CN108628974B - Public opinion information classification method and device, computer equipment and storage medium - Google Patents
Public opinion information classification method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN108628974B CN108628974B CN201810380769.9A CN201810380769A CN108628974B CN 108628974 B CN108628974 B CN 108628974B CN 201810380769 A CN201810380769 A CN 201810380769A CN 108628974 B CN108628974 B CN 108628974B
- Authority
- CN
- China
- Prior art keywords
- training
- sentences
- sentence
- weight matrix
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to a public opinion information classification method, a public opinion information classification device, computer equipment and a storage medium. The method comprises the following steps: establishing a classification model, wherein the classification model comprises a word vector model and a multilayer recurrent neural network; obtaining public opinion information, wherein the public opinion information comprises a plurality of sentences; training by using a word vector model to obtain sentence vectors corresponding to a plurality of sentences, and generating a weight matrix by using the sentence vectors corresponding to the plurality of sentences; acquiring codes corresponding to the sentences respectively, and inputting the codes of the sentences into the trained multilayer recurrent neural network; calculating based on the codes of the sentences and the weight matrix through the trained multilayer recurrent neural network, and outputting the categories of the sentences; and determining the category corresponding to the public opinion information according to the categories of the sentences. By adopting the method, a large amount of public sentiment information can be effectively classified.
Description
Technical Field
The application relates to the technical field of computers, in particular to a public opinion information classification method, device, computer equipment and storage medium.
Background
With the development of internet technology, people can know hot events at any time. Generally, a great deal of public sentiment information is generated by the hot events, and the development trend of the hot events can be clarified by analyzing the public sentiment information. Public opinion information may be of various kinds, e.g., micro blogs, comments, etc. Before public opinion information is analyzed, proper classification is required. Generally, public opinion information content is short, and text length is different. The traditional semantic expression model is difficult to classify effectively. Therefore, how to effectively classify a large amount of public opinion information becomes a technical problem to be solved at present.
Disclosure of Invention
In view of the above, it is necessary to provide a public opinion information classification method, apparatus, computer device and storage medium capable of effectively classifying a large amount of public opinion information.
A public opinion information classification method, the method comprising:
establishing a classification model, wherein the classification model comprises a word vector model and a multilayer recurrent neural network;
obtaining public opinion information, wherein the public opinion information comprises a plurality of sentences;
training by using a word vector model to obtain sentence vectors corresponding to a plurality of sentences, and generating a weight matrix by using the sentence vectors corresponding to the plurality of sentences;
acquiring codes corresponding to the sentences respectively, and inputting the codes of the sentences into the trained multilayer recurrent neural network;
calculating based on the codes of the sentences and the weight matrix through the trained multilayer recurrent neural network, and outputting the categories of the sentences;
and determining the category corresponding to the public opinion information according to the categories of a plurality of sentences.
In one embodiment, the method further comprises:
acquiring a training set corresponding to public opinion information, wherein the training set comprises a plurality of pieces of sample information, and the sample information comprises a plurality of training sentences and a plurality of training words corresponding to the training sentences;
training a word vector model through the training words to obtain word vectors corresponding to the training words;
training the word vector model through word vectors corresponding to a plurality of training sentences to obtain sentence vectors corresponding to the training sentences;
and training the multilayer cyclic neural network through sentence vectors corresponding to the training sentences to obtain categories corresponding to the training sentences.
In one embodiment, the training a word vector model by using the training word includes:
counting the vocabulary number of training words in a plurality of training sentences, and marking the maximum vocabulary number as a first input parameter;
adding a corresponding number of preset characters in the training sentence according to the difference value between the vocabulary quantity of the training sentence and the maximum vocabulary quantity corresponding to the first input parameter;
and training the word vector model through training words in a plurality of training sentences and the supplemented preset characters to obtain word vectors corresponding to a plurality of training words.
In one embodiment, the training the word vector model by word vectors corresponding to a plurality of training sentences includes:
counting the sentence number of training sentences in the sample information, and marking the maximum sentence number as a second input parameter;
adding a corresponding number of sentences in the sample information by using preset characters according to the difference value between the number of sentences in the sample information and the second input parameter;
and training the word vector model through a plurality of training sentences and newly added sentences to obtain sentence vectors corresponding to the training sentences.
In one embodiment, the training the word vector model through a plurality of training sentences and newly added sentences includes:
acquiring a mapping file corresponding to the training sentence, wherein the mapping file records the category corresponding to the training sentence;
generating a training weight matrix according to a plurality of training sentences and sentence vectors corresponding to the newly added sentences, wherein the training weight matrix corresponds to the sample information after the sentence number is increased;
and training through the multilayer recurrent neural network by utilizing a plurality of training sentences, newly-added sentences and corresponding training weight matrixes, and outputting categories corresponding to the training sentences.
In one embodiment, the multi-layer recurrent neural network nerve comprises a plurality of hidden layers; the training through the multilayer recurrent neural network by using the plurality of training sentences, the newly added sentences and the corresponding training weight matrix comprises:
distributing random vectors to each hidden layer as an initial weight matrix of the hidden layer;
setting a training weight matrix corresponding to the sample information with the increased sentence number between the input layer and the first layer hidden layer according to the second input parameter;
inputting codes corresponding to a plurality of training sentences and codes of the newly added sentences into an input layer of the multilayer recurrent neural network;
and the multilayer hidden layer is trained by utilizing the initial weight matrix and the training weight matrix, and the categories corresponding to the training sentences are output through the output layer.
A public opinion information classification device, the device comprising:
the model establishing module is used for establishing a classification model, and the classification model comprises a word vector model and a multilayer recurrent neural network;
the information acquisition module is used for acquiring public opinion information, wherein the public opinion information comprises a plurality of sentences;
the weight matrix generation module is used for obtaining sentence vectors corresponding to a plurality of sentences through training by using the word vector model and generating a weight matrix by using the sentence vectors corresponding to the sentences;
the classification module is used for acquiring codes corresponding to the sentences respectively and inputting the codes of the sentences to the trained multilayer recurrent neural network; the trained multilayer recurrent neural network operates based on the codes of a plurality of sentences and the weight matrix and outputs the categories of the plurality of sentences; and determining the category corresponding to the public opinion information according to the categories of a plurality of sentences.
