A text classification method, system, device and storage medium

Through the multi-layer semantic vector model processing and classification marking module, the text classification accuracy and recall rate in natural language processing are improved, and the problems of large workload and low accuracy of text classification in the prior art are solved.

CN111858923BActive Publication Date: 2024-11-19BEIJING DIDI INFINITY TECH & DEV CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN201911352244.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2019-12-24
Publication Date
2024-11-19
Estimated Expiration
2039-12-24

AI Technical Summary

Technical Problem

In the field of natural language processing, especially in semantic analysis and emotion classification tasks, the classification of massive texts is large and the accuracy is difficult to guarantee. The existing dictionary-based methods have low model recall due to the low dictionary coverage.

Method used

By obtaining the original text, using multiple processing models for vectorization, semantic vectors are determined, and the classification mark of text is determined through the classification model, and a multi-layer semantic vector model is used to improve classification accuracy.

Benefits of technology

It improves the accuracy and recall rate of text classification, solves the problem of insufficient classification training samples, and enhances the applicability and classification effect of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111858923B_ABST
    Figure CN111858923B_ABST
Patent Text Reader

Abstract

The embodiments of this specification disclose a text classification method, system, device and storage medium. The text classification method includes: obtaining a first original text; vectorizing the first original text through a first processing model to determine a first semantic vector of the first original text; determining a second semantic vector through a second processing model; the input of the second processing model includes the first semantic vector; determining one or more classification tags of the first original text through a third processing model; the input of the third processing model includes the second semantic vector. The method provided by the embodiments of this specification can classify text more effectively and improve the accuracy of classification.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of this specification relate to the field of natural language processing, and in particular to a text classification method, system, device and storage medium. Background Art

[0002] Natural Language Processing (NLP) is the basis for studying language issues in human-computer interaction. It integrates linguistics, computer science, mathematics and other disciplines, and is also one of the core topics of artificial intelligence. In specific natural language processing tasks such as semantic analysis and sentiment classification, it is often necessary to classify massive amounts of input text first. Classifying these texts manually is a huge workload, and the accuracy is difficult to guarantee. Dictionary-based text classification uses authoritative dictionaries and manually constructs features based on experience. The model has a high accuracy rate, but due to the low dictionary coverage, the model recall rate is low.

[0003] Therefore, it is desirable to provide a more efficient solution for classifying text. Summary of the invention

[0004] One aspect of an embodiment of the present specification provides a method for classifying text. The method includes: obtaining a first original text; vectorizing the first original text through a first processing model to determine a first semantic vector of the first original text; determining a second semantic vector through a second processing model; the input of the second processing model includes the first semantic vector; determining one or more classification tags of the first original text through a third processing model; the input of the third processing model includes the second semantic vector.

[0005] Another aspect of the embodiments of the present specification provides a system for classifying text. The system includes: a data acquisition module for acquiring a first original text; a semantic vector module for vectorizing the first original text through a first processing model to determine a first semantic vector of the first original text; and for determining a second semantic vector through a second processing model; the input of the second processing model includes the first semantic vector; a classification labeling module for determining one or more classification labels of the first original text through a third processing model; the input of the third processing model includes the second semantic vector.

[0006] Another aspect of the embodiments of this specification provides a device for classifying texts, wherein the device includes a processor and a memory; the memory is used to store instructions, wherein when the instructions are executed by the processor, the device implements operations corresponding to the method for classifying texts.

[0007] Another aspect of the embodiments of this specification provides a computer-readable storage medium, wherein the storage medium stores computer instructions, and when a computer reads the computer instructions in the storage medium, the computer executes a method for classifying text. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 An exemplary flow chart of a method for classifying text according to some embodiments of the present specification.

[0009] Figure 2 An exemplary flowchart of a method for training a second processing model according to some embodiments of the present specification.

[0010] Figure 3 The present invention is an exemplary flow chart of a method for classifying vehicle evaluation texts by combining the characteristics of the corresponding evaluator and the travel characteristics of the corresponding vehicle according to some embodiments of the present specification.

[0011] Figure 4 This is an exemplary flow chart of a method for classifying vehicle evaluation text in combination with historical evaluation classification according to some embodiments of this specification.

[0012] Figure 5 This is an exemplary flow chart of a method for classifying text in combination with a third semantic vector according to some embodiments of this specification. DETAILED DESCRIPTION

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0014] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0015] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprises" and "includes" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0016] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, the various steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0017] Natural Language Processing (NLP) is the basis for studying language issues in human-computer interaction. It integrates linguistics, computer science, mathematics and other disciplines, and is also one of the core topics of artificial intelligence. In specific natural language processing tasks such as semantic analysis and sentiment classification, it is often necessary to classify a large amount of input text first. However, in some classification tasks, there are many categories corresponding to the text, and the number of texts corresponding to some categories is small. If the machine learning method cannot obtain enough training samples, it will be difficult to achieve effective classification.

[0018] This specification provides a method for more efficient text classification.

[0019] First, the application scenario of this specification is described.

[0020] In business applications, there are a lot of texts that need to be classified. Through a computing system, the classification can be automatically performed and the classification results can be obtained.

[0021] A computing system refers to a system with computing capabilities, which can include various computers, such as servers and personal computers, or a computing platform composed of multiple computers connected in various structures.

[0022] A computing system may include a processor that can execute program instructions. The processor may include various common general-purpose central processing units (CPUs), graphics processing units (GPUs), microprocessors, application-specific integrated circuits (ASICs), or other types of integrated circuits.

[0023] The computing system may include a storage medium, which may store instructions or data. The storage medium may include a mass storage device, a removable storage device, a volatile read-write memory, a read-only memory (ROM), etc., or any combination thereof.

[0024] The computing system may also include a network for internal connection and external connection. It may also include a terminal for input or output. The network may be any one or more of a wired network or a wireless network. The terminal may include various devices with information receiving and / or sending functions, such as computers, mobile phones, text scanning devices, display devices, printers, etc.

