CN109992780B - Specific target emotion classification method based on deep neural network - Google Patents
Specific target emotion classification method based on deep neural network Download PDFInfo
- Publication number
- CN109992780B CN109992780B CN201910249992.4A CN201910249992A CN109992780B CN 109992780 B CN109992780 B CN 109992780B CN 201910249992 A CN201910249992 A CN 201910249992A CN 109992780 B CN109992780 B CN 109992780B
- Authority
- CN
- China
- Prior art keywords
- target
- specific
- specific target
- word
- word vector
- 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.)
- Expired - Fee Related
Links
- 230000008451 emotion Effects 0.000 title claims abstract description 60
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 25
- 239000013598 vector Substances 0.000 claims abstract description 82
- 238000012549 training Methods 0.000 claims abstract description 39
- 230000007246 mechanism Effects 0.000 claims abstract description 11
- 230000011218 segmentation Effects 0.000 claims abstract description 6
- 230000008569 process Effects 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 40
- 230000015654 memory Effects 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 7
- 230000002996 emotional effect Effects 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 3
- 238000003058 natural language processing Methods 0.000 abstract description 6
- 230000007787 long-term memory Effects 0.000 abstract description 4
- 230000006403 short-term memory Effects 0.000 abstract description 4
- 238000004458 analytical method Methods 0.000 description 8
- 238000010801 machine learning Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000013145 classification model Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000004576 sand Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention provides a specific target emotion classification method based on a deep neural network. Belonging to the field of text sentiment classification of natural language processing. Firstly, Chinese word segmentation, stop word removal and punctuation removal are carried out on a data set, then word2vec algorithm is adopted to train the processed corpus to obtain corresponding word vectors, then the training set is input into a long-term and short-term memory network model structure based on a target attention mechanism, a specific target and a specific aspect are embedded in the process of realizing attention weight training, the specific target is expressed by weighted summation embedded in the specific aspect, so that the model gives more correct attention to the specific target and the specific aspect, the real semantics of the target is better captured, and finally the emotion classification accuracy of the specific target is improved.
Description
Technical Field
The invention relates to comment text emotion classification, in particular to a depth neural network-based specific target emotion classification method, and belongs to the technical field of natural language processing.
Background
The emotion analysis method mainly comprises a rule-based method, a machine learning-based method and a deep neural network-based method. The rule-based method usually needs to construct an emotion dictionary or emotion collocation template, and then calculates the emotion tendency of the text by comparing emotion words or fixed collocation contained in the comment text, but the construction of a relatively complete emotion dictionary or related collocation rules is a major problem at present. The method based on machine learning mainly carries out feature extraction and modeling on the labeled training corpus, thereby automatically realizing judgment of emotion polarity by using a machine learning algorithm; the method mainly comprises a support vector machine, naive Bayes, maximum information entropy, a conditional random field and the like, but the effect of machine learning classification is usually determined by the selection of features, great uncertainty exists in manual feature selection, and functions used in modeling the materials are generally simple, deep features are difficult to capture, and the modeling capability and generalization capability have great limitations. With the development of deep learning and the freedom and diversification of language expression modes, the advantage of the deep neural network technology is gradually highlighted and becomes the mainstream technology in the natural language processing field, compared with the emotion analysis method based on rules and the emotion analysis method based on machine learning, the deep neural network method can capture more comprehensive and deeper text characteristics when facing the current complex and changeable language phenomena due to the complexity of the model and the function, namely, the deep neural network method has better comprehension capability on the text and can achieve better effect in the emotion analysis field.
The LSTM neural network model, also called long-short term memory network model, is a variant of the RNN model. The LSTM solves the problem of information disappearance or information explosion when the RNN model transmits information in a long distance, and the LSTM neural network model adds various gate structures to neural network nodes on the basis of the RNN model to control the information to flow at different moments. In order to control the flow of information, a memory unit is specially designed in an internal node of the LSTM neural network, and the deletion or addition of the information is controlled through a gate structure, wherein the gate is a method for selectively passing the information, and the nodes of the LSTM neural network are provided with three gate structures for protecting and controlling the states of the nodes, namely an input gate, a forgetting gate and an output gate. The attention mechanism is derived from the fact that more attention is allocated to key parts of things concerned by human brains, is applied to the field of visual images at first and then applied to a task of natural language processing, and has a good effect.