In one embodiment, the apparatus further comprises:
the public opinion information processing system comprises a first training module, a second training module and a third training module, wherein the first training module is used for acquiring a training set corresponding to public opinion information, the training set comprises a plurality of pieces of sample information, and the sample information comprises a plurality of training sentences and a plurality of training words corresponding to the training sentences; training a word vector model through the training words to obtain word vectors corresponding to the training words; training the word vector model through word vectors corresponding to a plurality of training sentences to obtain sentence vectors corresponding to the training sentences;
and the second training module is used for training the multilayer recurrent neural network through sentence vectors corresponding to the training sentences to obtain classes corresponding to the training sentences.
A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to realize the steps in the public opinion information classification method embodiment.
A computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the steps in the above-mentioned public opinion information classification method embodiments.
According to the public opinion information classification method, the public opinion information classification device, the computer equipment and the storage medium, when the public opinion information needs to be classified, the server can obtain a plurality of sentences in the public opinion information through word vector model training to obtain corresponding weight vectors, and then weight matrixes corresponding to the sentences are generated. And the server inputs the codes of the sentences into the trained multilayer recurrent neural network, and the trained multilayer recurrent neural network performs operation by using the codes of the sentences and the weight matrix to output the category of each sentence. The server can obtain the category of public opinion information according to the categories of the sentences. Because the weight vector of each sentence is obtained by training a word vector model, the multilayer recurrent neural network is obtained by training a weight matrix aiming at massive sentences. The description of the natural language is effectively mapped to the vector space, so that the convergence efficiency of the multilayer recurrent neural network is improved, and the accuracy of the classification effect is improved. Therefore, a large amount of public opinion information crawled on the network can be effectively classified.
Drawings
Fig. 1 is a diagram illustrating an application scenario of a public opinion information classification method according to an embodiment;
fig. 2 is a flow chart illustrating a public opinion information classification method according to an embodiment;
FIG. 3 is an expanded view of a layer 2 recurrent neural network over time in one embodiment;
FIG. 4 is an expanded view of a 4-layer recurrent neural network over time in one embodiment;
FIG. 5 is an expanded view of a 6-layer recurrent neural network over time in one embodiment;
FIG. 6 is a flowchart illustrating the steps of word vector model training and multi-layer recurrent neural network training in one embodiment;
fig. 7 is a block diagram illustrating a public opinion information classification apparatus according to an embodiment;
FIG. 8 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.
The public opinion information classification method provided by the application can be applied to the application environment shown in fig. 1. The server 102 and the plurality of web servers 104 are connected via a network. The server 102 may be implemented as a stand-alone server or a server cluster composed of a plurality of servers. The server 102 may crawl a plurality of public opinion information from a plurality of website servers 104 according to a preset frequency. The server 102 may identify each sentence of public opinion information according to punctuation. A classification model is established in the server 102, and the classification model comprises a word vector model and a multilayer recurrent neural network. The server 102 obtains sentence vectors corresponding to a plurality of sentences obtained through training of the word vector model, and generates a weight matrix by using the sentence vectors. The server 102 calls the trained multilayer recurrent neural network, obtains codes corresponding to the sentences, and inputs the codes of the sentences into the trained multilayer recurrent neural network. And the trained multilayer recurrent neural network utilizes the codes of the sentences and the weight matrix to carry out operation and outputs the categories of the sentences. The server 102 determines a category corresponding to the public opinion information according to the categories of the sentences. Therefore, effective classification of a large amount of public opinion information is realized.
In one embodiment, as shown in fig. 2, a public opinion information classification method is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes the following steps:
A classification model can be pre-established in the server, and the classification model comprises a word vector model and a multilayer recurrent neural network. The word vector model may employ a Skip-Gram model, i.e., the model may employ a neural network structure including an input vector, a hidden layer, and an output layer. In the conventional manner, the final result is output through the output layer of the model, and the final result is a probability distribution. Such probability distributions are not applicable to multi-layer recurrent neural networks. Therefore, in this embodiment, only by adopting the structures of the input vector and the hidden layer of the model, the weight vectors of a plurality of words are output through the hidden layer, and the operation is not continued through the output layer.
In the multi-layer recurrent neural network, a plurality of layers of hidden layers can be included, and the hidden layers include a forward-calculation layer and a backward-calculation layer, which can also be called a hidden layer for bidirectional calculation. The hidden layers of the first layer comprise a first forward calculation layer and a first backward calculation layer, the hidden layers of the second layer comprise a second forward calculation layer and a second backward calculation layer, the hidden layers of the third layer comprise a third forward calculation layer and a third backward calculation layer, and the rest can be done in the same way. The hidden layer of the first layer may also be referred to as the first hidden layer for short, and so on. Corresponding weight matrixes are arranged between the input layer and the hidden layer of the first layer, namely corresponding weight matrixes are respectively arranged between the input layer and the first forward calculation layer and between the input layer and the first backward calculation layer.
The server can crawl various public opinion information from a plurality of websites according to preset frequency. The type of public opinion information may include various sports, finance, entertainment, education, etc. Each piece of public opinion information can comprise a plurality of sentences, and each sentence comprises a plurality of words. The server can identify each sentence of the public sentiment information according to the punctuation marks. The server can also perform word segmentation processing on each sentence to obtain words in each sentence.
And 206, training by using the word vector model to obtain sentence vectors corresponding to a plurality of sentences, and generating a weight matrix by using the sentence vectors corresponding to the plurality of sentences.
In a conventional method, the weight matrices corresponding to the first forward estimation layer and the first backward estimation layer are initialized to random vectors, but this may result in poor convergence effect of the multi-layer recurrent neural network, and the output result cannot meet the requirement.
