

[0052] S110, collect text big data, and partially annotate the text big data to obtain a target text;

[0053] S120, data preprocessing is performed on the target text, the data preprocessing including data cleaning, data transformation and data reduction, to obtain model data;

[0054] S130, the model data is divided into a training set, a test set and a validation set, and a deep belief network model is constructed using the training set;

[0055] S140, a deep belief network model is obtained by verifying the validation set, and used to analyze the text big data.

[0056] As can be seen from Example 1, the data preprocessing stage involves data cleaning, data transformation and data reduction, etc. The data obtained by data collection will have a large amount of incomplete, inconsistent and abnormal data, which will seriously affect the execution efficiency of the later modeling, and even may cause the deviation of the evaluation result, so data preprocessing is needed. On the one hand, the quality of the data is improved, and on the other hand, the data is better adapted to the construction of the deep belief network model. Data cleaning "cleans" the data by filling in the missing values, smoothing the noise data, identifying or deleting outliers and solving inconsistencies. The main goal is to achieve the following targets: format standardization, abnormal data removal, error correction, and duplicate data removal. Data transformation converts the data into a form suitable for subsequent training and learning by smoothing aggregation, data generalization, normalization, etc. The amount of data obtained by data collection will be relatively large, which is not conducive to the later processing. Data reduction techniques can be used to obtain a reduced representation of the data set, which is much smaller, but still close to maintaining the integrity of the original data, and will not cause a large difference between the evaluation results before and after reduction.

[0057] Before the emergence of deep learning methods, the academic and industrial communities applied shallow artificial neural networks more, not because the performance of shallow artificial neural networks is better than that of deep artificial neural networks, but because deep neural networks are difficult to train and are prone to converge to local minimum, resulting in results worse than shallow models, which cannot meet the application requirements. As a probabilistic generative model, the deep belief network is stacked by multiple restricted Boltzmann machines (RBM) in sequence, and its structure is shown in Figure 2 As shown in Fig. 1, the deep belief network has a strong ability to fit complex data, and can process various complex types of data such as speech, image, text and numerical value by extracting data features layer by layer through multiple restricted Boltzmann machines, and is widely used in various fields. As a deep neural network model, the deep belief network has a unique training process that overcomes the difficulty of training multi-hidden layer neural networks. Its training process is divided into two steps:

[0058] (1) From the input layer, every two adjacent layers of nodes form an RBM; first, the first layer of RBM is trained by using the CD algorithm to obtain the parameters of the first layer of RBM; then, the parameters of the RBM are fixed, the output of the current RBM is taken as the input of the next RBM, and the next layer of RBM is continuously trained until all RBMs are trained to obtain the initial parameters of the network.

[0059] (2) An artificial neural network model with the same structure is constructed by using the weight parameters obtained in step (1), and then the entire model is fine-tuned in reverse by using the labeled data to obtain the final parameters of the model.

[0060] Through the layer-by-layer pre-training of the RBM in step (1), the deep belief network model learns the intrinsic characteristics of the data, obtains the initial weight close to the global optimum, effectively avoids the problem that the network model converges to the local optimum due to the random initialization of the weight parameters, and makes the model perform excellent performance in both classification and regression.

[0061] The preprocessed relevant data is divided into a training set, a test set and a validation set, and then a deep belief network model is constructed, and the trained model meeting the requirements is applied to text big data analysis.

[0062] In the construction of the deep belief network model, the big data text is obtained by random sampling, and there is an influence that causes the data to deviate, which leads to the generalization of the deep belief network, so it is hoped to separate the influence of the random variable, and the data preprocessing specifically includes the following steps:

[0063] Step one: establish the objective function:

[0064] Wherein, ζ(x, θ) represents the optimization objective of training the neural network, θ represents the network parameter, x represents a single data sample, and x represents the data set X of N data points, {x i} N i＝1 ;

[0065] Step two: formula (1) is optimized by first-order optimization, wherein, as the parameter θ on the target data set B is decreased along the negative gradient, the step size of the gradient is decreased, and k is an integer greater than 1, and the formula is represented as: k k