The specific target emotion analysis is an important subtask of emotion analysis and is deeper emotion analysis. Different from ordinary emotion analysis, the judgment of the emotion polarity of the specific target not only depends on the context information of the text, but also depends on the feature information of the specific target. For example, the sentence "the food at this restaurant is very good to eat but expensive, but the service is compelling," the "taste" aspect for the target "food" is a positive emotion, the "price" aspect for the target "food" is a negative emotion, and the "service" aspect for the target "restaurant" is a positive emotion. Therefore, even if the same sentence is used, the completely opposite emotional polarities may occur for the same target, and different targets may have different emotional polarities. However, most text emotion classification models based on neural networks do not pay correct attention to the emotion of a specific aspect of a specific target, and the classification effect is poor. The method realizes better capture of the real semantics of the target, enriches the semantic information in the text, improves the accuracy of emotion classification of the specific target, and is the main research direction of the method.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a specific target emotion classification method based on a deep neural network and a target attention mechanism, which is used for analyzing the emotion colors of specific targets and specific aspects contained in text data in a social network.
The invention can be realized by adopting the following system:
a specific target emotion classification method based on a deep neural network is characterized by comprising the following steps:
step one, acquiring a Chinese emotion classification data set and preprocessing a text, and dividing the emotion classification data set into a training set and a test set;
secondly, training a word vector model for the preprocessed data set by using a word2vec tool and mapping texts in the data set into a word vector set;
step three, inputting the word vector set of the training set into the LSTM, discarding or transferring information by using three gates in the LSTM with trainable parameters, and outputting a series of hidden vectors h ═ { h ═ h1,h2,…,hn};
Step four, putting the word vector matrix of the training set, the word vector matrix of the specific target and the word vector matrix of the specific aspect into a target attention mechanism to obtain each hiPositive weight p ofiSubsequently, a sentence representation Z is obtainedS;
Step five, generating a sentence ZSAnd judging the emotional polarity of the specific target by using the full connection layer and the softmax function.
Further, the text preprocessing specifically includes: :
preprocessing mainly comprises the steps of carrying out Chinese word segmentation, word stop and punctuation removal on the sentences marked with emotion polarities; 80% of the data sets were randomly selected as training sets and 20% as test sets.
Further, the training a word vector model on the preprocessed data set using a word2vec tool comprises:
after the word2vec model training is completed, the word2vec model may be used to map each word ω to a continuous feature vector eω∈RdWherein d represents the dimension of the word vector, and finally generating a word vector matrix E E ∈ Rv×dWhere V represents the size of the vocabulary in the dataset.
Further, the set of word vectors of the training set is input into the LSTM, information is discarded or passed using three gates in the LSTM with trainable parameters, and a series of hidden vectors h are output={h1,h2,…,hnThe method specifically comprises the following steps:
three gates in the LSTM include an input gate, a forgetting gate, and an output gate. Let xtFor input at t moment of certain node of LSTM neural network, htIs the output at time t, WxFor inputting the corresponding weight, WhIn order to output the corresponding weight, the flow of updating the LSTM neural network model through the gate structure control information is divided into the following steps:
calculating the value i at the moment t of the input gatetThe input gate controls the influence of the current input on the state value of the memory unit, and the calculation method is as follows
it=sigmoid(Wxixt+Whiht-1+Wcict-1+bi) (1)
Calculating the value f at the moment of forgetting to leave the door ttThe influence of the historical information on the state value of the memory unit is controlled by the forgetting gate, and the calculation method is as follows:
ft=sigmoid(Wxfxt+Whfht-1+Wcfct-1+bf) (2)
calculating the value of the candidate memory unit at the current timeAnd updating the value of the memory cell at the current moment, wherein the calculation method comprises the following steps:
ct=ft·ct-1+it·ct (4)
finally calculating the output information h at the moment ttThe information is determined by an output gate, and the calculation method is as follows:
ot=sigmoid(Wxoxt+Whoht-1+Wcoct-1+bo) (5)
ht=ot·tanh(ct) (6)
putting the word vector matrix of the training set, the word vector matrix of the specific target and the word vector matrix of the specific aspect into a target attention mechanism specifically comprises the following steps:
and calculating the word vector matrix of the training set and the word vector matrix of the specific target as follows:
where the representative Average returns the Average of the input vectors.Is a matrix of word vectors for a particular target,as a word vector matrix of the training set, cSIs operative to capture both target information and context information.