In this embodiment, the server uses a weight matrix corresponding to a plurality of sentences as a weight matrix between the input layer and the first hidden layer in the multi-layer recurrent neural network. The weight matrix is obtained by training a word vector model. The method can effectively map the description of the natural language to the vector space, improve the convergence efficiency of the multilayer recurrent neural network, and further improve the accuracy of the output effect.
The weight matrixes corresponding to the first forward estimation layer and the first backward estimation layer are different. The server can obtain a weight vector corresponding to each sentence according to the description sequence of the public opinion information, and the weight vector corresponding to each sentence can be a vector array. The server generates a corresponding weight matrix estimated forward by using the weight vectors corresponding to the sentences. The server can obtain the weight vector corresponding to each sentence again according to the reverse description sequence of the sentences in the public opinion information, and generates a backward reckoning weight matrix corresponding to the sentences. The weight matrix of forward calculation is the weight matrix between the input layer and the first forward calculation layer in the multi-layer recurrent neural network. The backward calculation weight matrix is the weight matrix between the input layer and the first backward calculation layer in the multilayer recurrent neural network.
By taking public sentiment information as an example of a microblog, the public sentiment can be that the Pingchang Donaoao just finishes and the Donaohui has entered Beijing. 2022 and oil-adding in Beijing Donao. And (5) oiling in China. The server can generate a weight matrix calculated forwards according to the forward description sequence of 'Pingchang Donaoao is just finished, donaohui meeting has entered Beijing time', '2022 Beijing Donao oiling' and 'Chinese oiling'. The server can also generate a backward-reckoning weight matrix according to a reverse description sequence of ' Chinese refueling ', ' 2022 Beijing winter refueling ', ' Pingchang winter refueling ' and ' Changchang winter afterward ' when the winter afterward meeting enters Beijing time '.
And 210, determining a category corresponding to the public opinion information according to the categories of the sentences.
The multilayer implicit layer in the multilayer recurrent neural network can be 2 layers, 4 layers or 6 layers, etc. Wherein, each layer hidden layer comprises a forward calculation layer and a backward calculation layer. As shown in fig. 3-5, which are time development diagrams of 2-layer, 4-layer, and 6-layer recurrent neural networks, respectively. Wherein Relu represents an activation function, lstm represents a long-short memory unit, and Softmax represents a classification function. w (— denotes a positive integer) denotes a weight matrix. As can be seen from the expanded view, each forward estimation layer and each backward estimation layer are provided with corresponding initial weight matrixes. Such as w2, w5 in fig. 3, w5, w6, w8 in fig. 4, and w3, w5, w7, w8, w10, w12 in fig. 5.
The multi-layer recurrent neural network may be pre-trained. When the multi-layer cyclic neural network is trained, a mapping file corresponding to public sentiment information can be used for training, and types corresponding to a plurality of sentences are recorded in the mapping file. Because the multilayer recurrent neural network only accepts numerical input, the server encodes a plurality of sentences of each piece of public opinion information during training. Specifically, the server generates a training table using the sample information before training. A plurality of training sentences are recorded in the training table, and each training sentence corresponds to a plurality of training words. The server encodes each training word, and then encodes each sentence according to the encoding of the training words.
And the server calls the trained multilayer circulating neural network and inputs the codes of a plurality of sentences in the public sentiment information into an input layer of the multilayer circulating neural network. The input layer activates the weight matrix of the first forward calculation layer through the activation function, activates the weight matrix of the first backward calculation layer, and starts to operate by combining the initial weight matrix of the first forward calculation layer and the initial weight matrix of the first backward calculation layer. Wherein there is no information flow between the forward estimation layer and the backward estimation layer.
The example will be described with the multi-layer recurrent neural network after training being a 4-layer recurrent neural network. The input in the input layer can be the codes of 'Pingchang Donao just finished, donao meeting already enters Beijing time', '2022 Beijing Donao oiling' and 'Chinese oiling'. w1 is a weight matrix of the first forward estimation layer, w3 is an initial weight matrix of the first forward estimation layer, and after Lstm operation, a forward estimation weight matrix w3 (w 3 is different from the initial w3, and the same reference numerals are used here for brevity) and a weight matrix w4 corresponding to the second forward estimation layer are output. w2 is a weight matrix of the first backward estimation layer, w6 is an initial weight matrix of the first backward estimation layer, and after Lstm operation, the backward estimation weight matrix w6 (in this case, w6 is different from the initial w6, and the same reference numerals are used for brevity) and a weight matrix w7 corresponding to the second backward estimation layer are output. And repeating the steps until the output layer outputs the category of each sentence in turn through the classification function.
The server counts the categories of a plurality of sentences in the public opinion information and sorts the category statistical number. And one or more categories are taken as categories corresponding to the public opinion information in a high-low order. For example, a category corresponding to a microblog may be sports, news, or the like.
In this embodiment, when public opinion information needs to be classified, the server may obtain a plurality of sentences in the public opinion information through word vector model training to obtain corresponding weight vectors, and then generate weight matrices corresponding to the plurality of sentences. And the server inputs the codes of the sentences into the trained multilayer recurrent neural network, and the trained multilayer recurrent neural network performs operation by using the codes of the sentences and the weight matrix to output the category of each sentence. The server can obtain the category of public opinion information according to the categories of the sentences. Because the weight vector of each sentence is obtained by training the word vector model, the multilayer recurrent neural network is obtained by training the weight matrix of massive sentences. The description of the natural language is effectively mapped to the vector space, so that the convergence efficiency of the multilayer recurrent neural network is improved, and the accuracy of the classification effect is improved. Therefore, a large amount of public opinion information crawled on the network can be effectively classified.
In one embodiment, the method further comprises: training a word vector model and training a multi-layer recurrent neural network. As shown in fig. 6, the following are included:
The server can crawl various public opinion information at a plurality of websites, and store the crawled public opinion information into a database. And the server takes the crawled public opinion information as a corpus to carry out pretreatment, including sentence segmentation, word segmentation, cleaning and the like. And the server establishes a corpus by utilizing the preprocessed corpus. And the server marks the preprocessed corpus into sample information according to a preset proportion in the corpus. The server generates a training set using the sample information. The training set comprises training sentences corresponding to a plurality of pieces of sample information and training words corresponding to the training sentences. The word vector model and the multilayer recurrent neural network can be trained in advance through a training set. The multi-layer recurrent neural network needs sentence vectors obtained by training a word vector model during training. When the word vector model trains sentence vectors of a plurality of sentences by using the training set, the word vector of each sentence needs to be relied on.
