CN109885670A - A kind of interaction attention coding sentiment analysis method towards topic text - Google Patents
A kind of interaction attention coding sentiment analysis method towards topic text Download PDFInfo
- Publication number
- CN109885670A CN109885670A CN201910112764.2A CN201910112764A CN109885670A CN 109885670 A CN109885670 A CN 109885670A CN 201910112764 A CN201910112764 A CN 201910112764A CN 109885670 A CN109885670 A CN 109885670A
- Authority
- CN
- China
- Prior art keywords
- topic
- text
- vector
- topic text
- emotion
- 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
Landscapes
- Machine Translation (AREA)
Abstract
The present invention proposes step 1 a kind of interaction attention coding sentiment analysis method towards topic text pre-processes topic text；Step 2, topic text code is carried out；Step 3, the interactive attention mechanism of topic object and topic text encodes；Step 4, sentiment classification model training；Step 5, emotion is predicted, is carried out sentiment analysis to topic object, is obtained the best emotional category of adaptive model device parameter.
Description
Technical field
The present invention relates to a kind of deep learning analysis method of text data, relate generally to a kind of towards topic text
Interaction attention encodes sentiment analysis method.
Background technique
Currently, network social intercourse media increasingly become the Important Platform that public opinion is delivered, including microblogging comment, newly
Hear topic etc..The topic comment data generated for it carries out one that sentiment analysis is always natural language processing field and grinds
Study carefully hot spot.To hold, emotion and the attitude of netizen masses' speech monitor public sentiment, network environment maintenance plays important work
With.However, network text information is many and diverse, the method for being difficult to it manually obtains information characteristics.Therefore, how automatical and efficient
The viewpoint of the public becomes increasingly important with attitude in the identification network information, this has also pushed the development of sentiment analysis technology.
The underlying task of sentiment analysis is to extract feature significant in text, is extracted valuable to sentiment analysis
Information.So far, the method for sentiment analysis is broadly divided into following three aspects.Sentiment analysis based on sentiment dictionary is most
Based on sentiment analysis method, emotion word, negative word and the degree adverb etc. relied in dictionary calculate feeling polarities
Value, in this, as the foundation of text emotion tendency.Sentiment analysis based on machine learning is an important breakthrough, is borrowed
The feature extraction of natural language processing tool, as the methods of naive Bayesian, support vector machines realize the feelings from feature level
Feel classifying and analyzing method, and final classification effect depends on the feature extraction of training text and the correctness of Emotion tagging.
With the rapid development of artificial intelligence study, the sentiment analysis based on deep learning is current main stream approach.Benefit
Word embedded technology is by text vector.Then it constructs deep learning model learning text feature, classify.Basis and allusion quotation
The sentiment analysis deep learning model of type includes: shot and long term memory models (LSTM/GRU, Bi-LSTM/Bi-GRU) and various changes
Improve in cataloged procedure into Recognition with Recurrent Neural Network model and remembers long-term Dependence Problem；Attention mechanism (attention mechanism)
Different weights can be assigned to the different piece of text, extracts key message, and model is enable to make more accurate judgement,
The expense in calculating and storage is not will increase simultaneously.To sum up, based on the neural network method of deep learning in certain journey
The two big main problems for optimizing the validity and accuracy of sentiment analysis on degree, but also facing: first is that the length of sequence text
Dependence Problem causes semantic disappearance；Second is that the analysis of topic text emotion does not fully consider the correlation of topic with text.Though vocabulary
The essential information of emotion can so be described, but single vocabulary lacks object, lacks correlation degree, and different vocabulary groups
It is all opposite to be combined the different even Sentiment orientations of obtained emotion degree.
The research of current sentiment analysis is mainly used in product scope, and the definition for viewpoint is a 5 tuple (ei,
aij,sijkl,hk,tl), wherein eiIt is entity, aijBe the entity in a certain respect, sijklIt is the feelings for this aspect of the entity
Sense, hkIt is viewpoint holder, tlIt is that viewpoint delivers the time.For this research, occur for entity some aspect or
The more fine-grained sentiment analysis of some object of person.Such as laptop, for some aspect " screen " of computer
Sentiment analysis, such as dining room carry out sentiment analysis research for the specific object " sushi " of the food in dining room.
In view of this thinking, the corresponding topic of topic text can be understood as topic existing for topic text, that
It is the sentiment analysis for being directed to a specific object i.e. topic itself that fine-grained sentiment analysis, which can be derived as,.To this
Invention is defined topic sentiment analysis, and wherein topic refers to specific entity or theme, and topic text refers to table
Text up to where the news of topic object out.And our task is to carry out sentiment analysis to topic itself, it is seen that emotion point
The granularity minimum for analysing object is vocabulary, but the most basic unit for expressing an emotion is then sentence.
Summary of the invention
In view of the above problems, the invention proposes a kind of, the interaction attention towards topic text encodes sentiment analysis side
Method, the present invention is the following steps are included: step 1, pre-processes topic text as shown in Figure 1；Step 2, topic text is carried out
Coding；Step 3, the interactive attention mechanism of topic object and topic text encodes；Step 4, sentiment classification model training；
Step 5, emotion is predicted, is carried out sentiment analysis to topic object, is obtained the best emotional category of adaptive model device parameter.
The beneficial effects of the present invention are: 1, need manually to extract semantic feature, present invention emotional semantic classification end to end,
Last feeling polarities are directly obtained by trained model to the topic text of input.2, the present invention is using topic as feelings
The object for feeling analysis, belongs to more fine-grained sentiment analysis.Present invention employs encoding from attention, text vocabulary is excavated
Between relationship, improve long text semantic dependency problem, and model the emotional expression of topic and context.3, the present invention combines
Two-way GRU encodes topic text.Merely using two-way GRU can integrating context information, but cannot embody with
Relationship between other words；It is simple to use the sequence signature then lost between word from attention mechanism.The present invention is double
The advantages of on the basis of to GRU using the two has been effectively combined from attention mechanism, so that the expression semanteme of sequence of terms is more
Add the accuracy rate of abundant, trained model higher.4, after the present invention respectively encodes topic text and topic with two-way GRU
The hidden state vector of forward direction and backward hidden state vector are done and are spliced, and the expression of topic text and the expression of topic object are respectively obtained,
And interactive attention mechanism is set, encode the correlation between topic text and topic object.Effectively excavate topic object with
Contain the related emotion information between text.
Detailed description of the invention
Fig. 1 is overall flow figure of the invention；
Fig. 2 is the two-way GRU topic text emotion analysis model based on interaction attention mechanism
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments,
The present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only used to explain this hair
It is bright, it is not intended to limit the present invention.In addition, technology involved in the various embodiments of the present invention described below is special
Sign can be combined with each other as long as they do not conflict with each other.
The technical solution adopted by the present invention is that the topic text emotion analysis method based on the interaction two-way GRU of attention, such as
Fig. 2 show the two-way GRU topic text emotion analysis model based on interaction attention mechanism, comprising the following steps:
Step 1, topic Text Pretreatment,
Step 1-1: the disclosure data with emotion and topic mark are pre-processed, rushing in data is removed
Prominent item.
Step 1-2: segmenting topic text and topic object, converts height for word using Word2vec interface
Dimensional vector respectively obtains the term vector coded representation of topic text and topic object.
Step 2, topic text code
Step 2-1, in order to alleviate subsequent two-way GRU cataloged procedure long text semantic dependency problem, the present invention first will words
Inscribe text encode from attention, obtain the relevance of word word, improve to a certain extent long sequence text it is semantic according to
Rely problem.Here the term vector of insertion is input in Multi-head Attention (bull attention) encoder, bull
Attention encoder integrates the multiple coding from attention mechanism, can encode the relevance letter of the high-order semanteme of word and word
Breath.Finally output obtains the weighting expression of each word of topic text.
EW=MultiHeadAttention (QWQ, KWK, VWV) (1)
Wherein, (Query) vector is inquiredKeyword (Key) vectorBe worth (Value) to
Amount(n indicates that text size, d indicate the dimension of term vector), EWFor network output.
Step 2-2, by treated, topic text vector is input in Bi-GRU neural network, carries out sequential coding.
Wherein, i indicates i-th current of word,The term vector of current word, as the input for Bi-GRU,For
The forward coding vector of i-th of word of topic text,Indicate its backward coding vector.During forward coding, from 1 moment to
T moment forward direction calculates, and the forward coding vector output at each moment is obtained and save, such as shown in (2).In backward cataloged procedure,
From t moment to 1 moment retrospectively calculate, the backward coding vector output at each moment is obtained and saves, such as shown in (3).
Topic object vectors are input in Bi-GRU neural network by step 2-3, carry out sequential coding.
Wherein, yiFor the input of Bi-GRU,For the forward coding vector of topic object,After topic object
To coding vector.
Step 3, the interactive attention mechanism of topic object and topic text encodes
After the Bi-GRU coding for obtaining text and topic object, the emotion in text is pair using topic as expression
As the semanteme of the semanteme and text of topic should have interactional effect.The present invention establishes a kind of topic object and text
Interactive encoding model, the illustraton of model is shown in Fig. 2.It is this coding using between topic object and text influence each other as
It was found that the important clue of related affective characteristics.
Step 3-1, the present invention are last by the hidden vector sum of the last one forward direction after topic text Bi-GRU coding respectively
One backward hidden vector does the sentence expression spliced as entire topic text.Secondly same place will be done to topic object
Reason obtains the expression of topic object.
Step 3-2 goes coding topic object to indicate the incidence relation between text using interactive attention mechanism,
Attention weight distribution indicates the correlation that each hidden vector of topic text is indicated with topic object:
aiFor attention weight, wherein s () is the scoring functions of additivity or multiplying property, in above-mentioned formula:
Wherein WTgThe weight parameter matrix learnt, b are needed for modelTgFor biasing.
The present invention obtains the study weight matrix of topic object expression and the hidden state vector of text using bilinear function.
Tanh is nonlinear function, TgTThe transposition of vector is indicated for topic object.
Step 3-3: likewise, going to the pass between coding topic text representation and topic using interactive attention mechanism
Connection relationship.
Wherein, MsTFor the transposition of topic object text representation vector.WMSThe weight matrix learnt, b are needed for modelMSFor
Biasing.Step 3-4: after calculating the interactive attention weight between topic text and topic object, ultimatum topic is obtained
The sentence vector of text indicates that the vector of Ms ' and topic object indicates Tg '.
Step 3-5: carrying out emotional semantic classification, and obtaining the final of sentence vector indicates after indicating with the vector of topic object, will
The two vectors are spliced together the expression vector as emotional semantic classification text, are denoted as R, this vector includes topic text
With the related information of topic object and the semantic information of emotion.Defining P (Ms, Tg) is the probability that model carries out emotional semantic classification.
Three classification (neutral, positive, passive) of emotion are carried out using softmax classifier.
P (Ms, Tg)=sof tmax (MLP (Wp·R+bp)) (14)
We use MLP (multi-layer perception (MLP)) and are used as nonlinear activation function, wherein Wp, bpFor the ginseng of model learning
Number.
Step 4, sentiment classification model training
Model training, training data and test data can be initial data is randomly ordered, do training by 80%,
20%, which does the method tested, separates.Model training is until obtain best appraisement system index.
It is implemented as in the step 4, the sentiment analysis data set of topic text is done into training by 80%,
20%, which does the method tested, separates, and wherein training data and test data can be initial data is randomly ordered, select depth
AdamOptimizer optimizer in study carries out model training, until obtaining best model evaluating system index.
Step 5, emotion is predicted
It is input with open field topic text by trained model equipment, sentiment analysis is carried out to topic object,
The best emotional category of available adaptive model device parameter.
It is implemented as in the step 5, is defeated with open field topic text by trained model equipment
Enter, including input topic text and topic object, model goes prediction to obtain the feelings for emotion object by the parameter succeeded in school
Inductance value, the emotional value are divided into neutrality, positive, passive, choose the highest emotional value classification of probability as analysis result.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；To the greatest extent
Present invention has been described in detail with reference to the aforementioned embodiments for pipe, those skilled in the art should understand that: it is still
It can modify to technical solution documented by previous embodiment or equivalent replacement of some of the technical features；
And these are modified or replaceed, the spirit for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution
And range.
Claims (6)
1. a kind of interaction attention towards topic text encodes sentiment analysis method, which is characterized in that step 1, to topic text
This is pre-processed；Step 2, topic text code is carried out；Step 3, the interactive attention machine of topic object and topic text
System coding；Step 4, sentiment classification model training；Step 5, emotion is predicted, is carried out sentiment analysis to topic object, is adapted to
The best emotional category of model equipment parameter.
2. the method as described in claim 1, which is characterized in that in the step 1, in the tool of the topic Text Pretreatment
Body is embodied as, step 1-1, pre-processes to the text data with emotion and topic mark, removes the conflict item in data；
Step 1-2: segmenting topic text and topic object, and convert high dimension vector for word, respectively obtains topic text
With the term vector coded representation of topic object.
3. method according to claim 2, which is characterized in that in the step 2, the topic text code specific implementation
For the topic text encode from attention, obtains the relevance of word word by step 2-1；Step 2-2, by the volume
Topic text vector after code is input in Bi-GRU neural network, carries out semantic coding；Step 2-3, by the topic object
Vector is input in Bi-GRU neural network, carries out semantic coding.
4. method as claimed in claim 3, which is characterized in that in being implemented as the step 3, establish a kind of topic pair
As the interactive encoding model with text, following steps are specifically included, step 3-1 respectively compiles topic text in the sequence
The last one backward hidden vector of the hidden vector sum of the last one forward direction after code does the sentence spliced as entire topic text
Indicate, then by topic object the last one forward direction after the sequential coding hidden vector sum after the last one to it is hidden to
Amount does the sentence expression spliced as entire topic object；Step 3-2 encodes the topic pair using interactive attention mechanism
Incidence relation between the expression and text of elephant indicates each hidden vector and topic of topic text using attention weight distribution
The correlation that object indicates；Step 3-3, using interactive attention mechanism go to encode the topic text expression and topic it
Between incidence relation；Step 3-4 is obtained to the end after calculating the interactive attention weight between topic text and topic object
Topic text sentence vector indicate and topic object vector indicate；Step 3-5: emotional semantic classification, the emotion point are carried out
The mode of class is to indicate to be spliced together with the expression of the vector of the topic object as emotion by the final of the sentence vector
The expression vector of classifying text.
5. method as claimed in claim 4, which is characterized in that be implemented as in the step 4, by topic text
Sentiment analysis data set does training by 80%, and 20%, which does the method tested, separates, and wherein training data and test data can be
Initial data is randomly ordered, and the AdamOptimizer optimizer selected in deep learning carries out model training, until obtaining most
Good model evaluating system index.
6. method as claimed in claim 5, which is characterized in that be implemented as in the step 5, by trained
Model equipment, is input, including input topic text and topic object with open field topic text, and model passes through the ginseng succeeded in school
Number goes prediction to obtain the emotional value for emotion object, and the emotional value is divided into neutrality, positive, passive, and it is highest to choose probability
Emotional value classification is as analysis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910112764.2A CN109885670A (en) | 2019-02-13 | 2019-02-13 | A kind of interaction attention coding sentiment analysis method towards topic text |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910112764.2A CN109885670A (en) | 2019-02-13 | 2019-02-13 | A kind of interaction attention coding sentiment analysis method towards topic text |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109885670A true CN109885670A (en) | 2019-06-14 |
Family
ID=66928122
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910112764.2A Pending CN109885670A (en) | 2019-02-13 | 2019-02-13 | A kind of interaction attention coding sentiment analysis method towards topic text |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109885670A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334190A (en) * | 2019-07-12 | 2019-10-15 | 电子科技大学 | A kind of reply automatic generation method towards open field conversational system |
| CN110334189A (en) * | 2019-07-11 | 2019-10-15 | 河南大学 | Method is determined based on the long microblog topic label in short-term and from attention neural network |
| CN110781306A (en) * | 2019-10-31 | 2020-02-11 | 山东师范大学 | English text aspect layer emotion classification method and system |
| CN110991190A (en) * | 2019-11-29 | 2020-04-10 | 华中科技大学 | Document theme enhanced self-attention network, text emotion prediction system and method |
| CN111259142A (en) * | 2020-01-14 | 2020-06-09 | 华南师范大学 | Specific target emotion classification method based on attention coding and graph convolution network |
| CN111274372A (en) * | 2020-01-15 | 2020-06-12 | 上海浦东发展银行股份有限公司 | Method, electronic device, and computer-readable storage medium for human-computer interaction |
| CN111414749A (en) * | 2020-03-18 | 2020-07-14 | 哈尔滨理工大学 | Social text dependency syntactic analysis system based on deep neural network |
| CN111460115A (en) * | 2020-03-17 | 2020-07-28 | 深圳市优必选科技股份有限公司 | Intelligent man-machine conversation model training method, model training device and electronic equipment |
| CN111522917A (en) * | 2020-04-21 | 2020-08-11 | 京东数字科技控股有限公司 | Dialogue emotion detection method and device, storage medium and electronic equipment |
| CN111581966A (en) * | 2020-04-30 | 2020-08-25 | 华南师范大学 | Context feature fusion aspect level emotion classification method and device |
| CN111753093A (en) * | 2020-07-02 | 2020-10-09 | 东北电力大学 | Method and device for evaluating level of network public opinion crisis |
| CN112100325A (en) * | 2020-08-28 | 2020-12-18 | 广州探迹科技有限公司 | Man-machine conversation emotion analysis method based on hierarchical gating circulation unit |
| CN113469479A (en) * | 2020-03-31 | 2021-10-01 | 阿里巴巴集团控股有限公司 | Contract risk prediction method and device |
| CN113536773A (en) * | 2021-07-20 | 2021-10-22 | 北京明略软件系统有限公司 | Commodity comment sentiment analysis method and system, electronic equipment and storage medium |
| CN113781398A (en) * | 2021-08-13 | 2021-12-10 | 浙江卡易智慧医疗科技有限公司 | Bone age identification method and model based on bidirectional gating circulation network |
| CN113806471A (en) * | 2020-06-17 | 2021-12-17 | 广东博智林机器人有限公司 | Text matching method and device, electronic equipment and storage medium |
| CN115910345A (en) * | 2022-12-22 | 2023-04-04 | 广东数业智能科技有限公司 | Mental health assessment intelligent early warning method and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016162357A (en) * | 2015-03-04 | 2016-09-05 | Kddi株式会社 | Analysis device and program of user's emotion to product |
| CN108133038A (en) * | 2018-01-10 | 2018-06-08 | 重庆邮电大学 | A kind of entity level emotional semantic classification system and method based on dynamic memory network |
| KR20180091496A (en) * | 2017-02-07 | 2018-08-16 | 주식회사 에스엘커뮤니케이션즈 | Method for public opinion making using social network based on emotion analysys |
| CN108804417A (en) * | 2018-05-21 | 2018-11-13 | 山东科技大学 | A kind of documentation level sentiment analysis method based on specific area emotion word |
| CN109145304A (en) * | 2018-09-07 | 2019-01-04 | 中山大学 | A kind of Chinese Opinion element sentiment analysis method based on word |
| CN109271522A (en) * | 2018-11-20 | 2019-01-25 | 深圳大学 | Comment sensibility classification method and system based on depth mixed model transfer learning |
| CN109325109A (en) * | 2018-08-27 | 2019-02-12 | 中国人民解放军国防科技大学 | Attention encoder-based extraction type news abstract generating device |
-
2019
- 2019-02-13 CN CN201910112764.2A patent/CN109885670A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016162357A (en) * | 2015-03-04 | 2016-09-05 | Kddi株式会社 | Analysis device and program of user's emotion to product |
| KR20180091496A (en) * | 2017-02-07 | 2018-08-16 | 주식회사 에스엘커뮤니케이션즈 | Method for public opinion making using social network based on emotion analysys |
| CN108133038A (en) * | 2018-01-10 | 2018-06-08 | 重庆邮电大学 | A kind of entity level emotional semantic classification system and method based on dynamic memory network |
| CN108804417A (en) * | 2018-05-21 | 2018-11-13 | 山东科技大学 | A kind of documentation level sentiment analysis method based on specific area emotion word |
| CN109325109A (en) * | 2018-08-27 | 2019-02-12 | 中国人民解放军国防科技大学 | Attention encoder-based extraction type news abstract generating device |
| CN109145304A (en) * | 2018-09-07 | 2019-01-04 | 中山大学 | A kind of Chinese Opinion element sentiment analysis method based on word |
| CN109271522A (en) * | 2018-11-20 | 2019-01-25 | 深圳大学 | Comment sensibility classification method and system based on depth mixed model transfer learning |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334189A (en) * | 2019-07-11 | 2019-10-15 | 河南大学 | Method is determined based on the long microblog topic label in short-term and from attention neural network |
| CN110334190A (en) * | 2019-07-12 | 2019-10-15 | 电子科技大学 | A kind of reply automatic generation method towards open field conversational system |
| CN110781306A (en) * | 2019-10-31 | 2020-02-11 | 山东师范大学 | English text aspect layer emotion classification method and system |
| CN110991190A (en) * | 2019-11-29 | 2020-04-10 | 华中科技大学 | Document theme enhanced self-attention network, text emotion prediction system and method |
| CN111259142A (en) * | 2020-01-14 | 2020-06-09 | 华南师范大学 | Specific target emotion classification method based on attention coding and graph convolution network |
| CN111274372A (en) * | 2020-01-15 | 2020-06-12 | 上海浦东发展银行股份有限公司 | Method, electronic device, and computer-readable storage medium for human-computer interaction |
| CN111460115B (en) * | 2020-03-17 | 2023-05-26 | 深圳市优必选科技股份有限公司 | Intelligent man-machine conversation model training method, model training device and electronic equipment |
| CN111460115A (en) * | 2020-03-17 | 2020-07-28 | 深圳市优必选科技股份有限公司 | Intelligent man-machine conversation model training method, model training device and electronic equipment |
| CN111414749A (en) * | 2020-03-18 | 2020-07-14 | 哈尔滨理工大学 | Social text dependency syntactic analysis system based on deep neural network |
| CN111414749B (en) * | 2020-03-18 | 2022-06-21 | 哈尔滨理工大学 | Social text dependency syntactic analysis system based on deep neural network |
| CN113469479A (en) * | 2020-03-31 | 2021-10-01 | 阿里巴巴集团控股有限公司 | Contract risk prediction method and device |
| CN111522917A (en) * | 2020-04-21 | 2020-08-11 | 京东数字科技控股有限公司 | Dialogue emotion detection method and device, storage medium and electronic equipment |
| CN111581966A (en) * | 2020-04-30 | 2020-08-25 | 华南师范大学 | Context feature fusion aspect level emotion classification method and device |
| CN111581966B (en) * | 2020-04-30 | 2023-04-21 | 华南师范大学 | Context feature-fused aspect-level emotion classification method and device |
| CN113806471A (en) * | 2020-06-17 | 2021-12-17 | 广东博智林机器人有限公司 | Text matching method and device, electronic equipment and storage medium |
| CN111753093A (en) * | 2020-07-02 | 2020-10-09 | 东北电力大学 | Method and device for evaluating level of network public opinion crisis |
| CN112100325A (en) * | 2020-08-28 | 2020-12-18 | 广州探迹科技有限公司 | Man-machine conversation emotion analysis method based on hierarchical gating circulation unit |
| CN113536773A (en) * | 2021-07-20 | 2021-10-22 | 北京明略软件系统有限公司 | Commodity comment sentiment analysis method and system, electronic equipment and storage medium |
| CN113781398A (en) * | 2021-08-13 | 2021-12-10 | 浙江卡易智慧医疗科技有限公司 | Bone age identification method and model based on bidirectional gating circulation network |
| CN115910345A (en) * | 2022-12-22 | 2023-04-04 | 广东数业智能科技有限公司 | Mental health assessment intelligent early warning method and storage medium |
| CN115910345B (en) * | 2022-12-22 | 2023-06-27 | 广东数业智能科技有限公司 | Intelligent early warning method for psychological health assessment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109885670A (en) | A kind of interaction attention coding sentiment analysis method towards topic text | |
| CN111767741B (en) | Text emotion analysis method based on deep learning and TFIDF algorithm | |
| CN110245229B (en) | Deep learning theme emotion classification method based on data enhancement | |
| CN110134946B (en) | Machine reading understanding method for complex data | |
| CN109558487A (en) | Document Classification Method based on the more attention networks of hierarchy | |
| CN110532554A (en) | A kind of Chinese abstraction generating method, system and storage medium | |
| CN108846017A (en) | The end-to-end classification method of extensive newsletter archive based on Bi-GRU and word vector | |
| CN108829662A (en) | A kind of conversation activity recognition methods and system based on condition random field structuring attention network | |
| CN111310474A (en) | Online course comment sentiment analysis method based on activation-pooling enhanced BERT model | |
| CN110083710A (en) | It is a kind of that generation method is defined based on Recognition with Recurrent Neural Network and the word of latent variable structure | |
| Bhoir et al. | Comparative analysis of different word embedding models | |
| CN110598219A (en) | Emotion analysis method for broad-bean-net movie comment | |
| CN109614487A (en) | A method of the emotional semantic classification based on tensor amalgamation mode | |
| CN106682089A (en) | RNNs-based method for automatic safety checking of short message | |
| CN108108468A (en) | A kind of short text sentiment analysis method and apparatus based on concept and text emotion | |
| CN107679225A (en) | A kind of reply generation method based on keyword | |
| CN103473380A (en) | Computer text sentiment classification method | |
| CN112561718A (en) | Case microblog evaluation object emotion tendency analysis method based on BilSTM weight sharing | |
| CN110472245A (en) | A kind of multiple labeling emotional intensity prediction technique based on stratification convolutional neural networks | |
| CN116205222A (en) | Aspect-level emotion analysis system and method based on multichannel attention fusion | |
| CN113094502A (en) | Multi-granularity takeaway user comment sentiment analysis method | |
| CN110866087A (en) | Entity-oriented text emotion analysis method based on topic model | |
| CN114282592A (en) | Deep learning-based industry text matching model method and device | |
| Sujana et al. | Lida: Language-independent data augmentation for text classification | |
| Bharathi et al. | Machine Learning Based Approach for Sentiment Analysis on Multilingual Code Mixing Text. |
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: 20190614 |
|
| RJ01 | Rejection of invention patent application after publication |