[0025] The computing system can obtain the text to be classified from the storage medium or from the outside through the network, and the processor executes the instructions to classify the text. The classification results can be output through the terminal, saved to the storage medium, output to the outside through the network or various interfaces, and can also be directly applied by the processor executing subsequent instructions.

[0026] For example, several servers of the online car-hailing platform can obtain the car evaluation text obtained by the car evaluation system, execute a program to classify it, and save the classification results in a database.

[0027] In the above computing system, the program instructions and / or data used may be generated through other processes, such as the training process of a machine learning model. These training processes may be performed in the above computing system or in other systems, and the instructions and / or data may be migrated to the above computing system.

[0028] For example, when classifying a car review text, the executed program includes a machine learning model. The machine learning model used may be trained in another server and then migrated to the server performing the classification.

[0029] The system described in this specification may be composed of a data acquisition module, a data output module and several data processing modules. These data processing modules may include a semantic vector module, a classification labeling module and a training module. The above modules are all executed in the computing system introduced in the application scenario, and each module includes its own instructions, which may be stored on a storage medium and executed in a processor. Different modules may be located on the same device or on different devices. Data may be transmitted between them through program interfaces, networks, etc., and data may be read from a storage device or written to a storage device.

[0030] The data acquisition module can be used to acquire the first original text. The first original text refers to the text that needs to be classified. In some embodiments, the first original text can be acquired from a database or other storage device, or through other interfaces. For details, see Figure 1 Step 110.

[0031] In some embodiments, the data acquisition module can also be used to obtain the vehicle evaluation text as the original text and obtain other vehicle evaluation original data. In some embodiments, the other vehicle evaluation original data includes the characteristics of the corresponding evaluator. The corresponding evaluator is the user who makes the vehicle evaluation, and the evaluation characteristics of the corresponding person refer to the characteristic vector of at least one or more attributes of the corresponding evaluator, which can be obtained from the database of the corresponding evaluator's registration information. In this embodiment, the input of the third processing model includes the characteristics of the evaluator. For details, see Figure 3 Step 340. In some embodiments, the other vehicle evaluation raw data includes travel characteristics of the corresponding vehicle. The corresponding vehicle is the vehicle corresponding to the vehicle evaluation. The travel characteristics of the corresponding vehicle refer to the characteristic vector of at least one or more attributes of the travel situation. Specifically, the travel characteristics may include each travel time period, each travel area and each travel mileage length. In this embodiment, the input of the third processing model includes travel characteristics. For details, see Figure 3 Step 350. In some embodiments, the other vehicle evaluation raw data includes historical evaluation classifications of the corresponding evaluator and historical evaluation classifications of the corresponding vehicle. The historical evaluation classifications of the corresponding evaluator refer to one or more classifications obtained by processing the content of one or more evaluations made by the corresponding evaluator. The historical evaluation classifications of the corresponding vehicle refer to the evaluation classifications that have been made for the vehicle. In some embodiments, the historical evaluation classifications are processed by the fifth processing model to obtain a historical evaluation vector. The fifth processing model can be an RNN model trained with data. In this embodiment, the input of the third processing model includes the historical evaluation vector. For details, see Figure 4 Step 440.

[0032] The semantic vector module can be used to vectorize the first original text through the first processing model to determine the first semantic vector of the first original text. In some embodiments, the first processing model can be a Word2Vec model, an ElMo model, a bag-of-words model, etc. The vectorization process is to input the text into the first processing model and output a corresponding vector or a group of vectors. For details, see Figure 1 Step 120.

[0033] The semantic vector module can also be used to determine a second semantic vector through a second processing model, the input of which includes the first semantic vector. The second processing model is a vector-to-vector conversion model, and the obtained vector represents the classification significance of the corresponding content to a greater extent. Figure 1 Step 130.

[0034] In some embodiments, the semantic vector module can also be used to perform vectorization processing on the first original text through a sixth processing model to determine a third semantic vector of the first original text. The sixth processing model is a vectorization processing model different from the first processing model. In some embodiments, the sixth processing model is an ELMO model. For details, see Figure 5 Step 540.

[0035] In some embodiments, the semantic vector module may also be used to determine a fourth semantic vector through a second processing model; the input of the second processing model includes the third semantic vector. Figure 5 Step 550.

[0036] The classification marking module can be used to determine one or more classification markings of the first original text through a third processing model; the input of the third processing model includes the second semantic vector. The third processing model is a classification model. The output of the third processing model is the evaluation value of the corresponding one or more classifications, and the one or more classification markings of the first original text are determined according to the evaluation value of the one or more classifications. For details, see Figure 1 Step 140.

[0037] In some embodiments, the classification tag module may also be used to determine one or more classification tags of the first original text through a third processing model; the input of the third processing model includes the fourth semantic vector. Figure 5 Step 560.

[0038] The training module can be used to determine the first training data based on the second original text. In the first training data, each sample includes a pair of third semantic vectors and a similar identifier; the third semantic vector is obtained by vectorizing the second original text through the first processing model; the similar identifier indicates whether the second original text corresponding to the pair of third semantic vectors belongs to the same category; each sample of the first training data is input into the fourth processing model; the fourth processing model includes two second processing models using the same parameters; the pair of third semantic vectors are respectively used as inputs of the two second processing models; the fourth processing model determines the loss function based on the output vectors of the two second processing models and the similar identifier in the sample; the fourth processing model determines the parameters of the included second processing model through training based on the loss function. For details, please refer to step 230 of step 2.

[0039] Figure 1 FIG. 1 is a flowchart of an exemplary text classification method according to some embodiments of this specification. Figure 1As shown, the text classification method 100 may include the following steps:

[0040] Step 110, obtaining a first original text. Specifically, step 110 may be performed by a data obtaining module.