Computing a weight vector q over all k aspect-specific embeddingstThe formula is as follows:
qt=softmax(Wt·cS+bt) (8)
wherein q istRepresenting a vector of weights embedded over all k particular aspects, each weight qtIndicating the likelihood that a particular object belongs to a relevant aspect, WtAnd btRespectively representing the weight matrix and the bias vector.
Computing a target-specific vector tsThe formula is as follows
ts=T·qt (9)
Wherein t issVector representing a specific target, T represents a word vector matrix of a specific aspect, T ∈ RK×dWherein K represents a specific aspect number.
Computing a positive weight piThe formula is as follows:
Computing a sentence representation ZSThe formula is as follows:
each hidden vector hiCorresponding to a positive weight piWherein the value piCalculated from the target attention model, piCan be interpreted as ω when determining the emotional polarity of a specific target aiIs the probability that the model is correctly focused on the word. ZSRepresenting sentences used for emotion classification.
In summary, the invention provides a method for classifying specific target emotion based on a deep neural network. Belonging to the field of text emotion classification of natural language processing. Firstly, Chinese word segmentation, stop word removal and punctuation removal are carried out on a data set, then word2vec algorithm is adopted to train the processed corpus to obtain corresponding word vectors, then the training set is input into a long-term and short-term memory network model structure based on a target attention mechanism, a specific target and a specific aspect are embedded in the process of realizing attention weight training, the specific target is expressed by weighted summation embedded in the specific aspect, so that the model gives more correct attention to the specific target and the specific aspect, the real semantics of the target is better captured, and finally the emotion classification accuracy of the specific target is improved.
Compared with the prior art, the invention has the following beneficial effects:
on the basis of a long-term and short-term memory network, a target attention mechanism is introduced, a specific target is represented by a weighted sum embedded in a specific aspect, so that the model can better capture the real semantics of the specific target, the model can give more correct attention to the specific target and the specific aspect, meanwhile, the influence of secondary information in a text is ignored or reduced, the real semantics of the target can be better captured, and finally, the emotion classification accuracy of the specific target is improved.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flow chart of the emotion classification model training of the present invention;
FIG. 2 is a view of the internal structure of the LSTM of the present invention;
FIG. 3 is the overall architecture of the emotion classification model of the present invention.
Detailed Description
In order to make the technical scheme in the embodiments of the present invention better understood and make the above objects, features and advantages of the present invention more obvious and understandable, the present invention provides an embodiment of a method for classifying specific target emotions based on a deep neural network, and the technical scheme of the present invention is further described in detail below with reference to the accompanying drawings:
the invention firstly provides a specific target emotion classification method based on a deep neural network, which comprises the following steps of:
s101, step 1: acquiring a Chinese emotion classification data set and preprocessing a text, and dividing the data set into a training set and a test set;
s102, step 2: training a word vector model for the preprocessed data set by using a word2vec tool and mapping texts in the data set into a word vector set;
s103, step 3: the word vector set of the training set is input into the LSTM, information is discarded or passed using three gates in the LSTM with trainable parameters, and a series of hidden vectors h ═ h is output1,h2,…,hn};
S104, step 4: putting the word vector matrix of the training set, the word vector matrix of the specific target and the word vector matrix of the specific aspect into a target attention mechanism to obtain each hiPositive weight p ofiThen, the sentence Z is obtainedS;
S105, step 5: according to the generated sentence ZSAnd judging the emotional polarity of the specific target by using the full connection layer and the softmax function.
