CN108363753A - Comment text sentiment classification model is trained and sensibility classification method, device and equipment - Google Patents

Comment text sentiment classification model is trained and sensibility classification method, device and equipment Download PDF

Info

Publication number
CN108363753A
CN108363753A CN201810086816.9A CN201810086816A CN108363753A CN 108363753 A CN108363753 A CN 108363753A CN 201810086816 A CN201810086816 A CN 201810086816A CN 108363753 A CN108363753 A CN 108363753A
Authority
CN
China
Prior art keywords
text
comment
level
sentence
comment text
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.)
Granted
Application number
CN201810086816.9A
Other languages
Chinese (zh)
Other versions
CN108363753B (en
Inventor
刘天亮
王静
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Post and Telecommunication University
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing Post and Telecommunication University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing Post and Telecommunication University filed Critical Nanjing Post and Telecommunication University
Priority to CN201810086816.9A priority Critical patent/CN108363753B/en
Publication of CN108363753A publication Critical patent/CN108363753A/en
Application granted granted Critical
Publication of CN108363753B publication Critical patent/CN108363753B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis

Abstract

The invention discloses the training of comment text sentiment classification model and sensibility classification method, device and equipment, belong to the text emotion classification field of natural language processing.Model training includes：Obtain comment text, associated subject and object information；Comment subject and object attention mechanism, which is incorporated, based on first layer Bi LSTM networks extracts Sentence-level character representation；Documentation level character representation is extracted incorporating comment subject and object attention mechanism based on second layer Bi LSTM networks；Documentation level Feature Mapping is classified to emotional category space using softmax using tanh nonlinear mapping function, the parameter in model is trained, obtains optimal text emotion disaggregated model.The present invention uses the two-way Bi LSTM network models of level and attention mechanism, and the semantic steady perception of context and the semantic meaning representation of text not only may be implemented, can also significantly improve the robustness of text emotion classification, improve classification accuracy rate.