[0041] In step 110, the first original text may be obtained from a database or other storage device, or through other interfaces.

[0042] The first original text refers to the text that needs to be classified, for example, it can be a user's evaluation text on the use of online car-hailing, or it can be the abstract of a patent specification, etc.

[0043] The following are examples of user evaluation texts on online ride-hailing services: “I feel safe and reliable,” “My family is very satisfied,” “The driver has a bad attitude,” “Average,” and “Okay.”

[0044] Step 120, performing vectorization processing on the first original text through a first processing model to determine a first semantic vector of the first original text. Specifically, step 120 can be performed by a semantic vector module.

[0045] In step 120, the first original text is vectorized by a first processing model to obtain a vector. In this specification, the obtained vector is referred to as a first semantic vector.

[0046] Vectorization processing is to input the text into the first processing model and output a corresponding vector or a set of vectors. The output vector is a digital representation of the text semantics. In some embodiments, the first processing model can be a Word2Vec model, an ElMo model, a bag-of-words model, etc., which is not limited in this specification.

[0047] Preferably, in some embodiments, the first processing model can use the BERT model. The essence of BERT is to learn a good feature representation for text by running a self-supervised learning method based on a massive corpus. We can directly use the feature representation generated by the BERT model as an embedded feature for a task. Therefore, BERT provides a model for transfer learning of other tasks, which can be directly called and used.

[0048] For example, the first semantic vector of “I feel safe and reliable” obtained by the BERT model in step 120 is: .

[0049] Step 130, determining a second semantic vector through a second processing model; the input of the second processing model includes the first semantic vector. Specifically, step 130 can be performed by a semantic vector module.

[0050] The second processing model is a vector-to-vector conversion model, and the obtained vector represents the classification significance of the corresponding content to a greater extent.

[0051] The input of the second processing model is the first semantic vector determined in step 120. The output of the second processing model is called the second semantic vector.

[0052] In some embodiments, the second processing model may use a machine learning model. The training method of the machine learning model is described in Figure 2 The machine learning model used may be a neural network (NN), a convolutional neural network (CNN), a long short-term memory network (LSTM), etc. Preferably, a long short-term memory network (LSTM) may be used in some embodiments.

[0053] For example, the second semantic vector obtained after the first semantic vector obtained in step 120 is processed by a long short-term memory network (LSTM) is: .

[0054] In some embodiments, the second process model may be calculated using a rule-based or statistical approach.

[0055] Step 140, determining one or more categories of the first original text through a third processing model; the input of the third processing model includes the second semantic vector. Specifically, step 140 can be performed by a classification labeling module.

[0056] The third processing model is a classification model.

[0057] The input of the third processing model is the second semantic vector determined in step 130. The output of the third processing model is the evaluation value of the corresponding one or more categories.

[0058] The output evaluation value can be saved to a storage device, or transmitted to other devices through an interface, or output in other common ways. Output can be carried out through a data output module.

[0059] Since the input vector is obtained from the first original text through the aforementioned steps, the classification of the first original text can be determined according to the output evaluation value.

[0060] In some embodiments, the output is an evaluation value of a category, and the value can be used to determine whether the first original text belongs to the category. The basis for the determination can be a threshold value, which can be set as needed.

[0061] For example, the classification is: whether it belongs to "positive evaluation", the threshold is set to 0.5, and if the output value of the first original text is greater than 0.5, it belongs to "positive evaluation", otherwise it does not belong to "positive evaluation".

[0062] In some embodiments, the output is evaluation values ​​of multiple categories, and it can be determined that the first original text belongs to the category with the largest evaluation value.

[0063] For example, the categories include: "positive evaluation", "negative evaluation", and "neutral evaluation". The evaluation values ​​of the first original text in each category are 0.3, 0.6, and 0.9 respectively, and the first original text belongs to the "neutral evaluation" with the largest evaluation value.

[0064] In some embodiments, the output is evaluation values ​​of multiple categories, and the first original text can be determined to belong to one or more categories according to the magnitude of these values. The basis for the judgment can be a threshold value, which can be set as needed.

[0065] For example, the categories are: "Familiarity with the road", "Service attitude", and "In-car environment". The threshold is set to 0.5. If the output value of the first original text "The driver's skills are OK, but his attitude is bad" is greater than the set threshold, it belongs to the corresponding category. If the evaluation values ​​of the first original text in the above categories are 0.6, 0.9, and 0.3 respectively, the first original text belongs to "Familiarity with the road" and "Service attitude".

[0066] In some embodiments, the third processing model may be a support vector machine (SVM), which may be a neural network.

[0067] Preferably, in some embodiments, the third processing model may use a logistic regression model, which may be obtained by training with sample data.

[0068] When training the third processing model, training data including training text and corresponding classification is obtained through the original text. The training text is processed by the first processing model and the second processing model in sequence to obtain vectors in the training samples. These vectors are used as inputs of the training samples and the corresponding classifications are used as identifiers to train the third processing model.

[0069] The process of processing through the first processing model and the second processing model in sequence is similar to the previous one and will not be repeated here.

[0070] The key to the above method is to obtain the second semantic vector through the second processing model. The second semantic vector can better represent the classification meaning of the original text than the first semantic vector, so that the third processing model can be classified more effectively. Compared with directly training the third processing model through the original text, the method provided in this specification can better solve the problem of insufficient training samples for certain classifications.

[0071] Figure 2 is a flowchart of a second processing model training method according to some embodiments of this specification. Figure 2As shown, the training method 200 of the second processing model may include the following steps:

[0072] In step 210, the second original text may be obtained from a database or other storage device, or through other interfaces. Specifically, step 210 may be performed by a data acquisition module.

[0073] The second original text refers to the text used to train the second processing model. The source and type of the second original text may be the same as the first original text, see step 110 .