The word vector model may employ a Skip-Gram model, i.e., the model may employ a neural network structure including an input vector, a hidden layer, and an output layer. In the conventional manner, the final result is output through the output layer of the model, and the final result is a probability distribution. Such probability distributions are not applicable to multi-layer recurrent neural networks. Therefore, in this embodiment, only the input vector of the model and the structure of the hidden layer are used, and the weight vectors of a plurality of words are output through the hidden layer, so that the operation is not continued through the output layer.
Because the word vector model and the multilayer recurrent neural network only accept numerical value input, the server generates a training table by using the sample information during training. A plurality of training sentences are recorded in the training table. The server also generates a corresponding training vocabulary according to the training words. The server encodes each training word, and then encodes each sentence according to the encoding of the training words.
When the classification model is trained, the server firstly trains the word vector model by taking the codes of a plurality of training words in the training set as input vectors to obtain word vectors corresponding to the training words. Secondly, the server trains the word vector model again by using the code of each sentence in the sample information and the word vectors of the corresponding words to obtain the sentence vector corresponding to the training sentence. And then, the server generates a training weight matrix by using the sentence vectors of the training sentences, and trains the multilayer recurrent neural network by using the training weight matrix and the codes of the sentences to obtain the corresponding category of each training sentence.
In a conventional method, since the weight matrices corresponding to the first forward estimation layer and the first backward estimation layer of the multi-layer recurrent neural network are initialized to random vectors, the convergence effect of the multi-layer recurrent neural network may be poor, and sentences cannot be classified effectively. In the embodiment, the word vector of each training word can be accurately obtained by training the training words in the sample information. And training by using the word vectors corresponding to the training words again to accurately obtain the sentence vectors corresponding to each training sentence. Therefore, the natural language is mapped to the vector space, the convergence effect of the multilayer recurrent neural network can be effectively improved, and effective classification of a plurality of sentences is realized.
In one embodiment, training the word vector model with the training words comprises: counting the vocabulary number of training words in a plurality of training sentences, and marking the maximum vocabulary number of the training words in the plurality of training sentences as a first input parameter; adding a corresponding number of preset characters in the training sentence according to the difference value between the vocabulary number of the training sentence and the maximum vocabulary number corresponding to the first input parameter; and training the word vector model through the training words in the training sentences and the supplemented preset characters to obtain word vectors corresponding to the training words.
Because the number of words of different sentences in the public opinion information is different, in order to enable the trained word vector model to be suitable for diversified sentences, the first input parameter is set for the word vector model in the embodiment. The server can count the number of the training words in the training sentences to obtain the number of the training words corresponding to each training sentence, and mark the maximum number of the training words in the training sentences as the first input parameter. For a training sentence with a vocabulary quantity smaller than the first input parameter, the server may increase a corresponding number of preset characters according to a difference between the vocabulary quantity of the training sentence and the first input parameter. The preset character may be a character which does not conflict with public opinion information, such as a null character. For example, the first input parameter is 20, the corresponding first output parameter is also 20, and assuming that the vocabulary number of a training sentence is 10, the server adds 10 preset characters to the training sentence. And the server trains the word vector model by using the codes corresponding to the training words and the codes of the supplemented preset characters, so as to obtain the weight vector corresponding to each training word and the preset characters. The padded predetermined character may also be referred to as a newly added character.
In one embodiment, training the word vector model with word vectors corresponding to a plurality of training sentences includes: counting the sentence number of training sentences in the sample information, and marking the maximum sentence number as a second input parameter; according to the difference value between the sentence number of the sample information and the second input parameter, adding a corresponding number of sentences in the sample information by using preset characters; and training the word vector model through a plurality of training sentences and newly added sentences to obtain sentence vectors corresponding to a plurality of training sentences.
Since the number of sentences in different public opinion information is different, in order to make the word vector model suitable for diversified public opinion information, the second input parameter is set for the word vector model in the embodiment. The server may count the number of sentences of the training sentences in the plurality of pieces of sample information, and mark the maximum number of sentences as the second input parameter. For sample information having a number of sentences smaller than the second input parameter, the server may add a corresponding number of sentences according to a difference between the number of sentences of the sample information and the second input parameter. The sentence to be added may be composed of preset characters. The preset character may be a character which does not conflict with public opinion information, such as a null character. And the server trains the word vector model again by using the plurality of training sentences and the word vectors corresponding to the supplemented sentences, thereby obtaining the weight vector corresponding to each training sentence. The sentence to be added may also be referred to as a new added sentence.
Further, before the training sentences are trained by the server, the number of words of the training words in each training sentence can be increased according to the first input parameter, so that the number of words of each training sentence after the preset characters are added reaches the value of the first input parameter. The server increases the number of sentences of each training sentence in the sample information according to the second input parameter, so that the number of sentences in each sample information reaches the value of the second input parameter. And the server utilizes the training sentences with the increased vocabulary quantity to train again through the word vector model to obtain sentence vectors corresponding to a plurality of training sentences. Therefore, the word vector model can be further fixed, and the universality of the trained word vector model is effectively improved.
In one embodiment, training the word vector model with a plurality of training sentences and a newly added sentence comprises: acquiring a mapping file corresponding to a training sentence, wherein the mapping file records the category corresponding to the training sentence; generating a training weight matrix according to the training sentences and sentence vectors corresponding to the newly added sentences, wherein the training weight matrix corresponds to the sample information after the number of the sentences is increased; and training by utilizing a plurality of training sentences, the newly-added sentences and the corresponding training weight matrix through a multilayer recurrent neural network, and outputting the corresponding categories of the training sentences.