Description

Comment text sentiment classification model is trained and sensibility classification method, device and equipment
Technical field
The present invention relates to comment text emotional semantic classifications, more particularly to one kind being based on the two-way LSTM (Bidirectional of level Long Short-Term Memory, Bi-LSTM) and attention mechanism comment text emotion model training with emotional semantic classification side Method, device and equipment, belong to natural language processing technique field.
Background technology
The key problem of text emotion classification is how effectively to indicate the emotional semantic of text.With Internet technology Fast development generates a large number of users on network for valuable comment text information such as focus incident, products, for example, microblogging, Electric business platform, food and drink platform etc..These comment informations include the abundant emotional color and Sentiment orientation of people.Sentiment analysis Purpose be exactly automatically from text extraction and sorted users for the subjective emotion information of product or event, help businessman or Government department completes the tasks such as data analysis and public sentiment monitoring.Therefore, sentiment analysis also becomes the weight of natural language processing field Want one of project.Sentiment analysis is divided into the extraction of emotion information, the classification of emotion information and the retrieval of emotion information and conclusion. Here the emotional semantic classification problem of the mainly text level solved.The emotional semantic classification task of text level is primarily to automatic point Class goes out in the emotion text that user generates for a certain product or the Sentiment orientation of event representation (actively or passive) or emotion Intensity (the 1-5 stars evaluation in such as film or dining room comment text).Emotional semantic classification is seen composition notebook by current most methods One kind of classification problem.By the method for machine learning, by Sentiment orientation or emotion scoring as the data for having supervision, training Grader carries out classification to text emotion becomes a kind of method of mainstream.Character representation in machine learning is to influence grader effect An important factor for fruit, therefore, the character representation of text emotion semanteme, become the key in text emotion classification problem and time-consuming One step.
Traditional character representation method includes that One-hot, N-Gram and domain expert pass through text or additional feelings Feel some validity features of dictionary design.However, Feature Engineering is a labor-intensive task, and need more field Knowledge.Therefore, the automatic study of feature becomes the emphasis of people's research gradually.Deep learning method based on neural network is exactly A kind of method of automatic learning characteristic.And as deep learning is in computer vision, speech recognition and natural language processing etc. The successful application in field, more and more the text emotion disaggregated model generation based on deep learning, the universal land productivity of these models The method that word is embedded in (Word Embedding, WE) carries out character representation, and this low dimensional term vector representation method can not only The problem that dimension present in word expression is excessive in conventional language model is well solved, moreover, the language of word can be effectively maintained Adopted information so that semantic similar word distance closer to.In addition, on the basis of word is embedded in, pass through convolutional neural networks (Convolutional Neural Network, CNN), recurrent neural network (Recursive Neural Network, RNN) With the neural network models such as Recognition with Recurrent Neural Network (Recurrent Neural Network, RNN), can indicate well sentence or The semantic information of person's text level, since deep learning has the automatic Extracting Ability of good feature, in this paper emotions point It is widely used in class problem.
However, most of text emotion disaggregated models based on neural network only considered the relevant feelings of content of text at present Sense is semantic, has ignored and comments on object information described in the relevant comment main information of text and content of text.Meanwhile having Studies have shown that the hobby of comment main body has important influence with the characteristics of comment object for the scoring for commenting on main body.It realizes The semantic steady perception of the context of text and semantic meaning representation, while merging attention mechanism will comment main body and comment object information It is combined with text semantic information so that semantic information is more abundant in text, is the main direction of studying of the present invention.
Invention content
Goal of the invention：In view of the deficiencies of the prior art, a kind of based on hierarchical B i-LSTM and note present invention aims at providing The comment text sentiment classification model training for power mechanism of anticipating and the method for emotional semantic classification is carried out based on the model and device and is set It is standby, Sentence-level and documentation level semantic feature are modeled by using two-way LSTM, it is special above to introduce forward direction LSTM elements captures Reference ceases, reversed LSTM elements captures following traits information, and will comment on main body and comment object information and believe with text semantic Breath combines, and captures abundant semantic feature information, improves the robustness and accuracy of text emotion classification.
Technical solution：For achieving the above object, the present invention uses following technical scheme：
A kind of comment text sentiment classification model training method, includes the following steps：
(1) obtain includes comment text, the training set text with the associated subject and object information of the comment text；
(2) word in the comment text of training set is transformed into term vector to indicate, is input to first layer Bi-LSTM networks In, in conjunction with front and back the associated subject and object information attention mechanism of word level comment text is multiplied by hidden layer output vector The weight trained extracts Sentence-level character representation；
(3) Sentence-level character representation is input in second layer Bi-LSTM networks, in conjunction with front and back to hidden layer output vector It is multiplied by the weight that the associated subject and object information attention mechanism of Sentence-level comment text trains, extracts the text of comment text Shelf grade character representation；
(4) use tanh nonlinear mapping function that documentation level semantic feature to obtain the final product is mapped to emotional category sky Between, classified using softmax, the parameter in model is trained, obtains optimal text emotion disaggregated model.
Preferably, in step (2), the Sentence-level language based on i-th of sentence that word level attention mechanism is extracted Adopted featureWherein,Indicate j-th of word in i-th of sentenceSignificance level, T is word in i-th of sentence Quantity,For the output of the forward direction LSTM models in first layer Bi-LSTM networksIt is defeated with backward LSTM models Go outCombination；Wherein, FormulaIt indicatesMain text information u and comment object text message p are commented on to wordThe calculating of significance level, Wherein parameter matrix WH、WU、WPIt continues to optimize to obtain optimal value in the training process with vector v, b.
Preferably, in step (3), the documentation level feature extracted based on Sentence-level attention mechanism Wherein, βiIndicate i-th of sentence y in documentiSignificance level, M is the quantity of sentence in document,For the second layer The output of forward direction LSTM models in Bi-LSTM networksWith the output of backward LSTM modelsCombination；Wherein, e (hi, u, p) and=vTtanh(WHhi+WUu+WPP+b), formula e (hi, u, p) and it indicates hi, comment main text information u and comment object text message p to sentence yiThe calculating of significance level, wherein parameter matrix WH、 WU、WPIt continues to optimize to obtain optimal value in the training process with vector v, b.
Preferably, step (4) includes：
(4.1) documentation level semantic feature d to obtain the final product is mapped to by emotion using tanh tanh nonlinear mapping functions Class number is the emotional category space of C,Wherein WcIt is the parameter matrix of file characteristics d, bcIt is inclined Difference vector, parameter matrix and bias vector obtain optimal value by training process optimization；
(4.2) adjudicate to obtain text emotion classification using softmax grader decisions, using cross entropy loss function as The optimization aim of model training constantly updates related ginseng in model by backpropagation BP algorithm counting loss functional gradient simultaneously Number obtains optimal models.
Preferably, the calculation formula for adjudicating to obtain text emotion classification using softmax grader decisions is：
Wherein,Be document semantic Feature Mapping it is the expression that emotional category is c, pcIt is the prediction that text emotion classification is c Probability.
Preferably, the calculation formula of the cross entropy loss function is：
Wherein, training setbiIt is i-th of sample text in training set, yiIt is sample biBenchmark Affective tag, N are training set sizes；I.e. if text emotion is classified as c,Value be 1, otherwiseValue be 0；pcIt is the prediction probability that text emotion classification is c.
A kind of text obtained using above-mentioned comment text sentiment classification model training method that another aspect of the present invention provides The method that this sentiment classification model carries out text emotion classification, includes the following steps：
Obtain includes comment text, the test set text with the associated subject and object information of the comment text；
It will be predicted to obtain emotional category in test set text input to the optimal text emotion disaggregated model.
A kind of comment text sentiment classification model training device that another aspect of the present invention provides, including：
Acquisition module, it includes comment text, the instruction with the associated subject and object information of the comment text to be used to obtain Practice collection text；
Sentence-level characteristic extracting module, for word in the comment text of training set text to be transformed into term vector expression, It is input in first layer Bi-LSTM networks, to hidden state output vector to be multiplied by word level comment text associated in conjunction with front and back The weight that subject and object information attention mechanism trains extracts Sentence-level character representation；
Documentation level characteristic extracting module, for Sentence-level character representation to be input in second layer Bi-LSTM networks, in conjunction with It is front and back to be multiplied by what the associated subject and object information attention mechanism of Sentence-level comment text trained to hidden state output vector Weight extracts the documentation level character representation of comment text；
And training module, for documentation level semantic feature to obtain the final product to be reflected using tanh nonlinear mapping function It is incident upon emotional category space, is classified using softmax, the parameter in model is trained, obtains optimal text emotion classification Model.
A kind of text obtained using above-mentioned comment text sentiment classification model training method that another aspect of the present invention provides This sentiment classification model carries out the device of text emotion classification, including：
Acquisition module, it includes comment text, the survey with the associated subject and object information of the comment text to be used to obtain Examination collection text；
And classification prediction module, for by test set text input to the optimal text emotion disaggregated model into Row prediction obtains emotional category.
The present invention also provides a kind of computer equipment, including memory, processor and storage on a memory and can located The computer program run on reason device, the processor realize the comment text sentiment classification model when executing described program Training method realizes the text sentiment classification method.
Advantageous effect：Compared with prior art, the present invention has the following technical effects：
The present invention is based on the documentation level text sentiment classification method of hierarchical B i-LSTM and attention mechanism, using two-way LSTM models the semantic feature of Sentence-level and text level.Positive LSTM captures semantic feature information above, reversed LSTM The semantic feature information of capture hereafter, while incorporating comment text host-guest relevant information and reinforcing text semantic expression, effect The semantic steady perception of context and the semantic meaning representation of text not only may be implemented in fruit, can also significantly improve text emotion classification Robustness, while the accuracy of text emotion classification can also be effectively improved.
Description of the drawings
Fig. 1 is the flow chart of comment text sentiment classification model training method in the embodiment of the present invention.
Fig. 2 is the hierarchical B i-LSTM model schematics proposed in the embodiment of the present invention.
Fig. 3 is the training of comment text sentiment classification model and sensibility classification method principle schematic in the embodiment of the present invention.