[0074] In step 220, first training data is determined based on the second original text. Each sample of the first training data includes a pair of third semantic vectors and a class identifier; the third semantic vector is obtained by vectorizing the second original text through the first processing model; the class identifier indicates whether the second original text corresponding to the pair of third semantic vectors belongs to the same category. Specifically, step 220 can be performed by a training module.

[0075] Each sample of the first training data includes a pair of third semantic vectors and a similar identifier.

[0076] The third semantic vector is obtained by vectorizing the second original text through the first processing model, and the vectorization processing method is similar to step 120.

[0077] The same category flag indicates whether the second original texts corresponding to the pair of third semantic vectors belong to the same category.

[0078] In some embodiments, if the second original texts corresponding to a pair of third semantic vectors belong to the same category, the same category identifier is 0, otherwise it is 1.

[0079] The following is a sample example of the first training data obtained from the second type of original text "user evaluation text of online car-hailing":

[0080] In some embodiments, the user's evaluation text of the online ride-hailing service may include "I feel safe and reliable", "My family is very satisfied", and "The driver has a bad attitude".

[0081] For example, we can select "I feel safe and reliable" and "The driver has a bad attitude" to obtain a pair of third semantic vectors, and use " "and" "I feel safe and reliable" is a positive evaluation, and "the driver's attitude is bad" is a negative evaluation. They can be regarded as not belonging to the same category, and the same category is marked as 1. Therefore, the training sample S1 can be obtained: ( , , 1).

[0082] For another example, we can select "I feel safe and reliable" and "My family is very satisfied" to obtain a pair of third semantic vectors, and use " "and" "I feel safe and reliable" and "My family is very satisfied" are both positive comments and can be regarded as belonging to the same category, and the same category is marked as 0. Therefore, we can obtain training sample S2: ( , , 0).

[0083] The same category identifier of the training data can be obtained by manually screening and marking the second original text, can be calculated based on the classification data of the second original text, or can be obtained by other methods.

[0084] In step 230, each sample of the first training data is input into a fourth processing model for training; the fourth processing model includes two second processing models using the same parameters; the fourth processing model determines a loss function based on the output vectors of the two second processing models and the same type identifiers in the sample; the fourth processing model determines the parameters of the included second processing model through training based on the loss function. Specifically, step 230 can be performed by a training module.

[0085] The fourth processing model includes the second processing model using two identical parameters. The second processing model is described in step 130 .

[0086] A pair of the third semantic vectors in each sample of the first training data is used as the input of two second processing models. For example, the " "and" ” are used as inputs of the two second processing models respectively.

[0087] In some embodiments, a pair of semantic vectors in each sample may be semantic vectors with the same identifier or semantic vectors with different identifiers. For example, a pair of semantic vectors may be "I feel safe and reliable" and "My family is very satisfied" both of which are positive evaluations. For another example, a pair of semantic vectors may be "I feel safe and reliable" which is positively evaluated and "The driver's attitude is bad" which is negatively evaluated.

[0088] The outputs of the two second processing models are each a vector. The output vector of the fourth processing model can be determined by calculating the output vectors of the two second processing models.

[0089] In some embodiments, the output vector of the fourth processing model may be determined by calculating the cosine distance between the output vectors of the two second processing models.

[0090] For example, if the output vectors of the two second processing models are , Indicates that the output vector of the fourth processing model = .

[0091] In some embodiments, the output vector of the fourth processing model may be determined by calculating the Euclidean distance.

[0092] For example, if the output vectors of the two second processing models are , Represents, then: The output vector of the fourth processing model = .

[0093] In other embodiments, other formulas may be used to determine the output vector of the fourth processing model.

[0094] The loss function is a function used to measure the degree of inconsistency between the model's predicted value and the true value. Its input is the predicted value of each sample of the machine learning model and the true value (identified value) of the corresponding sample, and its output can be a non-negative real value.

[0095] In some embodiments, the loss function of the fourth processing model can be determined based on the output vector of the fourth processing model and the same type identification of the corresponding sample. As can be seen from the above description, since the output vector of the fourth processing model can be determined by calculating the output vectors of the two second processing models, the loss function can be determined based on the output vectors of the two second processing models and the same type identification of the corresponding sample.

[0096] In some embodiments, the loss function may be a square loss function, an absolute value loss function, a logarithmic loss function, etc., which is not limited in this specification.

[0097] In some embodiments, the loss function can be expressed as:

[0098] formula:

[0099] in, Representative training samples Similar identifiers in represents the output vector of the fourth processing model, is the loss function value, indicating the predicted value of the fourth model "output vector "Same as the true value" "The degree of inconsistency.

[0100] Take the training samples S1 and S2 in step 230 as an example:

[0101] For example, , They represent the outputs of the two second processing models after the input sample S1, that is, the input of the fourth processing model , the output vector of the fourth processing model determined by the Euclidean distance can be calculated ;by , They are respectively represented as the outputs of the two second processing models after the input sample S2, that is, the input of the fourth processing model , the output vector of the fourth processing model determined by the Euclidean distance can be calculated Then the loss function .

[0102] The output vector of the fourth processing model is obtained by calculating the parameters of the second processing model, so the loss function also includes the parameters of the second processing model.

[0103] For example, the second processing model may be a convolutional neural network (CNN) classification model, and the parameters may include the convolution kernel of the convolutional layer of the convolutional neural network, that is, the weight when the convolutional layer extracts the feature vector.

[0104] For another example, the second processing model may be a long short-term memory network (LSTM) classification model, and the parameters may include weights of an input gate, an output gate, and a forget gate in the long short-term memory network.

[0105] The parameters of the second processing model included in the loss function can be optimized by minimizing the loss function to train the model.

[0106] The method to minimize the loss function can be gradient descent or other model training methods.

[0107] In some embodiments, the parameters of the two second processing models may be updated synchronously through training to obtain two trained second processing models, which are applied in step 130 .