[0066]

[0067] ​​Step three: remove the impact of small batch stochastic gradient descent corresponding to the sampling data points, to eliminate the error caused by random sampling data, improve the precision of the model, the update equation is:

[0068]

[0069] In formula (3), the complete batch gradient descent corresponds to the descent on the complete data set B=X, the stochastic gradient descent corresponds to the sampling of a single random data point B={x}, and the small batch stochastic gradient descent corresponds to the sampling data point B={x j} S j＝1 , x j～ X, when sampling without replacement, usually when all elements are exhausted, the update equation is usually analyzed as the update of the contaminated complete batch gradient, and the gradient noise generated by random small batch sampling.

[0070] The embodiment first performs preliminary content analysis on the collected text big data, the content analysis including same semantic analysis on the target text which has been annotated, the annotated content is not limited to atomic segmentation, recognition of unknown words, and part-of-speech annotation, and a semantic model conforming to formula (1) is established, and the specific entity information possessed by the same semantic analysis, including conditions, variables and targets, is specifically assigned, secondly, based on the requirement of formula (2), the part which cannot be described as semantics is iteratively calculated and optimized, then, based on formula (3), the interference factors such as redundant part, invalid condition, overlapping target and contradictory condition contained in the text big data are eliminated, so that the complexity is reduced, which is conducive to the computer to clearly understand the real query target of the instruction, after accurately understanding the semantic of the instruction, the work flow of the semantic analysis is determined, so that it accurately conforms to the input, target, constraint and the like of the instruction, and becomes an executable work flow. Starting from formula (1)-(3), the semantic association and dependence of annotated text to the whole text big data are utilized to select the corresponding candidate work flow instance set; finally, the application identifies various potential risks and threats occurring in the work flow link through the query case knowledge of the past work flow, and verifies the reliability of the work flow.

[0071] Embodiment 3, a specific implementation, includes:

[0072] S310, collect text big data, and partially annotate the text big data to obtain target text, for example:

[0073] The extracted part of the big data text is "The recording process of holography: divide the laser beam into two beams, one laser beam is directly projected on the holographic plate, called the reference beam; the other laser beam is projected on the object, and carries the light wave information of the object through reflection or transmission of the object, called the object beam. The object beam is also projected on the same area of the holographic plate and interferes with the reference beam. On the holographic plate, the object beam and the reference beam are coherently superimposed to form interference fringes, and a hologram is completed. The reproduction of the object image in the hologram: a laser beam is used to irradiate the hologram, and the frequency and transmission direction of the laser beam should be exactly the same as those of the reference beam. When the hologram is irradiated by the light wave, the original object light wave can be reproduced due to the principle of diffraction, thereby forming a realistic three-dimensional image of the original object. People can see different sides of the object from different angles, just like seeing the real object, but cannot touch the real object". The essence is interference recording and diffraction reproduction.

[0074] The comment setting interval symbol is: divide the laser beam into two beams, one laser beam is directly projected on the holographic plate, called the reference beam; the other laser beam is projected on the object, and carries the light wave information of the object through reflection or transmission of the object, called the object beam. The object beam is also projected on the same area of the holographic plate and interferes with the reference beam. On the holographic plate, the object beam and the reference beam are coherently superimposed to form interference fringes, and a hologram is completed. The reproduction of the object image in the hologram: a laser beam is used to irradiate the hologram, and the frequency and transmission direction of the laser beam should be exactly the same as those of the reference beam. When the hologram is irradiated by the light wave, the original object light wave can be reproduced due to the principle of diffraction, thereby forming a realistic three-dimensional image of the original object. People can see different sides of the object from different angles, just like seeing the real object, but cannot touch the real object".

