CN109783644A - A kind of cross-cutting emotional semantic classification system and method based on text representation study - Google Patents
A kind of cross-cutting emotional semantic classification system and method based on text representation study Download PDFInfo
- Publication number
- CN109783644A CN109783644A CN201910048910.XA CN201910048910A CN109783644A CN 109783644 A CN109783644 A CN 109783644A CN 201910048910 A CN201910048910 A CN 201910048910A CN 109783644 A CN109783644 A CN 109783644A
- Authority
- CN
- China
- Prior art keywords
- text
- domain
- module
- text representation
- cross
- 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.)
- Pending
Links
- 230000002996 emotional effect Effects 0.000 title claims abstract description 61
- 238000000034 method Methods 0.000 title claims abstract description 22
- 230000006870 function Effects 0.000 claims abstract description 13
- 238000012512 characterization method Methods 0.000 claims abstract description 7
- 239000000284 extract Substances 0.000 claims abstract description 6
- 238000005457 optimization Methods 0.000 claims description 14
- 230000008451 emotion Effects 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 6
- 230000008901 benefit Effects 0.000 claims description 4
- 230000006978 adaptation Effects 0.000 claims description 3
- 239000002131 composite material Substances 0.000 claims description 2
- 210000003739 neck Anatomy 0.000 claims 1
- 238000007781 pre-processing Methods 0.000 abstract 1
- 238000012549 training Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Abstract
The present invention relates to a kind of cross-cutting emotional semantic classification system and methods based on text representation study, include a comment text preprocessing module, for carrying out characterization to cross-domain texts；Source domain and the potential general field feature of target domain and affective characteristics are excavated for the character representation space that learning areas adapts to comprising a text representation study module；Comprising a text representation reinforced module, generates confrontation model and be used to automatically generate the source domain text representation of robust and differentiate its emotional category, model parameter is constantly updated according to error；Comprising an emotional category output module, final emotional semantic classification result is obtained using classification function.The present invention can extract the potential generic features of target domain and source domain automatically, and feature is abstracted and is combined, and finally identify the emotional category of target domain text.
Description
Technical field
The present invention relates to sentiment analysis and opining mining field, especially a kind of cross-cutting feelings based on text representation study
Feel categorizing system and method.
Background technique
Currently, many technical methods can be used for text emotion classification.Traditional text sentiment classification method is based primarily upon
There is the method for the machine learning of supervision, the main affective characteristics by extracting in text use support vector machines (SVMs) equal part
Class device carries out emotional semantic classification.Although this kind achieves certain effect based on the method for Feature Engineering and shallow-layer linear model,
But when training field is different from the distribution of the affective characteristics of testing field, the migration of affective characteristics is poor, traditional emotion
Cross-cutting emotional semantic classification can not be effectively performed in classification method, need to take a substantial amount of time and energy redesign target domain
Affective characteristics, and be limited to engineer rule and the validity of feature and the learning ability of model.And there is prison
It is higher to superintend and direct method accuracy, but when target domain does not have labeled data, how using the connection between field, thus study one
The problem of a cross-cutting Text Representation space is cross-cutting emotional semantic classification is crucial.
Currently, there is some cross-cutting sensibility classification method based on feature selecting.In cross-cutting emotional semantic classification,
The feature all frequently occurred in target domain and source domain is known as pivot feature, and researchers are from the primitive character in two fields
It concentrates and excavates pivot feature and non-pivot feature, construct the mapping relations between domain features, find unified affective characteristics space.
Such methods usually utilize the feature of engineer or use N metagrammar model extraction feature, are unable to fully efficiently to across neck
Domain text is indicated.
Currently, there is some cross-cutting sensibility classification methods based on character representation study.Recently as depth
The fast development of habit is indicated text using neural network model and learns in terms of semantic expressiveness and sentiment analysis utilization more
Tool advantage.These models are also used in cross-cutting emotional semantic classification by many researchers.Neural network model and feature selecting
Method is compared, and can automatically learn text representation from text data, so as to avoid a large amount of Feature Engineering, but needs mesh
The labeled data in mark field effectively train.Some researchers carry out domain classification and feelings using the method that field is fought simultaneously
Sense classification, thus the character representation space that one field of study adapts to, but do not fully consider that the noise in text representation study is asked
Topic, still there is very big exploration space
Therefore it is desirable to find a kind of more efficient cross-cutting sensibility classification method, and then cross-cutting emotional semantic classification is improved
Precision and the consumption for reducing manual time's energy.