[0108] The above embodiment has at least one of the following technical effects: (1) The output of the fourth processing model is used to evaluate and train the second processing model. The fourth processing model is not limited by the type of the second processing model and has wide applicability. (2) A large number of samples are selected to train the fourth processing model, which can further optimize the parameters of the second processing model, thereby improving the accuracy of the third processing model (classification model). This embodiment uses a large number of samples to train the classification models based on CNN and LSTM, and the results show significant improvement.

[0109] Figure 3 It is an exemplary flow chart shown according to some embodiments of this specification, which illustrates a method for text classification by combining assessor characteristics and travel characteristics.

[0110] like Figure 3 As shown, in some embodiments, the method for classifying vehicle evaluation text may include the following steps:

[0111] Step 310, obtaining the vehicle evaluation text. This step can refer to the description of step 110.

[0112] Step 320 , vectorize the vehicle evaluation text using a first processing model to determine a first semantic vector of the vehicle evaluation text. This step can refer to the description of step 120 .

[0113] Step 330 : Determine a second semantic vector through a second processing model; the input of the second processing model includes the first semantic vector. For this step, please refer to the description of step 130 .

[0114] Step 340, obtaining the characteristics of the corresponding evaluator, as described below.

[0115] Step 350, obtaining corresponding vehicle travel characteristics, see the following description for details.

[0116] Step 330 , step 340 , and step 350 may be performed in any order or simultaneously.

[0117] Step 360: Determine one or more categories of the vehicle evaluation text through a third processing model; the input of the third processing model includes the second semantic vector, the characteristics of the corresponding evaluator and / or the travel characteristics. See the following description for details.

[0118] In step 340, the characteristics of the corresponding evaluator may be obtained from a database or other storage device, or through other interfaces. Step 340 may be performed by a data acquisition module.

[0119] The corresponding evaluator is a user who makes an evaluation of the vehicle. The characteristics of the corresponding evaluator refer to a characteristic vector of at least one or more attributes of the corresponding evaluator.

[0120] For example, the attribute of the corresponding evaluator may include age. For example, the age attribute of the corresponding evaluator A is "18".

[0121] The characteristics of the corresponding evaluator may be obtained by performing vectorization processing on the attributes of the corresponding evaluator.

[0122] The vectorization process can be performed by conventional processing methods. For example, the features corresponding to the evaluation person A are .

[0123] In some embodiments, the characteristics of the corresponding evaluator can be obtained from a database of the corresponding evaluator's registration information. This specification does not limit this.

[0124] In step 350, the travel characteristics of the corresponding vehicle may be obtained from a database or other storage device, or through other interfaces. Step 350 may be performed by a data acquisition module.

[0125] The corresponding vehicle is the vehicle corresponding to the vehicle evaluation. The travel characteristics of the corresponding vehicle refer to the characteristic vector of at least one or more attributes of the travel situation.

[0126] For example, the attributes of the travel situation may include the weights corresponding to the travel time periods of the vehicle ("7:00 to 9:00", "9:00 to 17:00", "17:00 to 19:00", "19:00 to 23:00", "23:00 to 4:00" and "4:00 to 7:00"). For example, the travel time period attribute corresponding to vehicle A is ( .

[0127] For example, the attributes of travel conditions can also include the weights corresponding to the mileage of the vehicle in each travel area ("within the first ring road", "from the first ring road to the second ring road", "from the second ring road to the third ring road", "outside the third ring road" and "cross-region"). For example, the travel area attribute corresponding to vehicle A is .

[0128] For example, the attributes of the travel situation may also include the weights corresponding to the lengths of each trip ("within 5 kilometers", "5 to 10 kilometers", "10 to 20 kilometers" and "more than 20 kilometers") of each trip of the vehicle. For example, the trip length attribute corresponding to vehicle A is ( ).

[0129] The travel characteristics of the corresponding vehicle can be obtained by vectorizing the attributes of the travel situation.

[0130] For example, the attribute corresponding to vehicle A , and( ) are vectorized to obtain the corresponding vehicle features , and .

[0131] The vectorization process may be performed by conventional processing methods.

[0132] Step 360, determining one or more categories of the vehicle evaluation text through a third processing model; the input of the third processing model includes the second semantic vector, the characteristics of the corresponding evaluator and / or the travel characteristics. Step 360 can be performed by a classification labeling module.

[0133] The input data of the third processing model can be obtained by processing the second semantic vector, the characteristics of the corresponding evaluator and / or the travel characteristics.

[0134] In some embodiments, the second semantic vector, the features of the corresponding evaluator and / or the travel features can be combined into a matrix. For example, the corresponding evaluator A in step 340 makes a vehicle evaluation of the corresponding vehicle A in step 350: "I feel safe and reliable". The second semantic vector of the vehicle evaluation text "I feel safe and reliable" is is a 1×4 vector corresponding to the evaluator’s feature vector is a 1×1 vector, the travel feature vector is a 1×3 vector, and we can use “0” to fill the low-dimensional vector to get , , and obtain the combination matrix .

[0135] In some embodiments, weights may be set for the vectors in the combined matrix B, and the weights may be adjusted according to the training of the third processing model, for example: , for weight , and Can be pre-set manually.

[0136] This specification does not limit the specific method of combining multiple features.

[0137] The training method of the third processing model is similar to the training method described in step 140, except that in addition to using the second semantic vector, the corresponding evaluator's characteristics and / or travel characteristics need to be added to the sample data, which will not be repeated here.

[0138] The output of the third processing model and the model selection can be seen in step 140 .

[0139] Comparison of the above methods Figure 1 , adding the corresponding evaluator's characteristics and the corresponding vehicle travel characteristics as the input of the third processing model, and having at least one of the following technical effects: (1) providing context for the vehicle evaluation text, so that the third processing model's judgment of the sentiment tendency text is not limited to the literal meaning of the text, thereby improving the accuracy of the third processing model. (2) The third processing model can integrate multiple word vectors, reducing the data sparsity problem, thereby effectively constructing context information and accurately expressing the meaning of the vehicle evaluation text.