[0075] S320, data preprocessing is performed on the target text, the data preprocessing includes data cleaning, data transformation and data reduction, and model data is obtained;

[0076] Words with frequencies below the average are replaced with exclamation marks (!). The word with the highest frequency is then identified as "hologram". "Hologram" is removed, and the most frequent word is counted again to identify "laser". This process is repeated with a threshold set to obtain words such as "interference", "object beam", "reference beam", "diffraction", "reproduction", "of", "at", and "on". An objective function is established for each target word. For example, synonyms for "laser" include "laser beam", "laser beam", "pulse laser", and "infrared laser". Optimization is then performed to remove pause words like "of", "at", and "on", as well as random errors from synonyms. For instance, "laser beam" and "laser beam" can cause random interference in "object beam".

[0077] S330. Divide the model data into a training set, a test set, and a validation set, and use the training set to construct a deep belief network model;

[0078] The model data obtained in step S320 is divided into training, testing, and validation sets to construct a deep belief network model. Finally, in step S340, the target text result is obtained as "laser interference hologram, diffraction reconstruction" or "object beam and reference beam interference, diffraction reconstruction". This means that the result of the target text is an interference record and diffraction reconstruction.

[0079] S340. A deep belief network model is obtained by verifying the validation set, and used to analyze the text big data.

[0080] Example 4, as Figure 3 As shown, a text analysis device based on big data includes:

[0081] Acquisition module 10: Used to acquire large amounts of text data and to annotate some of the large amounts of text data to obtain the target text;

[0082] Data preprocessing module 20: used to preprocess the target text, the data preprocessing including data cleaning, data transformation and data reduction, to obtain model data;

[0083] Training module 30: used to divide the model data into training set, test set and validation set, and to use the training set to construct a deep belief network model;

[0084] Analysis module 40: used to verify the deep belief network model based on the validation set, and to analyze the text big data.

[0085] Example 5

[0086] like Figure 4As shown, an electronic device includes a memory 401 and a processor 402, the memory 401 is configured to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor 402 to implement the above-described big data-based text analysis method.

[0087] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the electronic device described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0088] A computer readable storage medium storing a computer program, the computer program enables a computer to execute the above-described big data-based text analysis method when executed.

[0089] Exemplarily, the computer program can be divided into one or more modules / units, one or more modules / units are stored in the memory 401 and executed by the processor 402, and the I / O interface transmission of data is completed by the input interface 405 and the output interface 406, to complete the present application, one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which is used to describe the execution process of the computer program in the computer device.

[0090] The computer device can be a desktop computer, a notebook computer, a palm computer, and a cloud server, etc. The computer device can include, but is not limited to, the memory 401, the processor 402, and those skilled in the art can understand that the embodiment is only an example of the computer device, and does not constitute a limitation on the computer device, and can include more or fewer components, or combine certain components, or different components, for example, the computer device can also include an inputter 407, a network access device, a bus, etc.

[0091] The processor 402 can be a central processing unit (CPU), and can also be other general-purpose processors 402, digital signal processors 402 (DSP), application-specific integrated circuits (ASIC), ready programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor 402 can be a microprocessor or the processor 402 can also be any conventional processor 402, etc.

[0092] The storage 401 can be an internal storage unit of the computer device, such as a hard disk or a memory of the computer device. The storage 401 can also be an external storage device of the computer device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like. Further, the storage 401 can include both an internal storage unit and an external storage device. The storage 401 is used to store computer programs and other programs and data required by the computer device. The storage 401 can also be used to temporarily store data to be output by the outputter 408. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM) 403, a random access memory (RAM) 404, a disk or a disc, and the like.

[0093] The above merely provides the specific embodiments of the present application, but the technical features of the present application are not limited to the above. Any changes or modifications made by those skilled in the art within the scope of the present application shall fall within the patent scope of the present application.