Fig. 4 is comment text sensibility classification method flow chart in the embodiment of the present invention.
Fig. 5 is comment text sentiment classification model training device structure chart in the embodiment of the present invention.
Fig. 6 is comment text emotional semantic classification structure drawing of device in the embodiment of the present invention.
Specific implementation mode
Technical scheme of the present invention is described in detail below in conjunction with the accompanying drawings：
As shown in Figure 1, a kind of comment text sentiment classification model training method disclosed by the embodiments of the present invention, includes mainly Following steps：
(1) training set text is obtained, each sample text wherein in training set includes comment text itself, and is commented with this The associated comment main body of paper sheet (i.e. commentator or linked groups) and comment object (objects such as comment product, news).Comment Text and relative subject and the information of object can be obtained from internet.
(2) word in the comment text of training set is transformed into term vector to indicate, is input to first layer Bi-LSTM networks In, in conjunction with front and back the associated subject and object information attention mechanism of word level comment text is multiplied by hidden layer output vector The weight trained extracts Sentence-level character representation.
(3) Sentence-level character representation is input in second layer Bi-LSTM networks, in conjunction with front and back to hidden layer output vector It is multiplied by the weight that the associated subject and object information attention mechanism of Sentence-level comment text trains, extracts the text of comment text Shelf grade character representation.
In step (2), each respective word in comment text is converted using word embedded technology (Word Embedding, WE) It indicates, and is input in Bi-LSTM networks to extract word level feature representation at feature vector.Word embedded technology will be in text Word is mapped to for the size of vocabulary compared with the real vector on lower dimensional space.In the semantic feature table of word level In showing, all words are all made of a word embeded matrix Lw∈Rn×|V|It indicating, wherein n indicates the dimension of word, | V | it is all lists The quantity of word.LwIn term vectorIt is generated by word2vector word embedding grammar pre-training.
In the first constructed level Bi-LSTM models, two-way LSTM (Bi-LSTM) networks introduce just simultaneously in hidden layer To LSTM units and reversed LSTM units；Positive LSTM elements captures characteristic information above, and reversed LSTM elements captures are hereafter Characteristic information；Compared with unidirectional LSTM, the characteristic information of more robusts can be captured.According to obtaining the word level mark sheet of text Show, the Sentence-level character representation of input Bi-LSTM networks output extraction text；Second Bi- is inputted using similar strategy cascade LSTM networks obtain the documentation level character representation of text by the Sentence-level character representation output extraction of text and are based on two layers with realizing The hierarchical B i-LSTM model constructions of Bi-LSTM network aware text document grade features.
It will be in the word level character representation input first layer Bi-LSTM networks of text, it is assumed that a sentence yiIn have T Word, j-th of word areBy sentence yiRegard a sequence as, the word in sentence is the component part of sentence sequence.Point The output of the forward direction LSTM and backward LSTM models in the Bi-LSTM networks are indescribably taken, and the Sentence-level for being considered as corresponding text is special Sign indicates：
In conjunction withWithForm sentence yiSemantic feature：
Sentence-level character representation to obtain the final product is inputted in second layer Bi-LSTM networks in step (3), similarly, it is assumed that one It is y to have M sentence, each sentence in text bi, i ∈ [0, M] regard text b as a sequence, and the sentence in text is text The component part of sequence.The forward direction LSTM in the Bi-LSTM networks and the output of backward LSTM models are extracted respectively, and are considered as corresponding The documentation level character representation of text：
In conjunction withWithObtain the semantic feature of final text b：
More accurate comment text emotional semantic indicates in order to obtain, will descriptive text related with comment subject and object This information trains word level and the weight of the hidden state output of Sentence-level using attention mechanism, in the semanteme for obtaining comment text When grade character representation, selection and the higher word of the comment subject and object degree of correlation or sentence are more literary to obtain This emotional semantic indicates.
Further being strengthened using word level attention mechanism (word attention) in step (2) is had with emotional factor Associated Sentence-level character representation assigns the power of semantic more relevant word bigger in conjunction with comment subject and object preference information Weight.Assuming that using vectorWithIt is indicated respectively with the related text message of comment main body and related with object is commented on Text message, wherein duAnd dpThe dimension with comment main body and the term vector for commenting on object related text is indicated respectively.Such as Fig. 2 It is shown, in Sentence-level character representation, it is assumed that siIndicate that the Sentence-level extracted based on word level attention mechanism is semantic special Sign, then
WhereinIndicate the significance level of j-th of word in i-th of sentence.DefinitionCalculating：
Wherein, e is to calculate wordImportance degree function：
Wherein, WH、WUAnd WPHidden layer output is indicated respectivelyComment on main text information u and comment visitor The relevant parameter matrix of body text message p, v are portrayed wordsImportance degree parameter vector, vTIt is the transposed vector of v, b It is bias vector, wherein parameter matrix WH、WU、WPIt continues to optimize to obtain optimal value in the training process with vector v, b.
Using the further reinforcing of Sentence-level attention mechanism (sentence attention) and emotional factor in step (3) Related documentation level character representation assigns the power of semantic more relevant sentence bigger in conjunction with comment subject and object preference information Weight.In documentation level character representation, it is assumed that d is the documentation level character representation extracted based on Sentence-level attention mechanism, that ：
Wherein, βiIndicate i-th of sentence y in documentiSignificance level, computational methods withUnanimously：Wherein, e (hi, u, p) and=vT tanh(WHhi+WUu+WPP+b), formula e (hi, u, p) and it indicatesMain text information u and comment object text message p are commented on to sentence yiThe calculating of significance level corresponds to Parameter is respectively WH、WUAnd WP, v is to portray sentence yiImportance degree parameter vector, vTIt is the transposed vector of v, in formula All parameters are continued to optimize to obtain optimal value in the training process with bias vector.βiIt is by softmax to e (hi, u, p) into Row normalization indicates.Then to get d be exactly final text semantic character representation.
To sum up, the step of Text character extraction is：Related words in text are used into word embedded technology (such as first Word2vector tools) it is converted into real vector expression, as first layer Bi-LSTM network inputs, and extract the Bi-LSTM The front and back output vector to hidden layer of networkMultiplied by with by word level attention mechanism (word Attention the weight) trainedObtain the Sentence-level character representation of corresponding text：Then again by the sentence Sub- grade feature siAs the input of second layer Bi-LSTM networks, and extract the Bi-LSTM networks it is front and back to hidden layer export to AmountMultiplied by the weight beta to be trained by Sentence-level attention mechanism (sentence attention) mechanismi, obtain Go out the documentation level character representation of corresponding text：
(4) use tanh nonlinear mapping function that documentation level semantic feature to obtain the final product is mapped to emotional category sky Between, classified using softmax, the parameter in model is trained, obtains optimal text emotion disaggregated model.
Step (3) obtains the documentation level character representation d of whole text, can be by it directly as the feature of text classifier Input.First, documentation level semantic feature d to obtain the final product is mapped to by feelings using tanh tanh nonlinear mapping functions Feel the emotional category space that class number is C, calculation formula：
Wherein, WcIt is the parameter matrix of file characteristics d, bcIt is bias vector,Then, classified using softmax Device decision adjudicates to obtain text emotion classification, calculation formula：
Wherein,Be document semantic Feature Mapping it is the expression that emotional category is c, pcBe text emotion classification be c it is pre- Survey probability.Optimization aim used here as cross entropy loss function as model training is calculated by backpropagation BP algorithm and is damaged It loses functional gradient and updates model parameter simultaneously, loss function calculates formula：
Wherein, training datasetbiIt is the i-th sample text that training data is concentrated, yiIt is sample bi's The affective tag of benchmark, N are training set sizes.I.e. if text emotion is classified as c,Value be 1, otherwiseValue be 0.
Disaggregated model is built according to above step, by the training set text input in data set to the model built In, (AdaDelta) method Optimized model parameter is adjusted using autoadapted learning rate and adjusting model parameter obtains on verification collection Optimal model parameters, to obtain optimal text emotion disaggregated model.Model training and the detailed schematic diagram of classification are as shown in Figure 3
As shown in figure 4, a kind of comment text sentiment classification model training device disclosed by the embodiments of the present invention, including：It obtains Modulus block, it includes comment text, the training set text with the associated subject and object information of the comment text to be used to obtain；Sentence Sub- grade characteristic extracting module indicates for word in the comment text of training set text to be transformed into term vector, is input to first In layer Bi-LSTM networks, the associated subject and object of word level comment text is multiplied by hidden state output vector in conjunction with front and back The weight that information attention mechanism trains extracts Sentence-level character representation；Documentation level characteristic extracting module is used for Sentence-level Character representation is input in second layer Bi-LSTM networks, and Sentence-level comment text is multiplied by hidden state output vector in conjunction with front and back The weight that associated subject and object information attention mechanism trains, extracts the documentation level character representation of comment text；And Training module, for documentation level semantic feature to obtain the final product to be mapped to emotional category sky using tanh nonlinear mapping function Between, classified using softmax, the parameter in model is trained, obtains optimal text emotion disaggregated model.
As shown in figure 5, a kind of comment text emotional semantic classification device disclosed by the embodiments of the present invention, including：Acquisition module is used In obtain include comment text, and the associated subject and object information of the comment text test set text；And classification is pre- Module is surveyed, for being predicted test set text input to the optimal text emotion disaggregated model to obtain emotional category.
Above-mentioned comment text sentiment classification model training device embodiment can be used for executing above-mentioned comment text emotion point Class model training method embodiment, comment text emotional semantic classification device embodiment can be used for executing above-mentioned comment text emotion point Class embodiment of the method, technical principle, it is solved the technical issues of and generation technique effect it is similar, the comment of foregoing description text The specific work process and related explanation of this emotional semantic classification and model training can refer to pair in aforementioned correlation method embodiment Process is answered, details are not described herein.
It will be understood by those skilled in the art that can carry out adaptively changing and it to the module in embodiment Be arranged in the one or more equipment different from the embodiment.Can in embodiment module or unit or component combine At a module or unit or component, and it can be divided into multiple submodule or subelement or sub-component in addition.
Based on technical concept identical with embodiment of the method, the embodiment of the present invention additionally provides a kind of computer equipment, should Computer equipment may include memory, processor and storage on a memory and the computer journey that can run on a processor Sequence.Wherein, above-mentioned comment text sensibility classification method embodiment or comment text are realized when computer program is loaded on processor Each step in this sentiment classification model training method embodiment.
Above example is merely illustrative of the invention's technical idea, and protection scope of the present invention cannot be limited with this, every According to technological thought proposed by the present invention, any change done on the basis of technical solution each falls within the scope of the present invention Within.

Claims (10)

1. a kind of comment text sentiment classification model training method, which is characterized in that include the following steps：
(1) obtain includes comment text, the training set text with the associated subject and object information of the comment text；
(2) word in the comment text of training set is transformed into term vector to indicate, is input in first layer Bi-LSTM networks, then In conjunction with front and back the associated subject and object information attention mechanism training of word level comment text is multiplied by hidden layer output vector The weight gone out extracts Sentence-level character representation；
(3) Sentence-level character representation is input in second layer Bi-LSTM networks, is multiplied by hidden layer output vector in conjunction with front and back The weight that the associated subject and object information attention mechanism of Sentence-level comment text trains, extracts the documentation level of comment text Character representation；
(4) documentation level semantic feature to obtain the final product is mapped to by emotional category space using tanh nonlinear mapping function, adopted Classified with softmax, the parameter in model is trained, obtains optimal text emotion disaggregated model.
2. comment text sentiment classification model training method according to claim 1, which is characterized in that in step (2), base In the Sentence-level semantic feature for i-th of sentence that word level attention mechanism is extractedWherein,Indicate the J-th of word in i sentenceSignificance level, T is the quantity of word in i-th of sentence,For first layer Bi- The output of forward direction LSTM models in LSTM networksWith the output of backward LSTM modelsCombination；Wherein,Formula It indicatesMain text information u and comment object text message p are commented on to wordThe calculating of significance level, wherein parameter matrix WH、WU、WPIt continues to optimize to obtain optimal value in the training process with vector v, b.
3. comment text sentiment classification model training method according to claim 1, which is characterized in that in step (3), base In the documentation level feature that Sentence-level attention mechanism is extractedWherein, βiIndicate i-th of sentence y in documenti's Significance level, M are the quantity of sentence in document,For the defeated of the forward direction LSTM models in second layer Bi-LSTM networks Go outWith the output of backward LSTM modelsCombination；Wherein, e (hi, u, p) and=vTtanh (WHhi+WUu+WPP+b), formula e (hi, u, p) and indicate hi, comment main text information u and comment object text message p to sentence yiThe calculating of significance level, wherein parameter matrix WH、WU、WPIt continues to optimize to obtain optimal value in the training process with vector v, b.
4. comment text sentiment classification model training method according to claim 1, which is characterized in that packet in step (4) It includes：
(4.1) documentation level semantic feature d to obtain the final product is mapped to by emotional category using tanh tanh nonlinear mapping functions Number is the emotional category space of C,Wherein WcIt is the parameter matrix of file characteristics d, bcDeviation to Amount, parameter matrix and bias vector obtain optimal value by training process optimization；
(4.2) it adjudicates to obtain text emotion classification using softmax grader decisions, using cross entropy loss function as model Trained optimization aim is constantly updated relevant parameter in model simultaneously by backpropagation BP algorithm counting loss functional gradient and is obtained To optimal models.
5. comment text sentiment classification model training method according to claim 4, which is characterized in that utilize softmax The calculation formula that grader decision adjudicates to obtain text emotion classification is：
Wherein,Be document semantic Feature Mapping it is the expression that emotional category is c, pcIt is that the prediction that text emotion classification is c is general Rate.
6. comment text sentiment classification model training method according to claim 5, which is characterized in that the cross entropy damage Lose function calculation formula be：
Wherein, training setbiIt is i-th of sample text in training set, yiIt is sample biBenchmark emotion Label, N are training set sizes；I.e. if text emotion is classified as c,Value be 1, otherwise Value be 0；pcIt is the prediction probability that text emotion classification is c.
7. a kind of text emotion obtained using comment text sentiment classification model training method according to claim 1 point The method that class model carries out text emotion classification, which is characterized in that include the following steps：
Obtain includes comment text, the test set text with the associated subject and object information of the comment text；
It will be predicted to obtain emotional category in test set text input to the optimal text emotion disaggregated model.
8. a kind of comment text sentiment classification model training device, which is characterized in that including：
Acquisition module, it includes comment text, the training set with the associated subject and object information of the comment text to be used to obtain Text；
Sentence-level characteristic extracting module indicates, input for word in the comment text of training set text to be transformed into term vector Into first layer Bi-LSTM networks, the associated main body of word level comment text is multiplied by hidden state output vector in conjunction with front and back The weight trained with object information attention mechanism extracts Sentence-level character representation；
Documentation level characteristic extracting module, for Sentence-level character representation to be input in second layer Bi-LSTM networks, in conjunction with front and back It is multiplied by the weight that the associated subject and object information attention mechanism of Sentence-level comment text trains to hidden state output vector, Extract the documentation level character representation of comment text；
And training module, for documentation level semantic feature to obtain the final product to be mapped to using tanh nonlinear mapping function Emotional category space, is classified using softmax, is trained to the parameter in model, and optimal text emotion disaggregated model is obtained.
9. a kind of text emotion obtained using comment text sentiment classification model training method according to claim 1 point Class model carries out the device of text emotion classification, which is characterized in that including：
Acquisition module, it includes comment text, the test set with the associated subject and object information of the comment text to be used to obtain Text；
And classification prediction module, it is pre- for carrying out test set text input to the optimal text emotion disaggregated model Measure emotional category.
10. a kind of computer equipment, including memory, processor and storage are on a memory and the meter that can run on a processor Calculation machine program, which is characterized in that the processor realizes the comment text emotion described in claim 1-6 when executing described program Text sentiment classification method described in disaggregated model training method or realization claim 7.
CN201810086816.9A 2018-01-30 2018-01-30 Comment text emotion classification model training and emotion classification method, device and equipment Active CN108363753B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810086816.9A CN108363753B (en) 2018-01-30 2018-01-30 Comment text emotion classification model training and emotion classification method, device and equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810086816.9A CN108363753B (en) 2018-01-30 2018-01-30 Comment text emotion classification model training and emotion classification method, device and equipment

Publications (2)

Publication Number Publication Date
CN108363753A true CN108363753A (en) 2018-08-03
CN108363753B CN108363753B (en) 2020-05-19

Family

ID=63007586

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810086816.9A Active CN108363753B (en) 2018-01-30 2018-01-30 Comment text emotion classification model training and emotion classification method, device and equipment

Country Status (1)

Country Link
CN (1) CN108363753B (en)

Cited By (81)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109145112A (en) * 2018-08-06 2019-01-04 北京航空航天大学 A kind of comment on commodity classification method based on global information attention mechanism
CN109241283A (en) * 2018-08-08 2019-01-18 广东工业大学 A kind of file classification method based on multi-angle capsule network
CN109284506A (en) * 2018-11-29 2019-01-29 重庆邮电大学 A kind of user comment sentiment analysis system and method based on attention convolutional neural networks
CN109472470A (en) * 2018-10-23 2019-03-15 重庆誉存大数据科技有限公司 In conjunction with the corporate news data classification of risks method of deep learning and logic rules
CN109492229A (en) * 2018-11-23 2019-03-19 中国科学技术大学 A kind of cross-cutting sensibility classification method and relevant apparatus
CN109508375A (en) * 2018-11-19 2019-03-22 重庆邮电大学 A kind of social affective classification method based on multi-modal fusion
CN109543030A (en) * 2018-10-12 2019-03-29 平安科技（深圳）有限公司 Customer service machine conference file classification method and device, equipment, storage medium
CN109558585A (en) * 2018-10-26 2019-04-02 深圳点猫科技有限公司 A kind of answer Automatic-searching method and electronic equipment based on educational system
CN109597997A (en) * 2018-12-07 2019-04-09 上海宏原信息科技有限公司 Based on comment entity, aspect grade sensibility classification method and device and its model training
CN109597995A (en) * 2018-12-04 2019-04-09 国网江西省电力有限公司信息通信分公司 A kind of document representation method based on BM25 weighted combination term vector
CN109670542A (en) * 2018-12-11 2019-04-23 田刚 A kind of false comment detection method based on comment external information
CN109670169A (en) * 2018-11-16 2019-04-23 中山大学 A kind of deep learning sensibility classification method based on feature extraction
CN109784280A (en) * 2019-01-18 2019-05-21 江南大学 Human bodys' response method based on Bi-LSTM-Attention model
CN109858034A (en) * 2019-02-25 2019-06-07 武汉大学 A kind of text sentiment classification method based on attention model and sentiment dictionary
CN109918501A (en) * 2019-01-18 2019-06-21 平安科技（深圳）有限公司 Method, apparatus, equipment and the storage medium of news article classification
CN109992780A (en) * 2019-03-29 2019-07-09 哈尔滨理工大学 One kind being based on deep neural network specific objective sensibility classification method
CN110032645A (en) * 2019-04-17 2019-07-19 携程旅游信息技术（上海）有限公司 Text emotion recognition methods, system, equipment and medium
CN110046353A (en) * 2019-04-22 2019-07-23 重庆理工大学 Aspect level emotion analysis method based on multi-language level mechanism
CN110083702A (en) * 2019-04-15 2019-08-02 中国科学院深圳先进技术研究院 A kind of aspect rank text emotion conversion method based on multi-task learning
CN110134966A (en) * 2019-05-21 2019-08-16 中电健康云科技有限公司 A kind of sensitive information determines method and device
CN110175237A (en) * 2019-05-14 2019-08-27 华东师范大学 It is a kind of towards multi-class secondary sensibility classification method
CN110188202A (en) * 2019-06-06 2019-08-30 北京百度网讯科技有限公司 Training method, device and the terminal of semantic relation identification model
CN110222178A (en) * 2019-05-24 2019-09-10 新华三大数据技术有限公司 Text sentiment classification method, device, electronic equipment and readable storage medium storing program for executing
CN110263165A (en) * 2019-06-14 2019-09-20 中山大学 A kind of user comment sentiment analysis method based on semi-supervised learning
CN110309503A (en) * 2019-05-21 2019-10-08 昆明理工大学 A kind of subjective item Rating Model and methods of marking based on deep learning BERT--CNN
CN110348016A (en) * 2019-07-15 2019-10-18 昆明理工大学 Text snippet generation method based on sentence association attention mechanism
CN110399547A (en) * 2018-04-17 2019-11-01 百度在线网络技术（北京）有限公司 For updating the method, apparatus, equipment and storage medium of model parameter
CN110517121A (en) * 2019-09-23 2019-11-29 重庆邮电大学 Method of Commodity Recommendation and the device for recommending the commodity based on comment text sentiment analysis
CN110598786A (en) * 2019-09-09 2019-12-20 京东方科技集团股份有限公司 Neural network training method, semantic classification method and semantic classification device
CN110600047A (en) * 2019-09-17 2019-12-20 南京邮电大学 Perceptual STARGAN-based many-to-many speaker conversion method
CN110751208A (en) * 2018-10-29 2020-02-04 山东大学 Criminal emotion recognition method for multi-mode feature fusion based on self-weight differential encoder
CN110765769A (en) * 2019-08-27 2020-02-07 电子科技大学 Entity attribute dependency emotion analysis method based on clause characteristics
CN110826340A (en) * 2019-11-06 2020-02-21 广东三维家信息科技有限公司 Evaluation text generation method and device and electronic equipment
CN110851601A (en) * 2019-11-08 2020-02-28 福州大学 Cross-domain emotion classification system and method based on layered attention mechanism
CN110879938A (en) * 2019-11-14 2020-03-13 中国联合网络通信集团有限公司 Text emotion classification method, device, equipment and storage medium
CN111046233A (en) * 2019-12-24 2020-04-21 浙江大学 Video label determination method based on video comment text
CN111078833A (en) * 2019-12-03 2020-04-28 哈尔滨工程大学 Text classification method based on neural network
CN111125354A (en) * 2018-10-31 2020-05-08 北京国双科技有限公司 Text classification method and device
CN111159400A (en) * 2019-12-19 2020-05-15 苏州大学 Product comment emotion classification method and system
CN111241842A (en) * 2018-11-27 2020-06-05 阿里巴巴集团控股有限公司 Text analysis method, device and system
CN111275521A (en) * 2020-01-16 2020-06-12 华南理工大学 Commodity recommendation method based on user comment and satisfaction level embedding
CN111291189A (en) * 2020-03-10 2020-06-16 北京芯盾时代科技有限公司 Text processing method and device and computer readable storage medium
CN111325027A (en) * 2020-02-19 2020-06-23 东南大学 Sparse data-oriented personalized emotion analysis method and device
CN111339440A (en) * 2020-02-19 2020-06-26 东南大学 Social emotion ordering method for news text based on hierarchical state neural network
CN111368082A (en) * 2020-03-03 2020-07-03 南京信息工程大学 Emotion analysis method for domain adaptive word embedding based on hierarchical network
CN111414122A (en) * 2019-12-26 2020-07-14 腾讯科技（深圳）有限公司 Intelligent text processing method and device, electronic equipment and storage medium
CN111522956A (en) * 2020-05-08 2020-08-11 河南理工大学 Text emotion classification method based on double channels and hierarchical attention network
CN111639183A (en) * 2020-05-19 2020-09-08 民生科技有限责任公司 Financial industry consensus public opinion analysis method and system based on deep learning algorithm
CN111858944A (en) * 2020-07-31 2020-10-30 电子科技大学 Entity aspect level emotion analysis method based on attention mechanism
CN111859978A (en) * 2020-06-11 2020-10-30 南京邮电大学 Emotion text generation method based on deep learning
CN111984931A (en) * 2020-08-20 2020-11-24 上海大学 Public opinion calculation and deduction method and system for social event web text
CN111985214A (en) * 2020-08-19 2020-11-24 四川长虹电器股份有限公司 Human-computer interaction negative emotion analysis method based on bilstm and attention
CN112347269A (en) * 2020-11-11 2021-02-09 重庆邮电大学 Method for recognizing argument pairs based on BERT and Att-BilSTM
CN112434516A (en) * 2020-12-18 2021-03-02 安徽商信政通信息技术股份有限公司 Self-adaptive comment emotion analysis system and method fusing text information
CN112597302A (en) * 2020-12-18 2021-04-02 东北林业大学 False comment detection method based on multi-dimensional comment representation
CN112685558A (en) * 2019-10-18 2021-04-20 普天信息技术有限公司 Emotion classification model training method and device
CN112699244A (en) * 2021-03-16 2021-04-23 成都信息工程大学 Deep learning-based method and system for classifying defect texts of power transmission and transformation equipment
CN112905739A (en) * 2021-02-05 2021-06-04 北京邮电大学 False comment detection model training method, detection method and electronic equipment
CN112950019A (en) * 2021-03-01 2021-06-11 昆明电力交易中心有限责任公司 Electricity selling company evaluation emotion classification method based on combined attention mechanism
CN113051367A (en) * 2021-03-22 2021-06-29 北京智慧星光信息技术有限公司 Deep learning early warning method and system based on semantic feature enhancement and electronic equipment
CN113065577A (en) * 2021-03-09 2021-07-02 北京工业大学 Multi-modal emotion classification method for targets
CN113159831A (en) * 2021-03-24 2021-07-23 湖南大学 Comment text sentiment analysis method based on improved capsule network
CN113255360A (en) * 2021-04-19 2021-08-13 国家计算机网络与信息安全管理中心 Document rating method and device based on hierarchical self-attention network
CN113297383A (en) * 2021-06-22 2021-08-24 苏州大学 Knowledge distillation-based speech emotion classification method
CN113297838A (en) * 2021-05-21 2021-08-24 华中科技大学鄂州工业技术研究院 Relationship extraction method based on graph neural network
CN113361617A (en) * 2021-06-15 2021-09-07 西南交通大学 Aspect level emotion analysis modeling method based on multivariate attention correction
WO2021174824A1 (en) * 2020-03-05 2021-09-10 苏州浪潮智能科技有限公司 Sentence-level convolution lstm training method, and device and readable medium
CN113407721A (en) * 2021-06-29 2021-09-17 哈尔滨工业大学（深圳） Method, device and computer storage medium for detecting log sequence abnormity
CN113553828A (en) * 2021-07-21 2021-10-26 南京邮电大学 Semantic code-based hierarchical level remote supervision relation extraction method
CN113762343A (en) * 2021-08-04 2021-12-07 德邦证券股份有限公司 Method, device and storage medium for processing public opinion information and training classification model
CN113779249A (en) * 2021-08-31 2021-12-10 华南师范大学 Cross-domain text emotion classification method and device, storage medium and electronic equipment
CN113806545A (en) * 2021-09-24 2021-12-17 重庆理工大学 Comment text emotion classification method based on label description generation
CN113849651A (en) * 2021-09-28 2021-12-28 平安科技（深圳）有限公司 Document-level emotional tendency-based emotion classification method, device, equipment and medium
CN113869065A (en) * 2021-10-15 2021-12-31 梧州学院 Emotion classification method and system based on 'word-phrase' attention mechanism
US20220223144A1 (en) * 2019-05-14 2022-07-14 Dolby Laboratories Licensing Corporation Method and apparatus for speech source separation based on a convolutional neural network
CN114791951A (en) * 2022-05-13 2022-07-26 青岛文达通科技股份有限公司 Emotion classification method and system based on capsule network
CN115544260A (en) * 2022-12-05 2022-12-30 湖南工商大学 Comparison optimization coding and decoding model and method for text emotion analysis
WO2023069017A3 (en) * 2021-10-19 2023-06-01 Grabtaxi Holdings Pte. Ltd. System and method for recognizing sentiment of user's feedback
CN116484005A (en) * 2023-06-25 2023-07-25 北京中关村科金技术有限公司 Classification model construction method, device and storage medium
WO2023204759A1 (en) * 2022-04-22 2023-10-26 Lemon Inc. Attribute and rating co-extraction
CN113869065B (en) * 2021-10-15 2024-04-12 梧州学院 Emotion classification method and system based on 'word-phrase' attention mechanism