[0140] Figure 4 is an exemplary flow chart shown according to some embodiments of the present specification, illustrating a method for text classification in combination with historical evaluation classification.

[0141] like Figure 4 As shown, in some embodiments, the text classification method may include the following steps:

[0142] Step 410, obtaining the vehicle evaluation text. This step can refer to the description of step 110.

[0143] Step 440, obtaining the historical evaluation classification of the corresponding evaluator and / or the historical evaluation classification of the corresponding vehicle, as described below.

[0144] Step 410 and step 440 may be performed in any order or simultaneously.

[0145] Step 420 , vectorize the vehicle evaluation text using a first processing model to determine a first semantic vector of the vehicle evaluation text. This step may refer to the description of step 120 .

[0146] Step 430 : Determine a second semantic vector through a second processing model; the input of the second processing model includes the first semantic vector. This step can refer to the description of step 130 .

[0147] Step 450, classify and process the historical evaluations through the fifth processing model to obtain a historical evaluation vector, as described below.

[0148] The execution order of step 450, step 420 and step 430 is not limited.

[0149] Step 460: Determine one or more categories of the vehicle evaluation text through a third processing model; the input of the third processing model includes the second semantic vector and the historical evaluation vector obtained in step 450. See the following description for details.

[0150] In step 440, the historical evaluation classification of the corresponding evaluator and / or the historical evaluation classification of the corresponding vehicle may be obtained from a database or other storage device, or through other interfaces. Specifically, step 440 may be performed by a data acquisition module.

[0151] The corresponding evaluator is the user who makes the evaluation of the vehicle. The historical evaluation classification of the corresponding evaluator refers to one or more classifications obtained by processing the content of one or more evaluations made by the corresponding evaluator. This classification can be made by the method described in this specification or by other methods.

[0152] For example, assessor A's historical evaluation of vehicle A was 1: "The driver took a shortcut and sent me to the airport on time", which corresponds to the classification of "road familiarity"; assessor A's historical evaluation of vehicle B was 2: "The strange smell in the car made me carsick", which corresponds to the classification of "in-car environment".

[0153] The corresponding vehicle is the vehicle corresponding to the vehicle evaluation. The historical evaluation classification of the corresponding vehicle refers to the evaluation classification of the vehicle that has been made. This classification can be made by the method described in this specification or by other methods.

[0154] For example, the evaluation 1 of vehicle C by the evaluator Zhang San in history is: "the driver's attitude is bad", and the corresponding classification is "service attitude"; the evaluation 2 of vehicle C by the evaluator Li Si in history is: "the interior of the car is clean", and the corresponding classification is "interior environment".

[0155] In step 450, the historical evaluations are classified and processed by the fifth model to obtain a historical evaluation vector. Step 450 may be executed by a data acquisition module.

[0156] Categorical values ​​can be represented by converting them into scalar values, vectors, or matrices using the above classifications.

[0157] This specification does not limit how to obtain this classification value.

[0158] Each classification obtained in step 440 can be processed and represented in a form similar to a vector, thereby obtaining one or more matrices.

[0159] For example, when the classification value is a scalar, the "road familiarity" and "in-vehicle environment" in the above example can correspond to classification numbers 1 and 2 respectively, and then the historical evaluation classification value of assessor A can be represented by the vector [1,2], thus obtaining the matrix .

[0160] For example, when the classification value is a word vector, the classification values ​​of the historical evaluation categories of vehicle A, "service attitude" and "in-vehicle environment", are , from which we can get the matrix .

[0161] For example, when the classification value is a probability distribution, the classification value corresponding to the historical evaluation 1 of vehicle A is the vector , the classification value of evaluation 2 is the vector , from which we can get the matrix .

[0162] The necessary alignment of the matrix can be performed.

[0163] The above matrix is ​​input into the fifth model, and the historical evaluation vector is output. The historical evaluation vector can be a vector value, which represents the overall tendency of the user's historical evaluation. For example: The historical evaluation vector can also be expressed in other forms.

[0164] The classification value corresponding to each dimension in the historical evaluation vector may be the same as the probability of belonging to the corresponding classification, or may be different from the probability of belonging to the corresponding classification.

[0165] In some embodiments, the fifth model can be a statistical model, a CNN model, an RNN model, etc.

[0166] Preferably, in some embodiments, the fifth model may use an RNN model. The RNN model may be obtained through training.

[0167] When using the RNN model, multiple historical evaluation values ​​can be input as a sequence.

[0168] In the above-mentioned models, the sample data for training can be obtained by preprocessing the historical evaluation classification of several users.

[0169] The annotation of sample data can be obtained in a variety of ways, such as manually analyzing the historical evaluation text of the corresponding user and determining it in combination with user portrait data, and this manual does not impose any restrictions on this.

[0170] In step 460, one or more categories of the first original text are determined by a third processing model, wherein the input of the third processing model includes the second semantic vector and the historical evaluation vector obtained in step 450. Step 460 may be performed by a classification labeling module.

[0171] The input data of the third processing model can be obtained by processing the second semantic vector and the historical evaluation vector.

[0172] In some embodiments, the second semantic vector and the historical evaluation vector may be combined into a matrix, for example: and the historical evaluation vector into a matrix. If and The dimensions are the same, then we can get the combined matrix ;if and If the dimensions are different, the vector with lower dimension can be padded with “0”, see step 360 .

[0173] In some embodiments, weights may be set for the vectors in the combination matrix C, see step 360 .

[0174] This specification does not limit the specific method of combining multiple features.

[0175] The training method of the third processing model is similar to the training method of step 130, except that a historical evaluation vector needs to be added to the sample data, which will not be described in detail here.