In order to fix the model structure of the multi-layer recurrent neural network, the multi-layer recurrent neural network after training has universality. In this embodiment, the second input parameters are set for all the multilayer recurrent neural networks. The server may generate a forward-estimated training weight matrix corresponding to each post-sentence sample information (i.e., post-sentence sample information complemented by the second input parameter) and a backward-estimated training weight matrix with reference to the above-described embodiment.
Referring to the manner in the above embodiment, the server obtains the code of each training sentence and the code corresponding to the newly added sentence, inputs the corresponding code to the input layer of the multi-layer recurrent neural network, sets the training weight matrix derived forward as the weight matrix of the first forward-derived layer, and sets the training weight matrix derived backward as the weight matrix of the first backward-derived layer. The server sets a plurality of forward-estimated weight matrixes between the input layer and the first forward-estimated layer according to the second input parameter. The server sets a plurality of backward reckoning weight matrixes between the input layer and the first backward reckoning layer according to the second input parameter. For example, if the second input parameter is 10, 10 forward-estimated weight matrices are set between the server input layer and the first forward-estimated layer, and 10 backward-estimated weight matrices are set between the server input layer and the first backward-estimated layer. That is, the server may set 10 w1 and 10 w2 in fig. 4. The w1 includes the weight matrix of forward estimation corresponding to 10 training sentences and the newly added sentence in the sample information. The w2 includes backward-estimated weight matrixes corresponding to 10 training sentences and the newly added sentences in the sample information. The server initializes the initial weight matrix of each forward calculation layer in the hidden layer and initializes the initial weight matrix of each backward calculation layer in the hidden layer. After initialization, the server trains the multi-layer recurrent neural network and outputs a category corresponding to each training sentence. For the output of the preset character, the preset character can also be output. The training result is not influenced.
In the training process, the sentence vector of each training sentence obtained by training the word vector model is adopted, so that the vector condition of each training sentence can be more accurately reflected, the convergence effect of the multilayer cyclic neural network is effectively improved, and the accuracy of the multilayer cyclic neural network training can be improved. By setting the second input parameters, the number of sentences corresponding to each piece of sample information is the same, and therefore the trained word vector model and the trained multilayer recurrent neural network have universality. And various models do not need to be trained, so that the workload of developers is effectively reduced.
Further, before the training of the multi-layer recurrent neural network, the first input parameters may be set to the word vector model in a manner as provided in the above embodiment, so that the number of words in each training sentence is the same. The number of sentences in the multiple sample information adopted by training is the same, and the number of words in each sentence is the same, so that the universality of the trained word vector model and the trained multilayer recurrent neural network can be further improved.
In one embodiment, the multi-layer recurrent neural network nerve comprises a plurality of hidden layers; utilizing the plurality of training sentences, the newly added sentences and the corresponding training weight matrix to train through the multilayer recurrent neural network comprises the following steps: allocating random vectors to each hidden layer as an initial weight matrix of the hidden layer; setting a training weight matrix corresponding to the sample information with the increased sentence number between the input layer and the first layer hidden layer according to the second input parameter; inputting the codes corresponding to the training sentences and the codes of the newly added sentences into an input layer of the multi-layer recurrent neural network; the multilayer hidden layer is trained by utilizing the initial weight matrix and the training weight matrix, and the categories corresponding to the training sentences are output through the output layer.
When the server trains the multi-layer recurrent neural network through the training words, each hidden layer needs to be initialized. Each hidden layer may include a forward dead reckoning layer and a backward dead reckoning layer. The forward and backward calculation layers of each hidden layer need to be initialized. In a traditional mode, initial weight matrixes corresponding to a forward calculation layer and a backward calculation layer of each hidden layer are initialized to be 0, but the generalization capability of the multi-layer recurrent neural network obtained by training in the mode is limited, and retraining is possibly needed if more public opinion information in different formats exists in the future.
In this embodiment, during initialization, the server allocates random vectors as initial weight matrices to the forward-estimation layer and the backward-estimation layer of each hidden layer. The random vector may be an array of preset lengths, for example, 200 or 300 dimensions. After the initialization is completed, the server sets a training weight matrix corresponding to the sample information with the increased sentence number between the input layer and the first hidden layer. And the server inputs the codes corresponding to the training sentences and the codes of the newly added sentences into an input layer of the multi-layer recurrent neural network. The initial weight matrix and the training weight matrix may be used for training through a plurality of hidden layers in a manner provided in the above embodiments, and the category of each training sentence may be output through an output layer.
Because each layer of hidden layer is configured with random vectors as an initial weight matrix during initialization, the generalization capability of the multilayer recurrent neural network can be effectively improved, and the method can be applied to more diversified public opinion information in the future. And a plurality of models are not required to be trained, so that the workload of developers is effectively reduced.
It should be understood that although the steps in the flowcharts of fig. 2 and 6 are shown in sequence as indicated by the arrows, the steps are not necessarily performed in sequence as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2 and 6 may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the sub-steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least some of the sub-steps or stages of other steps.
In one embodiment, as shown in fig. 7, there is provided a public opinion information classification apparatus including: a model building module 702, an information obtaining module 704, a weight matrix generating module 706, and a classifying module 708, wherein:
the model building module 702 is configured to build a classification model, where the classification model includes a word vector model and a multi-layer recurrent neural network.
The information obtaining module 704 is configured to obtain public opinion information, where the public opinion information includes a plurality of sentences.
The weight matrix generating module 706 is configured to train to obtain sentence vectors corresponding to multiple sentences by using the word vector model, and generate a weight matrix by using the sentence vectors corresponding to multiple sentences.
A classification module 708, configured to obtain codes corresponding to the multiple sentences, and input the codes of the multiple sentences to the trained multilayer recurrent neural network; the trained multilayer recurrent neural network carries out operation based on the codes of a plurality of sentences and the weight matrix, and outputs the categories of the plurality of sentences; and determining the category corresponding to the public opinion information according to the categories of the sentences.