Summary of the invention
In view of this, the purpose of the present invention is to propose to it is a kind of based on text representation study cross-cutting emotional semantic classification system and
Method can extract the potential generic features of target domain and source domain automatically, and feature is abstracted and is combined, finally
Identify the emotional category of target domain text.
The present invention is realized using following scheme: a kind of cross-cutting emotional semantic classification system based on text representation study, specifically
Include:
One Text Pretreatment module obtains source domain and target domain for carrying out characterization to cross-domain texts
Original text vector；
One text representation study module receives the output of Text Pretreatment module, the character representation adapted to for learning areas
Space excavates source domain and the potential general field feature of target domain and affective characteristics, obtains source domain and target domain
Text eigenvector；
One text representation reinforced module generates confrontation network model and is used to automatically generate the source domain text representation of robust and sentences
Its other emotional category constantly updates the parameter with optimization text representation study module according to error；
One emotional category output module, according to the Text eigenvector and benefit of the target domain of text representation study module output
The text emotion classification results of target domain are obtained with classification function.
Further, the Text Pretreatment module extracts the spy of source domain and target domain text using the N-gram syntax
Sign, and the character representation that noise reduction self-encoding encoder learns cross-domain texts, the affective tag without target domain are stacked using edge.
Preferably, edge stacks noise reduction self-encoding encoder using less calculation amount and with the scalability to high dimensional feature,
Realize the abstract of domain features.
Further, the character representation that the text representation study module adapts to field using neural network
It practises, while considering domain features and affective characteristics in different field text.
Further, the neural network abstracts text feature, to obtain the character representation of field adaptation
Vector, i.e. source domain Text eigenvector and target domain Text eigenvector.
Further, building generates confrontation network model in the text representation reinforced module, it is contemplated that text table dendrography
Practise noise characteristic problem present in module.
Further, the generation confrontation network model includes generating network and differentiating network, in the two confrontation study
Strengthen text representation space；The generation network synthesizes dummy copy by the way that noise vector is added in text representation vector to be confused
The puzzled judgement for differentiating network differentiates that network carries out the judgement of emotional semantic classification and true and false sample simultaneously, advanced optimizes text representation
Study module.
Further, the generation network generates random noise vector using normal distribution, and the composite vector of generation is logical
The optimization of module is crossed closer to source domain sample.
Further, the judgement for differentiating network and carrying out emotional semantic classification and true and false sample simultaneously, it is contemplated that domain features
With affective characteristics to the percentage contribution of cross-cutting emotional semantic classification, two factors are weighed to the influence degree of result.
Further, the emotional category output module using softmax function to obtained text representation vector into
Row processing, predicts the emotional category of each text.
The present invention also provides a kind of based on the cross-cutting emotional semantic classification system described above based on text representation study
Method, specifically includes the following steps:
Step S1: the Text Pretreatment module receives source domain text data and affective tag, target domain text data, right
Cross-domain texts carry out characterization, obtain the original text vector of source domain and target domain；
Step S2: the text representation study module adapts to the output of Text Pretreatment module as input, learning areas
Source domain and the potential general field feature of target domain and affective characteristics are excavated in character representation space, obtain source domain with
The Text eigenvector of target domain；
Step S3: the text representation reinforced module generates the source domain text table that confrontation network model is used to automatically generate robust
Show and differentiate its emotional category, the parameter with optimization text representation study module is constantly updated according to error；
Step S4: the emotional category output module receive the text feature of the text representation study module output after optimization to
Amount, and the text emotion classification results of target domain are obtained using classification function.
It particularly, will be constantly right according to the back-propagating of the propagated forward of information and error in the training stage of model
They are adjusted, successive optimization objective function.
Compared with prior art, the invention has the following beneficial effects: method proposed by the present invention can extract mesh automatically
The potential generic features in mark field and source domain, and feature is abstracted and is combined, finally identify target domain text
Emotional category.
Detailed description of the invention
Fig. 1 is the schematic illustration of the embodiment of the present invention.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and embodiments.