The preprocessing in the step 1 mainly comprises the steps of carrying out Chinese word segmentation, word stop and punctuation removal on the sentences marked with emotion polarities; 80% of the data sets were randomly selected as training sets and 20% as test sets.
After the word2vec model training is completed in step 2, the word2vec model may be used to map each word ω to a continuous feature vector eω∈RdWherein d represents the dimension of the word vector, and finally generating a word vector matrix E E ∈ Rv×dWhere V represents the size of the vocabulary in the dataset.
The three gates in the LSTM in step 3, as shown in fig. 2, include an input gate, a forgetting gate, and an output gate. Let xtFor input at t moment of certain node of LSTM neural network, htIs the output at time t, WxFor inputting the corresponding weight, WhIn order to output the corresponding weight value, the flow of updating the LSTM neural network model through the gate structure control information is to firstly calculate the value i of the input gate at the t momenttThe input gate controls the influence of the current input on the state value of the memory unit, and the calculation method is as follows
it=sigmoid(Wxixt+Whiht-1+Wcict-1+bi) (1)
Then calculating the value f of the moment t of forgetting to leave the doortThe influence of the historical information on the state value of the memory unit is controlled by the forgetting gate, and the calculation method is as follows:
ft=sigmoid(Wxfxt+Whfht-1+Wcfct-1+bf) (2)
recalculating current time candidate memory listValue of elementAnd updating the value of the memory cell at the current moment, wherein the calculation method comprises the following steps:
ct=ft·ct-1+it·ct (4)
finally calculating the output information h at the moment ttThe information is determined by an output gate, and the calculation method is as follows:
ot=sigmoid(Wxoxt+Whoht-1+Wcoct-1+bo) (5)
ht=ot·tanh(ct) (6)
in step 4, the word vector matrix of the training set and the word vector matrix of the specific target are first averaged, as shown in the following formula:
where Average represents the Average of the returned input vectors.Is a matrix of word vectors for a particular target,as a word vector matrix of the training set, cSIs operative to capture both target information and context information.
Next step cSPut into the softmax function to calculate the weight vector q embedded on all k specific aspectstThe formula is as follows:
qt=softmax(Wt·cS+bt) (8)
wherein q istRepresenting a vector of weights embedded over all k particular aspects, each weight qtIndicating the likelihood that a particular object belongs to a relevant aspect, WtAnd btRespectively representing the weight matrix and the bias vector.
Q is to betPerforming point multiplication with the word vector matrix of a specific aspect to calculate a vector t of a specific targetsThe formula is as follows
ts=T·qt (9)
Wherein t issVector representing a specific target, T represents a word vector matrix of a specific aspect, T ∈ RK×dWhere K represents a specific number of facets, much smaller than V.
Next, a positive weight p is calculatediThe formula is as follows:
Then the sentence representation Z is calculatedSThe formula is as follows:
each hidden vector hiCorresponding to a positive weight piWherein the value piCalculated from the target attention model, piCan be interpreted as ω when judging the emotion polarity of a specific target aiIs the probability that the model is correctly focused on the word. ZSRepresenting sentences used for emotion classification.
Finally, according to the generated sentence ZSThe emotion polarity of the specific target is judged by using the full connection layer and the softmax function, and a specific calculation flow is shown in fig. 3.
In summary, the invention provides a method for classifying specific target emotion based on a deep neural network. Belonging to the field of text emotion classification of natural language processing. Firstly, Chinese word segmentation, stop word removal and punctuation removal are carried out on a data set, then word2vec algorithm is adopted to train the processed corpus to obtain corresponding word vectors, then the training set is input into a long-term and short-term memory network model structure based on a target attention mechanism, a specific target and a specific aspect are embedded in the process of realizing attention weight training, the specific target is expressed by weighted summation embedded in the specific aspect, so that the model gives more correct attention to the specific target and the specific aspect, the real semantics of the target is better captured, and finally the emotion classification accuracy of the specific target is improved.