In one embodiment, the apparatus further comprises: a first training module 710 and a second training module 712, wherein:
a first training module 710, configured to obtain a training set corresponding to public opinion information, where the training set includes a plurality of pieces of sample information, and the sample information includes a plurality of training sentences and a plurality of training words corresponding to the training sentences; training the word vector model through the training words to obtain word vectors corresponding to the training words; training the word vector model through the word vectors corresponding to the training sentences to obtain sentence vectors corresponding to the training sentences;
the second training module 712 is configured to train the multi-layer recurrent neural network through sentence vectors corresponding to the training sentences to obtain classes corresponding to the training sentences.
In one embodiment, the first training module 710 is further configured to count vocabulary numbers of training words in a plurality of training sentences, and mark the maximum vocabulary number as a first input parameter; adding a corresponding number of preset characters in the training sentence according to the difference value between the vocabulary number of the training sentence and the maximum vocabulary number corresponding to the first input parameter; training the word vector model through training words in the training sentences and the supplemented preset characters to obtain word vectors corresponding to the training words.
In one embodiment, the first training module 710 is further configured to count the sentence number of the training sentences in the sample information, and mark the maximum sentence number as the second input parameter; according to the difference value between the sentence number of the sample information and the second input parameter, adding a corresponding number of sentences in the sample information by using preset characters; and training the word vector model through a plurality of training sentences and newly added sentences to obtain sentence vectors corresponding to a plurality of training sentences.
In one embodiment, the second training module 712 is further configured to obtain a mapping file corresponding to the training sentence, where the category corresponding to the training sentence is recorded in the mapping file; generating a training weight matrix according to the training sentences and sentence vectors corresponding to the newly added sentences, wherein the training weight matrix corresponds to the sample information after the sentence quantity is increased; and training by utilizing a plurality of training sentences, the newly-added sentences and the corresponding training weight matrix through a multilayer recurrent neural network, and outputting the corresponding categories of the training sentences.
In one embodiment, the second training module 712 is further configured to assign a random vector to each of the hidden layers as an initial weight matrix of the hidden layer; setting a training weight matrix corresponding to the sample information with the increased sentence number between the input layer and the first layer hidden layer according to the second input parameter; inputting the codes corresponding to the training sentences and the codes of the newly added sentences into an input layer of the multi-layer recurrent neural network; the multilayer hidden layer is trained by utilizing the initial weight matrix and the training weight matrix, and the categories corresponding to the training sentences are output through the output layer.
For specific limitations of the public opinion information classification device, reference may be made to the above limitations on the public opinion information classification method, which is not described herein again. All or part of the modules in the public opinion information classification device can be realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 8. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operating system and the computer program to run on the non-volatile storage medium. The database of the computer equipment is used for storing public opinion information, sample information and the like. The network interface of the computer device is used for communicating with an external server through a network connection. The computer program is executed by a processor to realize a public opinion information classification method.
Those skilled in the art will appreciate that the architecture shown in fig. 8 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory storing a computer program and a processor implementing the steps of the above-described method embodiments when the processor executes the computer program.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the respective method embodiments described above.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), rambus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
All possible combinations of the technical features in the above embodiments may not be described for the sake of brevity, but should be considered as being within the scope of the present disclosure as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, and these are all within the scope of protection of the present application. Therefore, the protection scope of the present patent application shall be subject to the appended claims.
Claims (10)
1. A public opinion information classification method, the method comprising:
establishing a classification model, wherein the classification model comprises a word vector model and a multilayer recurrent neural network;
acquiring public opinion information, wherein the public opinion information comprises a plurality of sentences;
training by using a word vector model to obtain sentence vectors corresponding to a plurality of sentences, and generating a weight matrix by using the sentence vectors corresponding to the plurality of sentences, wherein the method comprises the following steps: acquiring a weight vector of each corresponding sentence according to the description sequence of the public opinion information, and generating a corresponding forward-reckoning weight matrix by using the weight vectors corresponding to the sentences; acquiring the weight vector of each sentence again by utilizing the reverse description sequence of a plurality of sentences in the public opinion information, and generating backward reckoning weight matrixes corresponding to the plurality of sentences; the forward-reckoning weight matrix is used as a weight matrix between an input layer and a first forward-reckoning layer in the multilayer recurrent neural network, and the backward-reckoning weight matrix is used as a weight matrix between the input layer and the first backward-reckoning layer in the multilayer recurrent neural network;
acquiring codes corresponding to the sentences respectively, and inputting the codes of the sentences into the trained multilayer recurrent neural network; the trained multilayer recurrent neural network operates based on the codes of a plurality of sentences and the weight matrix and outputs the categories of the plurality of sentences;
and determining the category corresponding to the public opinion information according to the categories of the sentences.
2. The method of claim 1, further comprising:
acquiring a training set corresponding to public opinion information, wherein the training set comprises a plurality of pieces of sample information, and the sample information comprises a plurality of training sentences and a plurality of training words corresponding to the training sentences;
training a word vector model through the training words to obtain word vectors corresponding to the training words;
training the word vector model through word vectors corresponding to a plurality of training sentences to obtain sentence vectors corresponding to the training sentences;
and training the multilayer recurrent neural network through sentence vectors corresponding to the training sentences to obtain classes corresponding to the training sentences.
3. The method of claim 2, wherein training a word vector model using the training words comprises:
counting the vocabulary quantity of training words in a plurality of training sentences, and marking the maximum vocabulary quantity as a first input parameter;
adding a corresponding number of preset characters in the training sentence according to the difference value between the vocabulary number of the training sentence and the maximum vocabulary number corresponding to the first input parameter;
and training the word vector model through training words in a plurality of training sentences and the supplemented preset characters to obtain word vectors corresponding to the training words.
4. The method of claim 2, wherein the training the word vector model with word vectors corresponding to a plurality of training sentences comprises:
counting the sentence number of training sentences in the sample information, and marking the maximum sentence number as a second input parameter;
adding a corresponding number of sentences in the sample information by using preset characters according to the difference value between the number of sentences in the sample information and the second input parameter;
and training the word vector model through a plurality of training sentences and newly added sentences to obtain sentence vectors corresponding to the training sentences.