[0176] The output of the third processing model and the model selection can be seen in step 140 .

[0177] Comparison of the above methods Figure 1 , adding the characteristics of the corresponding evaluator and the corresponding vehicle travel characteristics as the input of the third processing model, and having at least one of the following technical effects: (1) providing a vector of the time dimension for the third processing model, avoiding the errors caused by the individual differences of different evaluators and different vehicles, thereby improving the accuracy of the third processing model; (2) enriching the word vector categories of the third processing model, reducing the data sparsity problem, and more accurately expressing the meaning of the vehicle evaluation text.

[0178] In some embodiments, by preferably using the RNN model as the fifth model, the context of user evaluations can be combined to generate the input vector of the model, and the context and trends of historical evaluations can be considered more in the model calculation process, thereby better representing the classification tendency of user evaluations.

[0179] Figure 5 is an exemplary flow chart shown according to some embodiments of the present specification, illustrating a method 500 for performing text classification in combination with a third semantic vector.

[0180] Step 510, obtaining the first original text. This step can refer to the description of step 110.

[0181] Step 520 : vectorize the first original text using a first processing model to determine a first semantic vector of the first original text. For this step, please refer to the description of step 120 .

[0182] Step 540, vectorize the first original text through the sixth processing model to determine the third semantic vector of the first original text. The sixth processing model can be an ELMO model, as described below.

[0183] Step 530 : Determine a second semantic vector through a second processing model; the input of the second processing model includes the first semantic vector. For this step, please refer to the description of step 130 .

[0184] Step 550 : Determine a fourth semantic vector through a second processing model; the input of the second processing model includes the third semantic vector. This step can refer to the description of step 130 .

[0185] Step 560: Determine one or more categories of the first original text through a third processing model; the input of the third processing model includes the second semantic vector and the fourth semantic vector.

[0186] Step 540, performing vectorization processing on the first original text through a sixth processing model to determine a third semantic vector of the first original text, and the sixth processing model may be an ELMO model. Step 540 may be performed by a semantic vector module.

[0187] The sixth processing model is a vectorized processing model different from the first processing model. In some embodiments, the sixth processing model can be a Word2Vec model, an ElMo model, a bag-of-words model, etc., which is not limited in this specification.

[0188] In some embodiments, input weights of the vectorized processing results of the first processing model and the sixth processing model in the third processing model may be set.

[0189] In some embodiments, the third processing model can also learn to automatically adjust the input weights of the vectorized processing results of the first processing model and the sixth processing model for different original texts.

[0190] Comparison of the above embodiments Figure 1 , a sixth processing model is added to obtain the third semantic vector as the input of the third processing model, which has at least one of the following technical effects: (1) different types of vectorization models can construct semantic information in different aspects, making the classification of the third processing model more comprehensive; (2) due to the different types of original texts, two different types of vectorization models contribute differently to a specific classification task. The third processing model can have the ability to distinguish by adjusting the weights, avoiding the negative impact of a certain type of processing model, thereby improving the accuracy of the third processing model.

[0191] The essence of ELMO is to use a language model to learn the vectorization of words on a large corpus in advance, and then use the training data without labels to fine-tune the pre-trained ELMO model, so that the context information of the training data can be used to obtain the vectorization of words in the current context. The ELMO model can solve the context problem and dynamically update the vectorization of words.

[0192] For example, the third semantic vector obtained by the ELMO model in step 540 for “I feel safe and reliable” is:

[0193] .

[0194] Step 560: Determine one or more categories of the first original text through a third processing model; the input of the third processing model includes the second semantic vector and the fourth semantic vector. Step 560 may be performed by a classification labeling module.

[0195] The input data of the third processing model can be obtained by processing the second semantic vector and the fourth semantic vector.

[0196] In some embodiments, the second semantic vector and the fourth semantic vector may be combined into a matrix. For example, the second semantic vector of the car evaluation text “I feel safe and reliable” is , the fourth semantic vector is , according to the method of step 360 and The same dimensions are combined into a matrix .

[0197] In some embodiments, weights may be set for the vectors in the combination matrix D, see step 360 .

[0198] In some embodiments, the third processing model can learn to automatically adjust the vector weights in the matrix D for different original texts.

[0199] This specification does not limit the specific method of combining multiple features.

[0200] The training method of the third processing model is similar to the training method of step 130, except that a fourth semantic vector needs to be added to the sample data, which will not be described in detail here.

[0201] The output of the third processing model and the model selection can be seen in step 140 .

[0202] The beneficial effects that may be brought about by the embodiments of the present application include but are not limited to: (1) In combination with the application scenario, the accuracy of the classification model can be improved by inputting different types of data information. (2) Multiple word vectors can be integrated to effectively construct context information and reduce data sparsity problems. (3) The training method of the classification model has a wide range of applicability and is not limited by the type of classification model. It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other possible beneficial effects.

[0203] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of the present application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements and amendments to the present application. Such modifications, improvements and amendments are suggested in the present application, so such modifications, improvements and amendments still belong to the spirit and scope of the exemplary embodiments of the present application.

[0204] At the same time, the present application uses specific words to describe the embodiments of the present application. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0205] In addition, it will be understood by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0206] A computer storage medium may include a propagated data signal containing computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, etc., or a suitable combination. A computer storage medium may be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, device or apparatus to communicate, propagate or transmit the program for use. The program code on the computer storage medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0207] The computer program codes required for the operation of each part of the present application can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C language, Visual Basic, Fortran 2003, Perl, COBOL 2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages, etc. The program code can be run entirely on the user's computer, or run on the user's computer as an independent software package, or run partially on the user's computer and partially on a remote computer, or run entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0208] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this application, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this application. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this application. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.