The above examples are intended to illustrate but not to limit the technical solutions of the present invention. Any modification or partial replacement without departing from the spirit and scope of the present invention should be covered in the claims of the present invention.
Claims (4)
1. A specific target emotion classification method based on a deep neural network is characterized by comprising the following steps:
step one, acquiring a Chinese emotion classification data set and preprocessing a text, and dividing the emotion classification data set into a training set and a test set;
secondly, training a word vector model for the preprocessed data set by using a word2vec tool and mapping texts in the data set into a word vector set;
step three, inputting the word vector set of the training set into the LSTM, discarding or transferring information by using three gates in the LSTM with trainable parameters, and outputting a series of hidden vectors h ═ { h { (h) }1,h2,…,hn};
Step four, putting the word vector matrix of the training set, the word vector matrix of the specific target and the word vector matrix of the specific aspect into a target attention mechanism to obtain each hiPositive weight p ofiSubsequently, a sentence representation Z is obtainedS;
Step five, generating a sentence ZSJudging the emotion polarity of a specific target by using a full connection layer and a softmax function;
putting the word vector matrix of the training set, the word vector matrix of the specific target and the word vector matrix of the specific aspect into a target attention mechanism specifically comprises the following steps:
(5.1) calculating the word vector matrix of the training set and the word vector matrix of the specific target as follows:
where Average represents the Average of the returned input vectors,is a matrix of word vectors for a particular target,as a word vector matrix of the training set, cSFor capturing target information and context information simultaneously;
(5.2) computing a weight vector q embedded on all k specific aspectstThe formula is as follows:
qt=soft max(Wt·cS+bt) (8)
wherein q istRepresenting a vector of weights embedded over all k particular aspects, each weight qtIndicating the likelihood that a particular object belongs to a relevant aspect, WtAnd btRespectively representing a weight matrix and an offset vector;
(5.3) calculating a target-specific vector tsThe formula is as follows
ts=T·qt (9)
Wherein t issTarget-specific vector, T represents aspect-specific word vector matrix, T ∈ RK×dWherein K represents a specific aspect number;
(5.4) calculation ofPositive weight piThe formula is as follows:
(5.5) calculating sentence representation ZSThe formula is as follows:
each hidden vector hiCorresponding to a positive weight piWherein the value piCalculated from the target attention model, piInterpreted as the probability, Z, of the word that the model is correctly interested in when judging the emotional polarity of a particular target aSRepresents a sentence for emotion classification;
in the process of realizing attention weight training, the specific target and the specific aspect are embedded, and the specific target is represented by the weighted sum of the embedding of the specific aspect, so that the emotion classification accuracy of the specific target is finally improved.
2. The method for classifying specific target emotion based on deep neural network as claimed in claim 1, wherein: the text preprocessing specifically comprises the following steps: carrying out Chinese word segmentation, word stop and punctuation removal on the sentences marked with emotion polarities; 80% of the data sets were randomly selected as training sets and 20% as test sets.
3. The method for classifying specific target emotion based on deep neural network as claimed in claim 1, wherein: training a word vector model using a word2vec tool on the preprocessed dataset comprises:
word2after the vec model training is completed, the word2vec model is used to map each word ω to a continuous feature vector eω∈RdWherein d represents the dimension of the word vector, and finally generating a word vector matrix E E ∈ Rv×dWhere V represents the size of the vocabulary in the dataset.