5. The method of claim 4, wherein the training the word vector model with a plurality of training sentences and a newly added sentence comprises:
acquiring a mapping file corresponding to the training sentence, wherein the mapping file records the category corresponding to the training sentence;
generating a training weight matrix according to a plurality of training sentences and sentence vectors corresponding to the newly added sentences, wherein the training weight matrix corresponds to the sample information after the sentence quantity is increased;
and training through the multilayer recurrent neural network by utilizing a plurality of training sentences, newly-added sentences and corresponding training weight matrixes, and outputting categories corresponding to the training sentences.
6. The method of claim 5, wherein the multi-layer recurrent neural network nerve comprises a plurality of hidden layers; the training through the multilayer recurrent neural network by using the plurality of training sentences, the newly added sentences and the corresponding training weight matrix comprises:
distributing random vectors to each hidden layer as an initial weight matrix of the hidden layer;
setting a training weight matrix corresponding to the sample information with the increased sentence number between the input layer and the first layer hidden layer according to the second input parameter;
inputting codes corresponding to a plurality of training sentences and codes of newly added sentences into an input layer of the multilayer recurrent neural network;
and the multilayer hidden layer is trained by utilizing the initial weight matrix and the training weight matrix, and the categories corresponding to the training sentences are output through the output layer.
7. The utility model provides a public opinion information classification device which characterized in that, the device includes:
the model establishing module is used for establishing a classification model, and the classification model comprises a word vector model and a multilayer recurrent neural network;
the information acquisition module is used for acquiring public opinion information, and the public opinion information comprises a plurality of sentences;
the weight matrix generation module is used for training by using the word vector model to obtain sentence vectors corresponding to a plurality of sentences, and generating the weight matrix by using the sentence vectors corresponding to the plurality of sentences, and comprises: acquiring a weight vector of each corresponding sentence according to the description sequence of the public opinion information, and generating a corresponding forward-reckoning weight matrix by using the weight vectors corresponding to the sentences; acquiring the weight vector of each sentence again by utilizing the reverse description sequence of a plurality of sentences in the public opinion information, and generating backward reckoning weight matrixes corresponding to the plurality of sentences; the forward-reckoning weight matrix is used as a weight matrix between an input layer and a first forward-reckoning layer in the multilayer recurrent neural network, and the backward-reckoning weight matrix is used as a weight matrix between the input layer and the first backward-reckoning layer in the multilayer recurrent neural network;
the classification module is used for acquiring codes corresponding to the sentences respectively and inputting the codes of the sentences to the trained multilayer recurrent neural network; the trained multilayer recurrent neural network carries out operation based on the codes of a plurality of sentences and the weight matrix and outputs the categories of the sentences; and determining the category corresponding to the public opinion information according to the categories of the sentences.
8. The apparatus of claim 7, further comprising:
the public opinion information processing device comprises a first training module, a second training module and a third training module, wherein the first training module is used for acquiring a training set corresponding to public opinion information, the training set comprises a plurality of pieces of sample information, and the sample information comprises a plurality of training sentences and a plurality of training words corresponding to the training sentences; training a word vector model through the training words to obtain word vectors corresponding to the training words; training the word vector model through word vectors corresponding to a plurality of training sentences to obtain sentence vectors corresponding to the training sentences;
and the second training module is used for training the multilayer recurrent neural network through sentence vectors corresponding to the training sentences to obtain classes corresponding to the training sentences.
9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 6 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810380769.9A CN108628974B (en) | 2018-04-25 | 2018-04-25 | Public opinion information classification method and device, computer equipment and storage medium |
| PCT/CN2018/097033 WO2019205318A1 (en) | 2018-04-25 | 2018-07-25 | Public opinion information classification method and apparatus, computer device, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810380769.9A CN108628974B (en) | 2018-04-25 | 2018-04-25 | Public opinion information classification method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108628974A CN108628974A (en) | 2018-10-09 |
| CN108628974B true CN108628974B (en) | 2023-04-18 |
Family
ID=63694487
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810380769.9A Active CN108628974B (en) | 2018-04-25 | 2018-04-25 | Public opinion information classification method and device, computer equipment and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN108628974B (en) |
| WO (1) | WO2019205318A1 (en) |
Families Citing this family (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109620154A (en) * | 2018-12-21 | 2019-04-16 | 平安科技（深圳）有限公司 | Borborygmus voice recognition method and relevant apparatus based on deep learning |
| CN110019819A (en) * | 2019-03-26 | 2019-07-16 | 方正株式(武汉)科技开发有限公司 | Method of generating classification model, electronic contract automatic content classification method and device |
| CN110377744B (en) * | 2019-07-26 | 2022-08-09 | 北京香侬慧语科技有限责任公司 | Public opinion classification method and device, storage medium and electronic equipment |
| CN112580329B (en) * | 2019-09-30 | 2024-02-20 | 北京国双科技有限公司 | Text noise data identification method, device, computer equipment and storage medium |
| CN113642302B (en) * | 2020-04-27 | 2024-04-02 | 阿里巴巴集团控股有限公司 | Training method and device for text filling model, text processing method and device |
| CN111581982B (en) * | 2020-05-06 | 2023-02-17 | 首都师范大学 | Ontology-based prediction method for public opinion early warning grade of medical dispute case |