[0209] Similarly, it should be noted that in order to simplify the description of the disclosure of this application and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this application, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the object of this application are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0210] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used for the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which can be changed according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the method of general digit retention. Although the numerical domains and parameters used to confirm the breadth of the range in some embodiments of the present application are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0211] Each patent, patent application, patent application disclosure, and other materials, such as articles, books, instructions, publications, documents, etc., cited in this application is hereby incorporated by reference in its entirety. Except for application history documents that are inconsistent with or conflicting with the content of this application, documents that limit the broadest scope of the claims of this application (currently or later attached to this application) are also excluded. It should be noted that if the description, definition, and / or use of terms in the attached materials of this application are inconsistent or conflicting with the content described in this application, the description, definition, and / or use of terms in this application shall prevail.

[0212] Finally, it should be understood that the embodiments described in this application are only used to illustrate the principles of the embodiments of the present application. Other variations may also fall within the scope of the present application. Therefore, as an example and not a limitation, the alternative configurations of the embodiments of the present application may be considered to be consistent with the teachings of the present application. Accordingly, the embodiments of the present application are not limited to the embodiments explicitly introduced and described in the present application.

Claims

1. A method for classifying text, comprising: Get the first original text; The first original text is vectorized by a first processing model to determine a first semantic vector of the first original text; Determining a second semantic vector through a second processing model; the input of the second processing model includes the first semantic vector; Determine one or more classification tags of the first original text through a third processing model; the input of the third processing model includes the second semantic vector; wherein the second processing model is obtained through the following training method: Based on the second original text, first training data is determined; in the first training data, each sample includes a pair of third semantic vectors and a similar identifier; The third semantic vector is obtained by vectorizing the second original text through the first processing model; The same category identifier indicates whether the second original texts corresponding to the pair of third semantic vectors belong to the same category; inputting each of said samples of said first training data into a fourth processing model; The fourth processing model includes two of the second processing models using the same parameters; the pair of third semantic vectors are respectively used as inputs of the two second processing models; the fourth processing model determines the loss function based on the distance between the output vectors of the two second processing models and the similar identifiers in the sample; the fourth processing model determines the parameters of the included second processing model through training based on the loss function.

2. The method for classifying text according to claim 1, characterized in that: Also includes: The first original text and the second original text are vehicle evaluation texts; Get the original data of other car evaluations, The other vehicle evaluation raw data include the characteristics of the corresponding evaluator; The input to the third processing model includes characteristics of the evaluator.

3. The method for classifying text according to claim 1, characterized in that: Also includes: The first original text and the second original text are vehicle evaluation texts; Get the original data of other car evaluations, The other vehicle evaluation raw data include corresponding travel characteristics; The travel characteristics include at least one of the following information: travel time period, travel area, and travel mileage; The input of the third processing model includes the travel characteristics.

4. The method for classifying text according to claim 1, characterized in that: Also includes: The first original text and the second original text are vehicle evaluation texts; Get the original data of other car evaluations, The other vehicle evaluation raw data includes the historical evaluation classification of the corresponding evaluator and the historical evaluation classification of the corresponding vehicle; Classifying and processing the historical evaluations through a fifth processing model to obtain a historical evaluation vector; The input to the third processing model includes the historical evaluation vector.

5. The method for classifying text according to claim 4, characterized in that: The fifth processing model is a RNN model trained with data.

6. A system for classifying text, comprising: A data acquisition module, used for acquiring a first original text; A semantic vector module, configured to perform vectorization processing on the first original text through a first processing model to determine a first semantic vector of the first original text; and to determine a second semantic vector through a second processing model; the input of the second processing model includes the first semantic vector; A classification labeling module, configured to determine one or more classification labels of the first original text through a third processing model; the input of the third processing model includes the second semantic vector; Training modules for: Based on the second original text, first training data is determined; in the first training data, each sample includes a pair of third semantic vectors and a similar identifier; The third semantic vector is obtained by vectorizing the second original text through the first processing model; The same category identifier indicates whether the second original texts corresponding to the pair of third semantic vectors belong to the same category; inputting each of said samples of said first training data into a fourth processing model; The fourth processing model includes two of the second processing models using the same parameters; the pair of third semantic vectors are respectively used as inputs of the two second processing models; the fourth processing model determines the loss function based on the distance between the output vectors of the two second processing models and the similar identifiers in the sample; the fourth processing model determines the parameters of the included second processing model through training based on the loss function.

7. The system for classifying text according to claim 6, characterized in that: The data acquisition module is also used for: Acquire a car evaluation text as the first original text and the second original text; Get the original data of other car evaluations, The other vehicle evaluation raw data include the characteristics of the corresponding evaluator; The input to the third processing model includes characteristics of the evaluator.

8. The system for classifying text according to claim 6, characterized in that: The data acquisition module is also used for: Acquire a car evaluation text as the first original text and the second original text; Get the original data of other car evaluations, The other vehicle evaluation raw data include corresponding travel characteristics; The travel characteristics include at least one of the following information: travel time period, travel area, and travel mileage; The input of the third processing model includes the travel characteristics.

9. The system for classifying text according to claim 6, characterized in that: The data acquisition module is also used for: Acquire a car evaluation text as the first original text and the second original text; Get the original data of other car evaluations, The other vehicle evaluation raw data includes the historical evaluation classification of the corresponding evaluator and the historical evaluation classification of the corresponding vehicle; Classifying and processing the historical evaluations through a fifth processing model to obtain a historical evaluation vector; The input to the third processing model includes the historical evaluation vector.

10. The system for classifying text according to claim 9, characterized in that: The fifth processing model is a RNN model trained with data.

11. A device for classifying text, the device comprising a processor and a memory; the memory is used to store instructions, characterized in that: When the instructions are executed by the processor, the device implements operations corresponding to the method for classifying text as described in any one of claims 1 to 5.

12. A computer-readable storage medium, characterized in that: The storage medium stores computer instructions. When the computer reads the computer instructions in the storage medium, the computer runs the method for classifying text as described in any one of claims 1 to 5.