4. The method for classifying specific target emotion based on deep neural network as claimed in claim 1, wherein: the set of word vectors of the training set is input into the LSTM, information is discarded or passed using three gates in the LSTM with trainable parameters, and a series of hidden vectors h ═ { h ═ is output1,h2,…,hnThe method specifically comprises the following steps:
three gates in the LSTM, including an input gate, a forgetting gate and an output gate; let xtFor input at t moment of certain node of LSTM neural network, htIs the output at time t, WxFor inputting the corresponding weight, WhIn order to output the corresponding weight, the flow of updating the LSTM neural network model through the gate structure control information is divided into four steps:
(4.1) calculating the value i at the moment t of the input gatetThe input gate controls the influence of the current input on the state value of the memory unit, and the calculation method is as follows
it=sigmoid(Wxixt+Whiht-1+Wcict-1+bi) (1)
(4.2) calculating the value f at the moment of forgetting to gate ttThe influence of the historical information on the state value of the memory unit is controlled by the forgetting gate, and the calculation method is as follows:
ft=sigmoid(Wxfxt+Whfht-1+Wcfct-1+bf) (2)
(4.3) calculating the value of the candidate memory cell at the current timeAnd updating the value of the memory cell at the current time, calculatingThe method comprises the following steps:
ct=ft·ct-1+it·ct (4)
(4.4) finally calculating the output information h at the moment ttThe information is determined by an output gate, and the calculation method is as follows:
ot=sigmoid(Wxoxt+Whoht-1+Wcoct-1+bo) (5)
ht=ot·tanh(ct) (6)。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910249992.4A CN109992780B (en) | 2019-03-29 | 2019-03-29 | Specific target emotion classification method based on deep neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910249992.4A CN109992780B (en) | 2019-03-29 | 2019-03-29 | Specific target emotion classification method based on deep neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109992780A CN109992780A (en) | 2019-07-09 |
CN109992780B true CN109992780B (en) | 2022-07-01 |
Family
ID=67131875
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910249992.4A Expired - Fee Related CN109992780B (en) | 2019-03-29 | 2019-03-29 | Specific target emotion classification method based on deep neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109992780B (en) |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110390017B (en) * | 2019-07-25 | 2022-12-27 | 中国民航大学 | Target emotion analysis method and system based on attention gating convolutional network |
CN110728298A (en) * | 2019-09-05 | 2020-01-24 | 北京三快在线科技有限公司 | Multi-task classification model training method, multi-task classification method and device |
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 |
CN110704622A (en) * | 2019-09-27 | 2020-01-17 | 北京明略软件系统有限公司 | Text emotion classification method and device and electronic equipment |
CN111191026A (en) * | 2019-12-10 | 2020-05-22 | 央视国际网络无锡有限公司 | Text classification method capable of calibrating specific segments |
CN111291189B (en) * | 2020-03-10 | 2020-12-04 | 北京芯盾时代科技有限公司 | Text processing method and device and computer readable storage medium |
CN111444728A (en) * | 2020-04-20 | 2020-07-24 | 复旦大学 | End-to-end aspect-based emotion analysis method |
CN112115243B (en) * | 2020-08-11 | 2023-06-16 | 南京理工大学 | Session representation learning method by modeling time-series time correlation |
CN112434161B (en) * | 2020-11-24 | 2023-01-03 | 哈尔滨工程大学 | Aspect-level emotion analysis method adopting bidirectional long-short term memory network |
CN112464281B (en) * | 2020-11-29 | 2022-11-18 | 深圳市索迪统计科技有限公司 | Network information analysis method based on privacy grouping and emotion recognition |
CN112699237B (en) * | 2020-12-24 | 2021-10-15 | 百度在线网络技术(北京)有限公司 | Label determination method, device and storage medium |
CN114357166A (en) * | 2021-12-31 | 2022-04-15 | 北京工业大学 | Text classification method based on deep learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107491490A (en) * | 2017-07-19 | 2017-12-19 | 华东师范大学 | Text sentiment classification method based on Emotion center |
CN108363753A (en) * | 2018-01-30 | 2018-08-03 | 南京邮电大学 | Comment text sentiment classification model is trained and sensibility classification method, device and equipment |
CN108460009A (en) * | 2017-12-14 | 2018-08-28 | 中山大学 | The attention mechanism Recognition with Recurrent Neural Network text emotion analytic approach of embedded sentiment dictionary |
CN108717439A (en) * | 2018-05-16 | 2018-10-30 | 哈尔滨理工大学 | A kind of Chinese Text Categorization merged based on attention mechanism and characteristic strengthening |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11205103B2 (en) * | 2016-12-09 | 2021-12-21 | The Research Foundation for the State University | Semisupervised autoencoder for sentiment analysis |
-
2019
- 2019-03-29 CN CN201910249992.4A patent/CN109992780B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107491490A (en) * | 2017-07-19 | 2017-12-19 | 华东师范大学 | Text sentiment classification method based on Emotion center |
CN108460009A (en) * | 2017-12-14 | 2018-08-28 | 中山大学 | The attention mechanism Recognition with Recurrent Neural Network text emotion analytic approach of embedded sentiment dictionary |
CN108363753A (en) * | 2018-01-30 | 2018-08-03 | 南京邮电大学 | Comment text sentiment classification model is trained and sensibility classification method, device and equipment |
CN108717439A (en) * | 2018-05-16 | 2018-10-30 | 哈尔滨理工大学 | A kind of Chinese Text Categorization merged based on attention mechanism and characteristic strengthening |
Non-Patent Citations (4)
Title |
---|
attention-based LSTM for aspect-level sentiment classification;Yequan Wang 等;《Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing》;20161105;606-615 * |
Improving sentiment analysis via sentence type classification using BiLSTM-CRF and CNN;Tao Chen 等;《Expert Systems With Applications》;20170415;221-230 * |
基于多注意力卷积神经网络的特定目标情感分析;梁斌等;《计算机研究与发展》;20170815(第08期);1724-1735 * |
基于改进注意力机制的中文细粒度情感分析;王振东;《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》;20210215;I138-2800 * |
Also Published As
Publication number | Publication date |
---|---|
CN109992780A (en) | 2019-07-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109992780B (en) | Specific target emotion classification method based on deep neural network | |
CN110502749B (en) | Text relation extraction method based on double-layer attention mechanism and bidirectional GRU | |
CN108628823B (en) | Named entity recognition method combining attention mechanism and multi-task collaborative training | |
Yao et al. | An improved LSTM structure for natural language processing | |
CN107273355B (en) | Chinese word vector generation method based on word and phrase joint training | |
Cui et al. | Consensus attention-based neural networks for Chinese reading comprehension | |
WO2022007823A1 (en) | Text data processing method and device | |
CN110609891A (en) | Visual dialog generation method based on context awareness graph neural network | |
Ghorbanali et al. | Ensemble transfer learning-based multimodal sentiment analysis using weighted convolutional neural networks | |
CN109308353B (en) | Training method and device for word embedding model | |
CN111401061A (en) | Method for identifying news opinion involved in case based on BERT and Bi L STM-Attention | |
CN109214006B (en) | Natural language reasoning method for image enhanced hierarchical semantic representation | |
CN109271636B (en) | Training method and device for word embedding model | |
CN112784041B (en) | Chinese short text sentiment orientation analysis method | |
Guo et al. | Who is answering whom? Finding “Reply-To” relations in group chats with deep bidirectional LSTM networks | |
CN113934835B (en) | Retrieval type reply dialogue method and system combining keywords and semantic understanding representation | |
CN111046157B (en) | Universal English man-machine conversation generation method and system based on balanced distribution | |
CN112560440A (en) | Deep learning-based syntax dependence method for aspect-level emotion analysis | |
Ermatita et al. | Sentiment Analysis of COVID-19 using Multimodal Fusion Neural Networks. | |
CN112349294A (en) | Voice processing method and device, computer readable medium and electronic equipment | |
WO2023116572A1 (en) | Word or sentence generation method and related device | |
CN111914084A (en) | Deep learning-based emotion label text generation and evaluation system | |
CN115510230A (en) | Mongolian emotion analysis method based on multi-dimensional feature fusion and comparative reinforcement learning mechanism | |
CN114519353A (en) | Model training method, emotion message generation device, emotion message generation equipment and emotion message generation medium | |
Windiatmoko et al. | Mi-Botway: A deep learning-based intelligent university enquiries chatbot |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220701 |