It is noted that described further below be all exemplary, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms used herein has usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular
Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
As shown in Figure 1, present embodiments providing a kind of cross-cutting emotional semantic classification system based on text representation study, specifically
Include:
One Text Pretreatment module obtains source domain and target domain for carrying out characterization to cross-domain texts
Original text vector；
One text representation study module receives the output of Text Pretreatment module, the character representation adapted to for learning areas
Space excavates source domain and the potential general field feature of target domain and affective characteristics, obtains source domain and target domain
Text eigenvector；
One text representation reinforced module generates confrontation network model and is used to automatically generate the source domain text representation of robust and sentences
Its other emotional category constantly updates the parameter with optimization text representation study module according to error；
One emotional category output module, according to the Text eigenvector and benefit of the target domain of text representation study module output
The text emotion classification results of target domain are obtained with classification function.
In the present embodiment, the Text Pretreatment module extracts source domain and target domain text using the N-gram syntax
Feature, and using edge stack noise reduction self-encoding encoder study cross-domain texts character representation, the emotion without target domain
Label.
Preferably, in the present embodiment, edge stacks noise reduction self-encoding encoder using less calculation amount and has to higher-dimension
The scalability of feature realizes the abstract of domain features.
Specifically, in Text Pretreatment module, since the input data of neural network is generally vector, so as to model
End-to-end training, it is therefore desirable to which vectorization expression is carried out to text data.It is pre- in text for the ease of the processing and analysis of data
Processing module, the present embodiment first segment the text of source domain and target domain and filter stop words；Then, text is extracted
Text data is converted into vector form from textual form by this unigram/bigram feature；It is last more robust in order to obtain
Character representation, training edge stack noise reduction self-encoding encoder to obtain the original text vector of source domain and target domain.
In the present embodiment, the text representation study module carries out the character representation that field adapts to using neural network
Study, while considering domain features and affective characteristics in different field text.
In the present embodiment, the neural network abstracts text feature, to obtain the feature of field adaptation
Indicate vector, i.e. source domain Text eigenvector and target domain Text eigenvector.
Specifically, the text representation study module is a multilayer perceptron, by optimizing the weight matrix of network, from
And the abstract vector for capturing text indicates.
In the present embodiment, building generates confrontation network model in the text representation reinforced module, it is contemplated that text table
Show noise characteristic problem present in study module.
In the present embodiment, the generation confrontation network model includes generating network and differentiating network, fights and learns in the two
Strengthen text representation space in habit；The generation network synthesizes dummy copy by the way that noise vector is added in text representation vector
To confuse the judgement for differentiating network, differentiates that network carries out the judgement of emotional semantic classification and true and false sample simultaneously, advanced optimize text
Indicate study module.
In the present embodiment, the generation network generates random noise vector using normal distribution, the synthesis of generation to
Amount is by the optimization of module closer to source domain sample.
In the present embodiment, the judgement for differentiating network and carrying out emotional semantic classification and true and false sample simultaneously, it is contemplated that field
Feature and affective characteristics weigh two factors to the influence degree of result to the percentage contribution of cross-cutting emotional semantic classification.
Specifically, the text representation reinforced module is a generation confrontation network model, by generation network and net is differentiated
Two parts of network form.The core for generating network be the noise vector generated at random using normal distribution and text representation vector into
Row splicing, the dummy copy text vector abstracted after feature extraction differentiate network to the false sample for generating network generation
This carries out true and false judgement with source domain authentic specimen, and is predicted the emotional category of source domain sample and calculate itself and reality
The error of border affective tag optimizes update to the parameter of text representation study module using stochastic gradient and back-propagating,
In generating network and the generation and confrontation study that differentiate network, to achieve the purpose that strengthen text representation, make text representation
The character representation space that the field of study module adapts to has more robustness.
In the present embodiment, the emotional category output module using softmax function to obtained text representation to
Amount is handled, and predicts the emotional category of each text.
Specifically, text representation study module has learnt the text representation of target domain and source domain, emotional category output
Module calculates gained vector using softmax classification function one by one, obtains the emotion of text expression according to the threshold value of setting
Class prediction value.In the training stage, the prediction of emotional category is carried out using the text representation of source domain and calculates itself and practical feelings
The error for feeling label, is iterated update to the parameter of whole system using stochastic gradient descent method and back-propagating；Otherwise, right
The text representation of target domain carries out the prediction of emotional category, and exports predicted value.