Claims

1. A text analysis method based on big data, characterized in that, include: Collect large amounts of text data and annotate some of it to obtain the target text; The target text is preprocessed, including data cleaning, data transformation, and data reduction, to obtain model data. The preprocessing steps include: Step 1: Establish the objective function Where ζ(x, θ) represents the optimization objective of training the neural network, θ represents the network parameters, and x represents a single data sample in a dataset X with N data points, {x i } Ni＝1 middle; Step 2: Optimize formula (1) using first-order optimization, where the parameter θ on the target dataset B decreases along the negative gradient т. k , where т k Let k represent the step size of the descent gradient, where k is an integer greater than 1. The formula is as follows: Step 3: Remove the influence of mini-batch stochastic gradient descent sampling data points to eliminate errors caused by random sampling, improve model accuracy, and update the equation as follows: The model data is divided into a training set, a test set, and a validation set, and a deep belief network model is constructed using the training set. The deep belief network model is obtained by validating the validation set and used to analyze the text big data.

2. The text analysis method based on big data according to claim 1, characterized in that, The construction depth The steps of a belief network model include: (1) Starting from the input layer, every two adjacent layers of nodes form an RBM; first, the first layer of the RBM is trained using the CD algorithm, resulting in... Get the parameters of the first-level RBM; then fix the parameters of the RBM, use the output of the current RBM as the input of the next RBM, and continue. Train the next layer RBM until all RBMs are trained, and obtain the initial parameters of the deep belief network; (2) Construct an artificial neural network model with the same structure using the initial parameters obtained in step (1), and then use labeled data to fine-tune the model in reverse to obtain the final parameters of the deep belief network model. The labeled data This refers to pre-defined data used to progressively penetrate the training model to obtain a deep belief network model that meets certain requirements. type.

3. The text analysis method based on big data according to claim 1, characterized in that, The text This large dataset has been partially annotated, including manual annotations and the use of text keyword extraction algorithms to extract key information from the text. Extract keywords and filter out incorrect keywords, including stop words.

4. A text analysis device based on big data, characterized in that, include: Acquisition module: Used to collect large amounts of text data and to annotate some of the text data to obtain the target text; Data preprocessing module: used to preprocess the target text, the data preprocessing includes data cleaning, data transformation and data reduction to obtain model data; Training module: used to divide the model data into training set, test set and validation set, and to construct a deep belief network model using the training set. The data preprocessing steps include: Step 1: Establish the objective function Where ζ(x, θ) represents the optimization objective of training the neural network, θ represents the network parameters, and x represents a single data sample in a dataset X with N data points, {x i } Ni＝1 middle; Step 2: Optimize formula (1) using first-order optimization, where the parameter θ on the target dataset B decreases along the negative gradient т. k , where т k Let k represent the step size of the descent gradient, where k is an integer greater than 1. The formula is as follows: Step 3: Remove the influence of mini-batch stochastic gradient descent sampling data points to eliminate errors caused by random sampling, improve model accuracy, and update the equation as follows: The model data is divided into a training set, a test set, and a validation set, and a deep belief network model is constructed using the training set. Analysis module: used to verify the deep belief network model based on the validation set, and then to analyze the text big data.

5. A text analysis device based on big data according to claim 4, characterized in that, The training module includes: (1) Starting from the input layer, every two adjacent layers of nodes form an RBM; first, the first layer of the RBM is trained using the CD algorithm, resulting in... Get the parameters of the first-level RBM; then fix the parameters of the RBM, use the output of the current RBM as the input of the next RBM, and continue. Train the next layer RBM until all RBMs are trained, and obtain the initial parameters of the deep belief network; (2) Construct an artificial neural network model with the same structure using the initial parameters obtained in step (1), and then use labeled data to fine-tune the model in reverse to obtain the final parameters of the deep belief network model. The labeled data This refers to pre-defined data used to progressively penetrate the training model to obtain a deep belief network model that meets certain requirements. type.

6. A text analysis device based on big data according to claim 4, characterized in that, The acquisition module This includes extracting keywords from the large amount of text data using manual annotation or text keyword extraction algorithms, and filtering out keywords including... Incorrect keywords, including stop words.

7. An electronic device, characterized in that, It includes a memory and a processor, the memory being used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement a big data-based text analysis method as described in any one of claims 1 to 3.

8. A computer-readable storage medium storing a computer program, characterized in that, The computer program enables the computer to implement a big data-based text analysis method as described in any one of claims 1 to 3 when executed.

Images