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103678607A (en) * 2013-12-16 2014-03-26 合肥工业大学 Building method of emotion marking system
CN106570179A (en) * 2016-11-10 2017-04-19 中国科学院信息工程研究所 Evaluative text-oriented kernel entity identification method and apparatus
CN107133211A (en) * 2017-04-26 2017-09-05 中国人民大学 A kind of composition methods of marking based on notice mechanism
CN107316654A (en) * 2017-07-24 2017-11-03 湖南大学 Emotion identification method based on DIS NV features
CN107451118A (en) * 2017-07-21 2017-12-08 西安电子科技大学 Sentence-level sensibility classification method based on Weakly supervised deep learning
CN107544957A (en) * 2017-07-05 2018-01-05 华北电力大学 A kind of Sentiment orientation analysis method of business product target word
CN107590134A (en) * 2017-10-26 2018-01-16 福建亿榕信息技术有限公司 Text sentiment classification method, storage medium and computer

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103678607A (en) * 2013-12-16 2014-03-26 合肥工业大学 Building method of emotion marking system
CN106570179A (en) * 2016-11-10 2017-04-19 中国科学院信息工程研究所 Evaluative text-oriented kernel entity identification method and apparatus
CN107133211A (en) * 2017-04-26 2017-09-05 中国人民大学 A kind of composition methods of marking based on notice mechanism
CN107544957A (en) * 2017-07-05 2018-01-05 华北电力大学 A kind of Sentiment orientation analysis method of business product target word
CN107451118A (en) * 2017-07-21 2017-12-08 西安电子科技大学 Sentence-level sensibility classification method based on Weakly supervised deep learning
CN107316654A (en) * 2017-07-24 2017-11-03 湖南大学 Emotion identification method based on DIS NV features
CN107590134A (en) * 2017-10-26 2018-01-16 福建亿榕信息技术有限公司 Text sentiment classification method, storage medium and computer

Cited By (116)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110399547A (en) * 2018-04-17 2019-11-01 百度在线网络技术（北京）有限公司 For updating the method, apparatus, equipment and storage medium of model parameter
CN109145112B (en) * 2018-08-06 2021-08-06 北京航空航天大学 Commodity comment classification method based on global information attention mechanism
CN109145112A (en) * 2018-08-06 2019-01-04 北京航空航天大学 A kind of comment on commodity classification method based on global information attention mechanism
CN109241283A (en) * 2018-08-08 2019-01-18 广东工业大学 A kind of file classification method based on multi-angle capsule network
CN109543030A (en) * 2018-10-12 2019-03-29 平安科技（深圳）有限公司 Customer service machine conference file classification method and device, equipment, storage medium
CN109543030B (en) * 2018-10-12 2023-04-07 平安科技（深圳）有限公司 Method, device, equipment and storage medium for classifying session texts of customer service robot
CN109472470A (en) * 2018-10-23 2019-03-15 重庆誉存大数据科技有限公司 In conjunction with the corporate news data classification of risks method of deep learning and logic rules
CN109558585A (en) * 2018-10-26 2019-04-02 深圳点猫科技有限公司 A kind of answer Automatic-searching method and electronic equipment based on educational system
CN110751208B (en) * 2018-10-29 2020-06-30 山东大学 Criminal emotion recognition method for multi-mode feature fusion based on self-weight differential encoder
CN110751208A (en) * 2018-10-29 2020-02-04 山东大学 Criminal emotion recognition method for multi-mode feature fusion based on self-weight differential encoder
CN111125354A (en) * 2018-10-31 2020-05-08 北京国双科技有限公司 Text classification method and device
CN109670169A (en) * 2018-11-16 2019-04-23 中山大学 A kind of deep learning sensibility classification method based on feature extraction
CN109508375A (en) * 2018-11-19 2019-03-22 重庆邮电大学 A kind of social affective classification method based on multi-modal fusion
CN109492229A (en) * 2018-11-23 2019-03-19 中国科学技术大学 A kind of cross-cutting sensibility classification method and relevant apparatus
CN109492229B (en) * 2018-11-23 2020-10-27 中国科学技术大学 Cross-domain emotion classification method and related device
CN111241842A (en) * 2018-11-27 2020-06-05 阿里巴巴集团控股有限公司 Text analysis method, device and system
CN111241842B (en) * 2018-11-27 2023-05-30 阿里巴巴集团控股有限公司 Text analysis method, device and system
CN109284506B (en) * 2018-11-29 2023-09-29 重庆邮电大学 User comment emotion analysis system and method based on attention convolution neural network
CN109284506A (en) * 2018-11-29 2019-01-29 重庆邮电大学 A kind of user comment sentiment analysis system and method based on attention convolutional neural networks
CN109597995A (en) * 2018-12-04 2019-04-09 国网江西省电力有限公司信息通信分公司 A kind of document representation method based on BM25 weighted combination term vector
CN109597997A (en) * 2018-12-07 2019-04-09 上海宏原信息科技有限公司 Based on comment entity, aspect grade sensibility classification method and device and its model training
CN109597997B (en) * 2018-12-07 2023-05-02 上海宏原信息科技有限公司 Comment entity and aspect-level emotion classification method and device and model training thereof
CN109670542A (en) * 2018-12-11 2019-04-23 田刚 A kind of false comment detection method based on comment external information
CN109918501A (en) * 2019-01-18 2019-06-21 平安科技（深圳）有限公司 Method, apparatus, equipment and the storage medium of news article classification
CN109784280A (en) * 2019-01-18 2019-05-21 江南大学 Human bodys' response method based on Bi-LSTM-Attention model
CN109858034B (en) * 2019-02-25 2023-02-03 武汉大学 Text emotion classification method based on attention model and emotion dictionary
CN109858034A (en) * 2019-02-25 2019-06-07 武汉大学 A kind of text sentiment classification method based on attention model and sentiment dictionary
CN109992780A (en) * 2019-03-29 2019-07-09 哈尔滨理工大学 One kind being based on deep neural network specific objective sensibility classification method
CN109992780B (en) * 2019-03-29 2022-07-01 哈尔滨理工大学 Specific target emotion classification method based on deep neural network
CN110083702A (en) * 2019-04-15 2019-08-02 中国科学院深圳先进技术研究院 A kind of aspect rank text emotion conversion method based on multi-task learning
CN110083702B (en) * 2019-04-15 2021-04-09 中国科学院深圳先进技术研究院 Aspect level text emotion conversion method based on multi-task learning
CN110032645A (en) * 2019-04-17 2019-07-19 携程旅游信息技术（上海）有限公司 Text emotion recognition methods, system, equipment and medium
CN110032645B (en) * 2019-04-17 2021-02-09 携程旅游信息技术（上海）有限公司 Text emotion recognition method, system, device and medium
CN110046353B (en) * 2019-04-22 2022-05-13 重庆理工大学 Aspect level emotion analysis method based on multi-language level mechanism
CN110046353A (en) * 2019-04-22 2019-07-23 重庆理工大学 Aspect level emotion analysis method based on multi-language level mechanism
CN110175237B (en) * 2019-05-14 2023-02-03 华东师范大学 Multi-category-oriented secondary emotion classification method
CN110175237A (en) * 2019-05-14 2019-08-27 华东师范大学 It is a kind of towards multi-class secondary sensibility classification method
US20220223144A1 (en) * 2019-05-14 2022-07-14 Dolby Laboratories Licensing Corporation Method and apparatus for speech source separation based on a convolutional neural network
CN110134966A (en) * 2019-05-21 2019-08-16 中电健康云科技有限公司 A kind of sensitive information determines method and device
CN110309503A (en) * 2019-05-21 2019-10-08 昆明理工大学 A kind of subjective item Rating Model and methods of marking based on deep learning BERT--CNN
CN110222178A (en) * 2019-05-24 2019-09-10 新华三大数据技术有限公司 Text sentiment classification method, device, electronic equipment and readable storage medium storing program for executing
CN110188202A (en) * 2019-06-06 2019-08-30 北京百度网讯科技有限公司 Training method, device and the terminal of semantic relation identification model
CN110263165A (en) * 2019-06-14 2019-09-20 中山大学 A kind of user comment sentiment analysis method based on semi-supervised learning
CN110348016B (en) * 2019-07-15 2022-06-14 昆明理工大学 Text abstract generation method based on sentence correlation attention mechanism
CN110348016A (en) * 2019-07-15 2019-10-18 昆明理工大学 Text snippet generation method based on sentence association attention mechanism
CN110765769A (en) * 2019-08-27 2020-02-07 电子科技大学 Entity attribute dependency emotion analysis method based on clause characteristics
US11934790B2 (en) * 2019-09-09 2024-03-19 Boe Technology Group Co., Ltd. Neural network training method and apparatus, semantic classification method and apparatus and medium
US20220075955A1 (en) * 2019-09-09 2022-03-10 Boe Technology Group Co., Ltd. Neural network training method and apparatus, semantic classification method and apparatus and medium
CN110598786B (en) * 2019-09-09 2022-01-07 京东方科技集团股份有限公司 Neural network training method, semantic classification method and semantic classification device
CN110598786A (en) * 2019-09-09 2019-12-20 京东方科技集团股份有限公司 Neural network training method, semantic classification method and semantic classification device
CN110600047A (en) * 2019-09-17 2019-12-20 南京邮电大学 Perceptual STARGAN-based many-to-many speaker conversion method
CN110517121A (en) * 2019-09-23 2019-11-29 重庆邮电大学 Method of Commodity Recommendation and the device for recommending the commodity based on comment text sentiment analysis
CN112685558A (en) * 2019-10-18 2021-04-20 普天信息技术有限公司 Emotion classification model training method and device
CN110826340A (en) * 2019-11-06 2020-02-21 广东三维家信息科技有限公司 Evaluation text generation method and device and electronic equipment
CN110851601A (en) * 2019-11-08 2020-02-28 福州大学 Cross-domain emotion classification system and method based on layered attention mechanism
CN110879938A (en) * 2019-11-14 2020-03-13 中国联合网络通信集团有限公司 Text emotion classification method, device, equipment and storage medium
CN111078833B (en) * 2019-12-03 2022-05-20 哈尔滨工程大学 Text classification method based on neural network
CN111078833A (en) * 2019-12-03 2020-04-28 哈尔滨工程大学 Text classification method based on neural network
CN111159400B (en) * 2019-12-19 2023-09-26 苏州大学 Product comment emotion classification method and system
CN111159400A (en) * 2019-12-19 2020-05-15 苏州大学 Product comment emotion classification method and system
CN111046233A (en) * 2019-12-24 2020-04-21 浙江大学 Video label determination method based on video comment text
CN111046233B (en) * 2019-12-24 2022-05-13 浙江大学 Video label determination method based on video comment text
CN111414122A (en) * 2019-12-26 2020-07-14 腾讯科技（深圳）有限公司 Intelligent text processing method and device, electronic equipment and storage medium
CN111275521B (en) * 2020-01-16 2022-06-14 华南理工大学 Commodity recommendation method based on user comment and satisfaction level embedding
CN111275521A (en) * 2020-01-16 2020-06-12 华南理工大学 Commodity recommendation method based on user comment and satisfaction level embedding
CN111339440B (en) * 2020-02-19 2024-01-23 东南大学 Social emotion sequencing method based on hierarchical state neural network for news text
CN111339440A (en) * 2020-02-19 2020-06-26 东南大学 Social emotion ordering method for news text based on hierarchical state neural network
CN111325027A (en) * 2020-02-19 2020-06-23 东南大学 Sparse data-oriented personalized emotion analysis method and device
CN111368082A (en) * 2020-03-03 2020-07-03 南京信息工程大学 Emotion analysis method for domain adaptive word embedding based on hierarchical network
WO2021174824A1 (en) * 2020-03-05 2021-09-10 苏州浪潮智能科技有限公司 Sentence-level convolution lstm training method, and device and readable medium
CN111291189A (en) * 2020-03-10 2020-06-16 北京芯盾时代科技有限公司 Text processing method and device and computer readable storage medium
CN111522956A (en) * 2020-05-08 2020-08-11 河南理工大学 Text emotion classification method based on double channels and hierarchical attention network
CN111639183A (en) * 2020-05-19 2020-09-08 民生科技有限责任公司 Financial industry consensus public opinion analysis method and system based on deep learning algorithm
CN111639183B (en) * 2020-05-19 2023-11-28 民生科技有限责任公司 Financial co-industry public opinion analysis method and system based on deep learning algorithm
CN111859978B (en) * 2020-06-11 2023-06-20 南京邮电大学 Deep learning-based emotion text generation method
CN111859978A (en) * 2020-06-11 2020-10-30 南京邮电大学 Emotion text generation method based on deep learning
CN111858944B (en) * 2020-07-31 2022-11-22 电子科技大学 Entity aspect level emotion analysis method based on attention mechanism
CN111858944A (en) * 2020-07-31 2020-10-30 电子科技大学 Entity aspect level emotion analysis method based on attention mechanism
CN111985214A (en) * 2020-08-19 2020-11-24 四川长虹电器股份有限公司 Human-computer interaction negative emotion analysis method based on bilstm and attention
CN111984931A (en) * 2020-08-20 2020-11-24 上海大学 Public opinion calculation and deduction method and system for social event web text
CN111984931B (en) * 2020-08-20 2022-06-03 上海大学 Public opinion calculation and deduction method and system for social event web text
CN112347269A (en) * 2020-11-11 2021-02-09 重庆邮电大学 Method for recognizing argument pairs based on BERT and Att-BilSTM
CN112434516A (en) * 2020-12-18 2021-03-02 安徽商信政通信息技术股份有限公司 Self-adaptive comment emotion analysis system and method fusing text information
CN112597302A (en) * 2020-12-18 2021-04-02 东北林业大学 False comment detection method based on multi-dimensional comment representation
CN112597302B (en) * 2020-12-18 2022-04-29 东北林业大学 False comment detection method based on multi-dimensional comment representation
CN112905739A (en) * 2021-02-05 2021-06-04 北京邮电大学 False comment detection model training method, detection method and electronic equipment
CN112950019A (en) * 2021-03-01 2021-06-11 昆明电力交易中心有限责任公司 Electricity selling company evaluation emotion classification method based on combined attention mechanism
CN112950019B (en) * 2021-03-01 2024-03-29 昆明电力交易中心有限责任公司 Electricity selling company evaluation emotion classification method based on joint attention mechanism
CN113065577A (en) * 2021-03-09 2021-07-02 北京工业大学 Multi-modal emotion classification method for targets
CN112699244A (en) * 2021-03-16 2021-04-23 成都信息工程大学 Deep learning-based method and system for classifying defect texts of power transmission and transformation equipment
CN113051367B (en) * 2021-03-22 2023-11-21 北京智慧星光信息技术有限公司 Deep learning early warning method and system based on semantic feature reinforcement and electronic equipment
CN113051367A (en) * 2021-03-22 2021-06-29 北京智慧星光信息技术有限公司 Deep learning early warning method and system based on semantic feature enhancement and electronic equipment
CN113159831A (en) * 2021-03-24 2021-07-23 湖南大学 Comment text sentiment analysis method based on improved capsule network
CN113255360A (en) * 2021-04-19 2021-08-13 国家计算机网络与信息安全管理中心 Document rating method and device based on hierarchical self-attention network
CN113297838A (en) * 2021-05-21 2021-08-24 华中科技大学鄂州工业技术研究院 Relationship extraction method based on graph neural network
CN113361617A (en) * 2021-06-15 2021-09-07 西南交通大学 Aspect level emotion analysis modeling method based on multivariate attention correction
CN113297383B (en) * 2021-06-22 2023-08-04 苏州大学 Speech emotion classification method based on knowledge distillation
CN113297383A (en) * 2021-06-22 2021-08-24 苏州大学 Knowledge distillation-based speech emotion classification method
CN113407721A (en) * 2021-06-29 2021-09-17 哈尔滨工业大学（深圳） Method, device and computer storage medium for detecting log sequence abnormity
CN113553828B (en) * 2021-07-21 2023-06-16 南京邮电大学 Hierarchical remote supervision relation extraction method based on original sense code
CN113553828A (en) * 2021-07-21 2021-10-26 南京邮电大学 Semantic code-based hierarchical level remote supervision relation extraction method
CN113762343A (en) * 2021-08-04 2021-12-07 德邦证券股份有限公司 Method, device and storage medium for processing public opinion information and training classification model
CN113762343B (en) * 2021-08-04 2024-03-15 德邦证券股份有限公司 Method, device and storage medium for processing public opinion information and training classification model
CN113779249A (en) * 2021-08-31 2021-12-10 华南师范大学 Cross-domain text emotion classification method and device, storage medium and electronic equipment
CN113806545A (en) * 2021-09-24 2021-12-17 重庆理工大学 Comment text emotion classification method based on label description generation
CN113806545B (en) * 2021-09-24 2022-06-17 重庆理工大学 Comment text emotion classification method based on label description generation
CN113849651A (en) * 2021-09-28 2021-12-28 平安科技（深圳）有限公司 Document-level emotional tendency-based emotion classification method, device, equipment and medium
CN113849651B (en) * 2021-09-28 2024-04-09 平安科技（深圳）有限公司 Emotion classification method, device, equipment and medium based on document-level emotion tendencies
CN113869065A (en) * 2021-10-15 2021-12-31 梧州学院 Emotion classification method and system based on 'word-phrase' attention mechanism
CN113869065B (en) * 2021-10-15 2024-04-12 梧州学院 Emotion classification method and system based on 'word-phrase' attention mechanism
WO2023069017A3 (en) * 2021-10-19 2023-06-01 Grabtaxi Holdings Pte. Ltd. System and method for recognizing sentiment of user's feedback
WO2023204759A1 (en) * 2022-04-22 2023-10-26 Lemon Inc. Attribute and rating co-extraction
CN114791951A (en) * 2022-05-13 2022-07-26 青岛文达通科技股份有限公司 Emotion classification method and system based on capsule network
CN115544260A (en) * 2022-12-05 2022-12-30 湖南工商大学 Comparison optimization coding and decoding model and method for text emotion analysis
CN116484005B (en) * 2023-06-25 2023-09-08 北京中关村科金技术有限公司 Classification model construction method, device and storage medium
CN116484005A (en) * 2023-06-25 2023-07-25 北京中关村科金技术有限公司 Classification model construction method, device and storage medium

Also Published As

Publication number Publication date
CN108363753B (en) 2020-05-19

Similar Documents

Publication Publication Date Title
CN108363753A (en) Comment text sentiment classification model is trained and sensibility classification method, device and equipment
Abdullah et al. SEDAT: sentiment and emotion detection in Arabic text using CNN-LSTM deep learning
CN108492200B (en) User attribute inference method and device based on convolutional neural network
CN109933664B (en) Fine-grained emotion analysis improvement method based on emotion word embedding
Lai et al. Fine-grained emotion classification of Chinese microblogs based on graph convolution networks
CN109145112A (en) A kind of comment on commodity classification method based on global information attention mechanism
CN109284506A (en) A kind of user comment sentiment analysis system and method based on attention convolutional neural networks
CN109614487B (en) Sentiment classification method based on tensor fusion mode
Sivakumar et al. Review on word2vec word embedding neural net
CN107247702A (en) A kind of text emotion analysis and processing method and system
CN110096575B (en) Psychological portrait method facing microblog user
CN111274398A (en) Method and system for analyzing comment emotion of aspect-level user product
CN103034626A (en) Emotion analyzing system and method
CN108052505A (en) Text emotion analysis method and device, storage medium, terminal
CN110765769B (en) Clause feature-based entity attribute dependency emotion analysis method
CN105809186A (en) Emotion classification method and system
CN107688870A (en) A kind of the classification factor visual analysis method and device of the deep neural network based on text flow input
CN112256866A (en) Text fine-grained emotion analysis method based on deep learning
CN110297986A (en) A kind of Sentiment orientation analysis method of hot microblog topic
CN110909529A (en) User emotion analysis and prejudgment system of company image promotion system
CN114417851A (en) Emotion analysis method based on keyword weighted information
CN113934835B (en) Retrieval type reply dialogue method and system combining keywords and semantic understanding representation
Palani et al. T-BERT--Model for Sentiment Analysis of Micro-blogs Integrating Topic Model and BERT
Leng et al. Deepreviewer: Collaborative grammar and innovation neural network for automatic paper review
CN111859955A (en) Public opinion data analysis model based on deep learning

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