The present embodiment additionally provides a kind of based on the cross-cutting emotional semantic classification system described above based on text representation study
The method of system, specifically includes the following steps:
Step S1: the Text Pretreatment module receives source domain text data and affective tag, target domain text data, right
Cross-domain texts carry out characterization, obtain the original text vector of source domain and target domain；
Step S2: the text representation study module adapts to the output of Text Pretreatment module as input, learning areas
Source domain and the potential general field feature of target domain and affective characteristics are excavated in character representation space, obtain source domain with
The Text eigenvector of target domain；
Step S3: the text representation reinforced module generates the source domain text table that confrontation network model is used to automatically generate robust
Show and differentiate its emotional category, the parameter with optimization text representation study module is constantly updated according to error；
Step S4: the emotional category output module receive the text feature of the text representation study module output after optimization to
Amount, and the text emotion classification results of target domain are obtained using classification function.
Particularly, in the present embodiment, in the training stage of model, according to the backward biography of the propagated forward of information and error
Broadcasting constantly to be adjusted them, successive optimization objective function.
The foregoing is merely presently preferred embodiments of the present invention, all equivalent changes done according to scope of the present invention patent with
Modification, is all covered by the present invention.
Claims (10)
1. a kind of cross-cutting emotional semantic classification system based on text representation study, it is characterised in that: including
One Text Pretreatment module obtains source domain and target domain for carrying out characterization to cross-domain texts
Original text vector；
One text representation study module receives the output of Text Pretreatment module, the character representation adapted to for learning areas
Space excavates source domain and the potential general field feature of target domain and affective characteristics, obtains source domain and target domain
Text eigenvector；
One text representation reinforced module generates confrontation network model and is used to automatically generate the source domain text representation of robust and sentences
Its other emotional category constantly updates the parameter with optimization text representation study module according to error；
One emotional category output module, according to the Text eigenvector and benefit of the target domain of text representation study module output
The text emotion classification results of target domain are obtained with classification function.
2. a kind of cross-cutting emotional semantic classification system based on text representation study according to claim 1, it is characterised in that:
The Text Pretreatment module extracts the feature of source domain and target domain text using the N-gram syntax, and is stacked using edge
Noise reduction self-encoding encoder learns the character representation of cross-domain texts, the affective tag without target domain.
3. a kind of cross-cutting emotional semantic classification system based on text representation study according to claim 1, it is characterised in that:
The text representation study module is learnt using the character representation that neural network adapts to field, while considering different necks
Domain features and affective characteristics in the text of domain.
4. a kind of cross-cutting emotional semantic classification system based on text representation study according to claim 3, it is characterised in that:
The neural network abstracts text feature, to obtain the character representation vector of field adaptation, i.e. source domain text
Feature vector and target domain Text eigenvector.
5. a kind of cross-cutting emotional semantic classification system based on text representation study according to claim 1, it is characterised in that:
Building generates confrontation network model in the text representation reinforced module, it is contemplated that noise present in text representation study module
Characteristic Problem.
6. a kind of cross-cutting emotional semantic classification system based on text representation study according to claim 5, it is characterised in that:
The generation confrontation network model includes generating network and differentiating network, strengthens text representation space in the two confrontation study；
The generation network by text representation vector be added noise vector synthesize dummy copy with confuse differentiate network judgement,
Differentiate that network carries out the judgement of emotional semantic classification and true and false sample simultaneously, advanced optimizes text representation study module.
7. a kind of cross-cutting emotional semantic classification system based on text representation study according to claim 6, it is characterised in that:
The generation network generates random noise vector using normal distribution, and the composite vector of generation is closer by the optimization of module
Source domain sample.
8. a kind of cross-cutting emotional semantic classification system based on text representation study according to claim 6, it is characterised in that:
The judgement for differentiating network and carrying out emotional semantic classification and true and false sample simultaneously, it is contemplated that domain features and affective characteristics are to cross-cutting
The percentage contribution of emotional semantic classification weighs two factors to the influence degree of result.
9. a kind of cross-cutting emotional semantic classification system based on text representation study according to claim 1, it is characterised in that:
The emotional category output module is handled obtained text representation vector using softmax function, predicts each text
Emotional category.
10. a kind of based on the described in any item cross-cutting emotional semantic classification systems based on text representation study of claim 1-9
Method, it is characterised in that: the following steps are included:
Step S1: the Text Pretreatment module receives source domain text data and affective tag, target domain text data, right
Cross-domain texts carry out characterization, obtain the original text vector of source domain and target domain；
Step S2: the text representation study module adapts to the output of Text Pretreatment module as input, learning areas
Source domain and the potential general field feature of target domain and affective characteristics are excavated in character representation space, obtain source domain with
The Text eigenvector of target domain；
Step S3: the text representation reinforced module generates the source domain text table that confrontation network model is used to automatically generate robust
Show and differentiate its emotional category, the parameter with optimization text representation study module is constantly updated according to error；
Step S4: the emotional category output module receive the text feature of the text representation study module output after optimization to
Amount, and the text emotion classification results of target domain are obtained using classification function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910048910.XA CN109783644A (en) | 2019-01-18 | 2019-01-18 | A kind of cross-cutting emotional semantic classification system and method based on text representation study |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910048910.XA CN109783644A (en) | 2019-01-18 | 2019-01-18 | A kind of cross-cutting emotional semantic classification system and method based on text representation study |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109783644A true CN109783644A (en) | 2019-05-21 |
Family
ID=66500884
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910048910.XA Pending CN109783644A (en) | 2019-01-18 | 2019-01-18 | A kind of cross-cutting emotional semantic classification system and method based on text representation study |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109783644A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110489753A (en) * | 2019-08-15 | 2019-11-22 | 昆明理工大学 | Improve the corresponding cross-cutting sensibility classification method of study of neuromechanism of feature selecting |
| CN110727844A (en) * | 2019-10-21 | 2020-01-24 | 东北林业大学 | Online commented commodity feature viewpoint extraction method based on generation countermeasure network |
| CN110796199A (en) * | 2019-10-30 | 2020-02-14 | 腾讯科技（深圳）有限公司 | Image processing method and device and electronic medical 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 |
| CN111998936A (en) * | 2020-08-25 | 2020-11-27 | 四川长虹电器股份有限公司 | Equipment abnormal sound detection method and system based on transfer learning |
| CN112199505A (en) * | 2020-10-30 | 2021-01-08 | 福州大学 | Cross-domain emotion classification method and system based on feature representation learning |
| 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 |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106096004A (en) * | 2016-06-23 | 2016-11-09 | 北京工业大学 | A kind of method setting up extensive cross-domain texts emotional orientation analysis framework |
| US20180165554A1 (en) * | 2016-12-09 | 2018-06-14 | The Research Foundation For The State University Of New York | Semisupervised autoencoder for sentiment analysis |
| CN108470061A (en) * | 2018-03-26 | 2018-08-31 | 福州大学 | A kind of emotional semantic classification system for visual angle grade text |
| US20180314716A1 (en) * | 2017-04-27 | 2018-11-01 | Sk Telecom Co., Ltd. | Method for learning cross-domain relations based on generative adversarial networks |
| CN108846128A (en) * | 2018-06-30 | 2018-11-20 | 合肥工业大学 | A kind of cross-domain texts classification method based on adaptive noise encoder |
| CN109213876A (en) * | 2018-08-02 | 2019-01-15 | 宁夏大学 | Based on the cross-module state search method for generating confrontation network |
-
2019
- 2019-01-18 CN CN201910048910.XA patent/CN109783644A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106096004A (en) * | 2016-06-23 | 2016-11-09 | 北京工业大学 | A kind of method setting up extensive cross-domain texts emotional orientation analysis framework |
| US20180165554A1 (en) * | 2016-12-09 | 2018-06-14 | The Research Foundation For The State University Of New York | Semisupervised autoencoder for sentiment analysis |
| US20180314716A1 (en) * | 2017-04-27 | 2018-11-01 | Sk Telecom Co., Ltd. | Method for learning cross-domain relations based on generative adversarial networks |
| CN108470061A (en) * | 2018-03-26 | 2018-08-31 | 福州大学 | A kind of emotional semantic classification system for visual angle grade text |
| CN108846128A (en) * | 2018-06-30 | 2018-11-20 | 合肥工业大学 | A kind of cross-domain texts classification method based on adaptive noise encoder |
| CN109213876A (en) * | 2018-08-02 | 2019-01-15 | 宁夏大学 | Based on the cross-module state search method for generating confrontation network |
Non-Patent Citations (3)
| Title |
|---|
| SWAMI SANKARANARAYANAN: "Generate To Adapt: Aligning Domains using Generative Adversarial Networks", 《2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| 廖祥文: "结合表示学习和迁移学习的跨领域情感分类", 《北京大学学报(自然科学版)》 * |
| 杨帅: "用于域适应的多边缘降噪自动编码器", 《计算机科学与探索》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110489753A (en) * | 2019-08-15 | 2019-11-22 | 昆明理工大学 | Improve the corresponding cross-cutting sensibility classification method of study of neuromechanism of feature selecting |
| CN110489753B (en) * | 2019-08-15 | 2022-06-14 | 昆明理工大学 | Neural structure corresponding learning cross-domain emotion classification method for improving feature selection |
| CN110727844A (en) * | 2019-10-21 | 2020-01-24 | 东北林业大学 | Online commented commodity feature viewpoint extraction method based on generation countermeasure network |
| CN110727844B (en) * | 2019-10-21 | 2022-07-01 | 东北林业大学 | Online commented commodity feature viewpoint extraction method based on generation countermeasure network |
| CN110796199A (en) * | 2019-10-30 | 2020-02-14 | 腾讯科技（深圳）有限公司 | Image processing method and device and electronic medical 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 |
| CN111998936A (en) * | 2020-08-25 | 2020-11-27 | 四川长虹电器股份有限公司 | Equipment abnormal sound detection method and system based on transfer learning |
| CN112199505A (en) * | 2020-10-30 | 2021-01-08 | 福州大学 | Cross-domain emotion classification method and system based on feature representation learning |
| CN112199505B (en) * | 2020-10-30 | 2022-06-03 | 福州大学 | Cross-domain emotion classification method and system based on feature representation learning |
| 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 |
| CN113779249B (en) * | 2021-08-31 | 2022-08-16 | 华南师范大学 | Cross-domain text emotion classification method and device, storage medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109783644A (en) | A kind of cross-cutting emotional semantic classification system and method based on text representation study | |
| CN109213995B (en) | Cross-language text similarity evaluation technology based on bilingual word embedding | |
| Khan et al. | A survey on the state-of-the-art machine learning models in the context of NLP | |
| Cui et al. | Consensus attention-based neural networks for Chinese reading comprehension | |
| US9779085B2 (en) | Multilingual embeddings for natural language processing | |
| CN110569508A (en) | Method and system for classifying emotional tendencies by fusing part-of-speech and self-attention mechanism | |
| CN110287494A (en) | A method of the short text Similarity matching based on deep learning BERT algorithm | |
| CN107943911A (en) | Data pick-up method, apparatus, computer equipment and readable storage medium storing program for executing | |
| CN110245229A (en) | A kind of deep learning theme sensibility classification method based on data enhancing | |
| CN112668319B (en) | Vietnamese news event detection method based on Chinese information and Vietnamese statement method guidance | |
| CN106599032A (en) | Text event extraction method in combination of sparse coding and structural perceptron | |
| CN106682089A (en) | RNNs-based method for automatic safety checking of short message | |
| CN109710769A (en) | A kind of waterborne troops's comment detection system and method based on capsule network | |
| CN110232123A (en) | The sentiment analysis method and device thereof of text calculate equipment and readable medium | |
| CN113157859B (en) | Event detection method based on upper concept information | |
| CN112883714B (en) | ABSC task syntactic constraint method based on dependency graph convolution and transfer learning | |
| CN109726745A (en) | A kind of sensibility classification method based on target incorporating description knowledge | |
| CN110347836A (en) | Incorporate the more bilingual the emotion of news classification method of the Chinese of viewpoint sentence feature | |
| CN111538841B (en) | Comment emotion analysis method, device and system based on knowledge mutual distillation | |
| Li et al. | Definition extraction with lstm recurrent neural networks | |
| CN116578705A (en) | Microblog emotion classification method based on pre-training language model and integrated neural network | |
| Marulli et al. | A comparison of character and word embeddings in bidirectional LSTMs for POS tagging in Italian | |
| CN109670169B (en) | Deep learning emotion classification method based on feature extraction | |
| Lê et al. | On Vietnamese sentiment analysis: A transfer learning method | |
| CN110569355A (en) | Viewpoint target extraction and target emotion classification combined method and system based on word blocks |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190521 |
|
| RJ01 | Rejection of invention patent application after publication |