| CN111738017B (en) * | 2020-06-24 | 2024-09-10 | 深圳前海微众银行股份有限公司 | Intention recognition method, device, equipment and storage medium |
| CN112036439B (en) * | 2020-07-30 | 2023-09-01 | 平安科技（深圳）有限公司 | Dependency relationship classification method and related equipment |
| CN111881687B (en) * | 2020-08-03 | 2024-02-20 | 浪潮云信息技术股份公司 | Relation extraction method and device based on context coding and multi-layer perceptron |
| CN112016296B (en) * | 2020-09-07 | 2023-08-25 | 平安科技（深圳）有限公司 | Sentence vector generation method, sentence vector generation device, sentence vector generation equipment and sentence vector storage medium |
| CN112115268B (en) * | 2020-09-28 | 2024-04-09 | 支付宝(杭州)信息技术有限公司 | Training method and device based on feature encoder, and classifying method and device |
| CN112183030A (en) * | 2020-10-10 | 2021-01-05 | 深圳壹账通智能科技有限公司 | Event extraction method and device based on preset neural network, computer equipment and storage medium |
| CN114386394A (en) * | 2020-10-16 | 2022-04-22 | 电科云(北京)科技有限公司 | Prediction model training method, prediction method and prediction device for platform public opinion data theme |
| CN112632984A (en) * | 2020-11-20 | 2021-04-09 | 南京理工大学 | Graph model mobile application classification method based on description text word frequency |
| CN112560505A (en) | 2020-12-09 | 2021-03-26 | 北京百度网讯科技有限公司 | Recognition method and device of conversation intention, electronic equipment and storage medium |
| CN112862672B (en) * | 2021-02-10 | 2024-04-16 | 厦门美图之家科技有限公司 | Liu-bang generation method, device, computer equipment and storage medium |
| CN113515626A (en) * | 2021-05-19 | 2021-10-19 | 中国工商银行股份有限公司 | Method, device and equipment for determining public opinion category |
| CN113190762B (en) * | 2021-05-31 | 2024-08-13 | 南京报业集团有限责任公司 | Network public opinion monitoring method |
| CN113468872B (en) * | 2021-06-09 | 2024-04-16 | 大连理工大学 | Biomedical relation extraction method and system based on sentence level graph convolution |
| CN113723096A (en) * | 2021-07-23 | 2021-11-30 | 智慧芽信息科技(苏州)有限公司 | Text recognition method and device, computer-readable storage medium and electronic equipment |
| CN113643060A (en) * | 2021-08-12 | 2021-11-12 | 工银科技有限公司 | Product price prediction method and device |
| CN113946680B (en) * | 2021-10-20 | 2024-04-16 | 河南师范大学 | Online network rumor identification method based on graph embedding and information flow analysis |
| CN117407527B (en) * | 2023-10-19 | 2024-09-27 | 中诚力锘(北京)教育咨询集团有限公司 | Education field public opinion big data classification method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106326346A (en) * | 2016-08-06 | 2017-01-11 | 上海高欣计算机系统有限公司 | Text classification method and terminal device |
| CN107066560A (en) * | 2017-03-30 | 2017-08-18 | 东软集团股份有限公司 | The method and apparatus of text classification |
| CN107766577A (en) * | 2017-11-15 | 2018-03-06 | 北京百度网讯科技有限公司 | A kind of public sentiment monitoring method, device, equipment and storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101414300B (en) * | 2008-11-28 | 2010-06-16 | 电子科技大学 | Method for sorting and processing internet public feelings information |
| CN104899335A (en) * | 2015-06-25 | 2015-09-09 | 四川友联信息技术有限公司 | Method for performing sentiment classification on network public sentiment of information |
| CN107045524B (en) * | 2016-12-30 | 2019-12-27 | 中央民族大学 | Method and system for classifying network text public sentiments |
| CN107239529B (en) * | 2017-05-27 | 2020-06-09 | 中国矿业大学 | Public opinion hotspot category classification method based on deep learning |
-
2018
- 2018-04-25 CN CN201810380769.9A patent/CN108628974B/en active Active
- 2018-07-25 WO PCT/CN2018/097033 patent/WO2019205318A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106326346A (en) * | 2016-08-06 | 2017-01-11 | 上海高欣计算机系统有限公司 | Text classification method and terminal device |
| CN107066560A (en) * | 2017-03-30 | 2017-08-18 | 东软集团股份有限公司 | The method and apparatus of text classification |
| CN107766577A (en) * | 2017-11-15 | 2018-03-06 | 北京百度网讯科技有限公司 | A kind of public sentiment monitoring method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108628974A (en) | 2018-10-09 |
| WO2019205318A1 (en) | 2019-10-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108628974B (en) | Public opinion information classification method and device, computer equipment and storage medium | |
| US11308405B2 (en) | Human-computer dialogue method and apparatus | |
| CN110765265B (en) | Information classification extraction method and device, computer equipment and storage medium | |
| CN108563782B (en) | Commodity information format processing method and device, computer equipment and storage medium | |
| CN109446302B (en) | Question-answer data processing method and device based on machine learning and computer equipment | |
| US20180276525A1 (en) | Method and neural network system for human-computer interaction, and user equipment | |
| CN109960725B (en) | Text classification processing method and device based on emotion and computer equipment | |
| CN111431742B (en) | Network information detection method, device, storage medium and computer equipment | |
| KR20210024173A (en) | Related recommendation method, device, computer equipment and storage medium | |
| CN108491406B (en) | Information classification method and device, computer equipment and storage medium | |
| CN111859986A (en) | Semantic matching method, device, equipment and medium based on multitask twin network | |
| CN110888911A (en) | Sample data processing method and device, computer equipment and storage medium | |
| CN112882974A (en) | JSON data conversion method and device, computer equipment and storage medium | |
| CN111078847A (en) | Power consumer intention identification method and device, computer equipment and storage medium | |
| CN110362798B (en) | Method, apparatus, computer device and storage medium for judging information retrieval analysis | |
| CN112766319A (en) | Dialogue intention recognition model training method and device, computer equipment and medium | |
| CN113254620B (en) | Response method, device and equipment based on graph neural network and storage medium | |
| CN113761868A (en) | Text processing method and device, electronic equipment and readable storage medium | |
| CN113342927B (en) | Sensitive word recognition method, device, equipment and storage medium | |
| CN112464660B (en) | Text classification model construction method and text data processing method | |
| CN113761193A (en) | Log classification method and device, computer equipment and storage medium | |
| CN111552785A (en) | Method and device for updating database of human-computer interaction system, computer equipment and medium | |
| CN115238124A (en) | Video character retrieval method, device, equipment and storage medium | |
| CN112016297B (en) | Intention recognition model testing method and device, computer equipment and storage medium | |
| EP2846499A1 (en) | Method And Device For Classifying A Message |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |