CN109408805A - A kind of Tibetan language sentiment analysis method and system based on interacting depth study - Google Patents
A kind of Tibetan language sentiment analysis method and system based on interacting depth study Download PDFInfo
- Publication number
- CN109408805A CN109408805A CN201811045667.8A CN201811045667A CN109408805A CN 109408805 A CN109408805 A CN 109408805A CN 201811045667 A CN201811045667 A CN 201811045667A CN 109408805 A CN109408805 A CN 109408805A
- Authority
- CN
- China
- Prior art keywords
- tibetan language
- term vector
- feature
- sentiment analysis
- sentence
- 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
Classifications
-
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Abstract
The present invention relates to a kind of Tibetan language sentiment analysis method and system based on interacting depth study, and method includes S1, obtain Tibetan language sentence, and are pre-processed to obtain pretreatment Tibetan language to Tibetan language sentence；S2 carries out the Tibetan language after Tibetan language is segmented to pretreatment Tibetan language；S3 carries out preliminary treatment to the Tibetan language after participle and obtains term vector, and establishes term vector model according to term vector；Term vector model is introduced into three-layer coil product neural network and is iterated processing to extract the first term vector feature by S4；First term vector feature is introduced into two layers of shot and long term memory network and is iterated processing to extract the second term vector feature by S5；S6 is classified after being compiled the second term vector feature by classification function.By analyzing Tibetan language feature using convolutional layer and two layers of shot and long term memory network layer converged network, the global measurement of text can be retained and deeper semantic relation can be excavated, and preferable classifying quality can be obtained.
Description
Technical field
The present invention relates to computer application technology more particularly to a kind of Tibetan language emotions point based on interacting depth study
Analyse method and system.
Background technique
There are mainly three types of existing Tibetan language sentiment analysis methods, and one is based on sentiment dictionary: based on sentiment dictionary
The method of Tibetan language sentiment analysis is to judge Tibetan language feelings by the weight superposition calculation of emotion word in Tibetan language sentence or emotion phrase
Sense tendency.Adversative in Tibetan language sentence can change the Sentiment orientation of whole sentence, therefore carry out adversative processing to Tibetan language sentence.
Degree word in Tibetan language sentence can weaken or reinforce the intensity of whole sentence emotion word or emotion phrase, therefore will be in Tibetan language sentence
Degree word handled.Tibetan language Sentiment orientation is calculated finally by weight.Second of method based on machine learning,
If support vector machines (SVM) is the learning model for having supervision in machine learning field, commonly used to carry out pattern-recognition,
Classification and regression analysis.The third method from coding based on deep learning is one kind of neural network from coding, and
A kind of unsupervised learning algorithm, it has used back-propagation algorithm, and target value is allowed to be equal to input value.
However, above-mentioned Tibetan language sentiment analysis method has the following deficiencies: the Tibetan language sentiment analysis method based on sentiment dictionary
It is most basic method, when calculating Tibetan language sentence, will lead to its semantic serious loss, accuracy rate is low.Based on machine learning
Method solves Tibetan language matter of semantics to a certain extent, and accuracy rate is less desirable.Based on deep learning from coding be also
Solves Tibetan language matter of semantics to a certain extent, accuracy rate is also not highly desirable.And current domestic Tibetan language sentiment analysis method
Solves Tibetan language matter of semantics to a certain extent, but all less desirable in sentiment analysis effect and semantic structure.
Summary of the invention
In order to solve the above-mentioned technical problem the present invention provides a kind of Tibetan language sentiment analysis method based on interacting depth study.
The technical scheme to solve the above technical problems is that a kind of Tibetan language emotion point based on interacting depth study
Analysis method, including the following steps:
S1 obtains Tibetan language sentence, and pre-processes to the Tibetan language sentence, obtains pretreatment Tibetan language.
S2 carries out Tibetan language participle to the pretreatment Tibetan language, the Tibetan language after being segmented.
S3 carries out preliminary treatment to the Tibetan language after the participle, obtains term vector, and establish dimension according to the term vector
For the term vector model of 256 Tibetan language sentence.
The term vector model is introduced into three-layer coil product neural network and is iterated processing to extract the first word by S4
Vector characteristics, the dimension of the first term vector feature are 128.
The first term vector feature is introduced into two layers of shot and long term memory network and is iterated processing to extract by S5
Second term vector feature, the dimension of the second term vector feature are 64.
S6 is classified after being compiled the second term vector feature by classification function to obtain sentiment analysis knot
Fruit.
The invention has the advantages that by utilizing convolutional layer and two layers of shot and long term to remember in interacting deep learning model
Network (LSTM) layer converged network analyzes Tibetan language feature, not only retain text global measurement can excavate again it is deeper
Semantic relation, and preferable classifying quality can be obtained.
Based on the above technical solution, the present invention can also be improved as follows.
Further, the specific implementation of the S1 are as follows:
The Tibetan language sentence in network is crawled by web crawlers, and filter out the expression in the Tibetan language sentence, symbol,
Chinese sentence obtains pretreatment Tibetan language.
Beneficial effect using above-mentioned further scheme is, by crawling the Tibetan language sentence in network using web crawlers
The acquisition accuracy and efficiency of Tibetan language sentence can be improved in son；By filtering out the expression in the Tibetan language sentence, symbol, Chinese
Sentence is so as to subsequent Tibetan language sentiment analysis.
Further, the specific implementation of the S3 are as follows:
Tibetan language after the participle is introduced into word2vec term vector tool and is handled, term vector model is obtained.
Beneficial effect using above-mentioned further scheme can be used directly to by using word2vec term vector tool
Term vector model is constructed, term vector model construction efficiency is improved.
Further, the S5 specifically includes the following steps:
The activation primitive in two layers of shot and long term memory network is set Softmax by S51.
The parameter of Dropout in two layers of shot and long term memory network is set as 0.5 by S52.
Beneficial effect using above-mentioned further scheme is, by the way that the parameter of the Dropout in activation primitive to be set as
0.5, preferable emotional semantic classification effect can be obtained.
Further, the S6 specifically includes the following steps:
S61 is compiled the second term vector feature by compile function, obtains compiling secondary vector feature,
In, the loss function in the compile function is set as binary_crossentropy, and majorized function is set as Adam (lr=
0.01, beta_1=0.9, beta_2=0.999, epsilon=1e-08).
S62 classifies to obtain sentiment analysis to the compiling secondary vector feature by softsign classification function
As a result.
Beneficial effect using above-mentioned further scheme is, by the way that loss function is set as binary_crossentropy,
Majorized function is set as Adam (lr=0.01, beta_1=0.9, beta_2=0.999, epsilon=1e-08), can be obtained more
Good emotional semantic classification effect.
In order to solve the above-mentioned technical problem the present invention also provides a kind of Tibetan language sentiment analysis system based on interacting depth study
System.
The technical scheme to solve the above technical problems is that the Tibetan language sentiment analysis system based on interacting depth study
System, comprising:
Tibetan language preprocessing module is pre-processed for obtaining Tibetan language sentence, and to the Tibetan language sentence, is pre-processed
Tibetan language.
Word segmentation module, for carrying out Tibetan language participle to the pretreatment Tibetan language, the Tibetan language after being segmented.
Term vector model building module obtains term vector, and root for carrying out preliminary treatment to the Tibetan language after the participle
The term vector model for the Tibetan language sentence that dimension is 256 is established according to the term vector.
First iteration module is iterated processing for the term vector model to be introduced into three-layer coil product neural network
To extract the first term vector feature, the dimension of the first term vector feature is 128.
Secondary iteration module changes for the first term vector feature to be introduced into two layers of shot and long term memory network
To extract the second term vector feature, the dimension of the second term vector feature is 64 for generation processing.
Compiling and categorization module, for being classified after being compiled the second term vector feature by classification function
To obtain sentiment analysis result.
Further, the Tibetan language preprocessing module is specifically used for crawling the Tibetan language sentence in network by web crawlers,
And the expression in the Tibetan language sentence, symbol, Chinese sentence are filtered out, obtain pretreatment Tibetan language.
Further, the term vector model building module is specifically used for the Tibetan language after the participle being introduced into word2vec
It is handled in term vector tool, obtains term vector model.
Further, the secondary iteration module further include:
Activation primitive setting unit, for setting the activation primitive in two layers of shot and long term memory network to
Softmax。
Parameter set unit, for the parameter of the Dropout in two layers of shot and long term memory network to be set as 0.5.
Further, the compiling and categorization module include:
Compilation unit obtains compiling secondary vector for being compiled by compile function to the second term vector feature
Feature, wherein the loss function in the compile function is set as binary_crossentropy, and majorized function is set as Adam
(lr=0.01, beta_1=0.9, beta_2=0.999, epsilon=1e-08).
Taxon classifies to obtain emotion to the compiling secondary vector feature by softsign classification function
Analyze result.
Detailed description of the invention
Fig. 1 is the method flow diagram of the Tibetan language sentiment analysis based on interacting depth study of the embodiment of the present invention；
Fig. 2 is the interacting deep learning method illustraton of model of the embodiment of the present invention；
Fig. 3 is the structural schematic diagram of the Tibetan language sentiment analysis system based on interacting depth study of the embodiment of the present invention.
Specific embodiment
The principle and features of the present invention will be described below with reference to the accompanying drawings, and the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the invention.
As shown in Figure 1, the Tibetan language sentiment analysis method provided in an embodiment of the present invention based on interacting depth study, including with
Under several steps:
S1 obtains Tibetan language sentence, and pre-processes to the Tibetan language sentence, obtains pretreatment Tibetan language.
S2 carries out Tibetan language participle to the pretreatment Tibetan language, the Tibetan language after being segmented.
S3 carries out preliminary treatment to the Tibetan language after the participle, obtains term vector, and establish dimension according to the term vector
For the term vector model of 256 Tibetan language sentence.
The term vector model is introduced into three-layer coil product neural network and is iterated processing to extract the first word by S4
Vector characteristics, the dimension of the first term vector feature are 128.
The first term vector feature is introduced into two layers of shot and long term memory network and is iterated processing to extract by S5
Second term vector feature, the dimension of the second term vector feature are 64.
S6 is classified after being compiled the second term vector feature by classification function to obtain sentiment analysis knot
Fruit.
In practical application scene, firstly, crawling the Tibetan language sentence in network by web crawlers under Python environment
Tibetan language microblogging in son, such as Tencent, Sina weibo.Then Tibetan language sentence is pre-processed, is pre-processed primarily to going
Except expression, symbol, Chinese sentence etc., pure Tibetan language microblogging, that is, pretreatment Tibetan language are extracted, to facilitate subsequent Tibetan language emotion
Analysis.
It should be noted that Python is a coherent and powerful object oriented program language, it is similar
In Perl, Ruby, Scheme, or Java.
After extracting pure Tibetan language microblogging, pretreatment Tibetan language need to be segmented, can use maximum matching in participle and calculate
Method carries out Tibetan language participle with the method combined is manually segmented by the segmenting method of Tibetan language sentiment dictionary, so as to Tibetan language later
Sentiment analysis.
After being segmented to Tibetan language, be re-introduced into word2vec term vector tool to treated Tibetan language sentence carry out word to
Preliminary treatment is measured, to establish the term vector model of the term vector model foundation Tibetan language sentence for the Tibetan language sentence that dimension is 256.
Then, it is 256 by the Tibetan language sentence term vector dimension built, guides to the convolutional neural networks containing three layers
(CNN) it is iterated, input the last layer convolutional layer is connected with full articulamentum, achievees the purpose that feature purifies, last every hiding
Sentence can all export the term vector of one 128 dimension, i.e. the first term vector feature.
The term vector feature of convolutional layer output is introduced again two layers of LSTM layer, every layer contains 128 hidden layers, with one
Full articulamentum is connected, and activation primitive Softmax, over-fitting, the parameter of Dropout are set as 0.5 in order to prevent, be iterated and
Analysis, last every Tibetan language sentence can all export the term vector feature of one 64 dimension, i.e. the second term vector feature.64 are used herein
The low dimensional term vector feature of dimension, can save vector space.
It should be noted that Softmax is normalization exponential function.
The parameter of Dropout is set as 0.5 herein, because having debugged the effect when parameter is 0.3,0.4,0.6,0.8
The effect of emotional semantic classification is good when there is no 0.5, and when the parameter is 0.5, about classifying quality improves 3% or so.
It should be noted that LSTM (Long Short-Term Memory) is shot and long term memory network, it is a kind of time
Recurrent neural network is suitable for being spaced and postpone relatively long critical event in processing and predicted time sequence.
Finally, needing to compile using compile () method after the model as shown in Figure 2 put up according to above-mentioned steps
Translate model, when compilation model must indicate that loss (loss) function and majorized function, loss function are set as binary_
Crossentropy, majorized function are set as Adam (lr=0.01, beta_1=0.9, beta_2=0.999, epsilon=1e-
08), classify further according to Tibetan language sentence of the sofisign classification function to the term vector that output 64 is tieed up, it is preferable to achieve
The sentiment analysis result of classifying quality, wherein the emotions such as passive, positive that sentiment analysis result is distinguished.Herein by multiple
Test show that majorized function is set as Adam (lr=0.01, beta_1=0.9, beta_2=0.999, epsilon=1e-08)
Compared with the setting of other majorized function, this setting classifying quality is more preferable.
It should be noted that compile () method is used to compile regular expression during the execution of the script, in a program
Alternatively referred to as compile function, and loss function and majorized function are then the necessary conditions for executing compile () method, quite
Parameter in compile () method, wherein binary_crossentropy is logarithm loss function；Adam is adaptive
Moments estimation, full name are adaptive moment estimation,.
It is as shown in Figure 3: the Tibetan language sentiment analysis system provided in an embodiment of the present invention based on interacting depth study, comprising:
Tibetan language preprocessing module is pre-processed for obtaining Tibetan language sentence, and to the Tibetan language sentence, is pre-processed
Tibetan language.
Word segmentation module, for carrying out Tibetan language participle to the pretreatment Tibetan language, the Tibetan language after being segmented.
Term vector model building module obtains term vector, and root for carrying out preliminary treatment to the Tibetan language after the participle
The term vector model for the Tibetan language sentence that dimension is 256 is established according to the term vector.
First iteration module is iterated processing for the term vector model to be introduced into three-layer coil product neural network
To extract the first term vector feature, the dimension of the first term vector feature is 128.
Secondary iteration module changes for the first term vector feature to be introduced into two layers of shot and long term memory network
To extract the second term vector feature, the dimension of the second term vector feature is 64 for generation processing.
Compiling and categorization module, for being classified after being compiled the second term vector feature by classification function
To obtain sentiment analysis result.
Optionally, the Tibetan language preprocessing module is specifically used for crawling the Tibetan language sentence in network by web crawlers,
And the expression in the Tibetan language sentence, symbol, Chinese sentence are filtered out, obtain pretreatment Tibetan language.
Optionally, the term vector model building module is specifically used for the Tibetan language after the participle being introduced into word2vec
It is handled in term vector tool, obtains term vector model.
Optionally, the secondary iteration module further include:
Activation primitive setting unit, for setting the activation primitive in two layers of shot and long term memory network to
Softmax。
Parameter set unit, for the parameter of the Dropout in two layers of shot and long term memory network to be set as 0.5.
Optionally, the compiling and categorization module include:
Compilation unit obtains compiling secondary vector for being compiled by compile function to the second term vector feature
Feature, wherein the loss function in the compile () is set as binary_crossentropy, and majorized function is set as Adam
(lr=0.01, beta_1=0.9, beta_2=0.999, epsilon=1e-08).
Taxon classifies to obtain emotion to the compiling secondary vector feature by softsign classification function
Analyze result.
In conclusion the Tibetan language sentiment analysis method and system provided in this embodiment based on interacting depth study, for
Tibetan language sentence sentiment analysis and Tibetan language semanteme lack problem, propose and carry out term vector calculating using Word2vec tool, then
The method for introducing interacting depth study, effectively raises the accuracy rate of Tibetan language sentiment analysis.Interacting deep learning model utilizes
Convolutional layer and LSTM layers of converged network are secondary analyzing Tibetan language feature, not only retaining the global measurement of text and capable of excavating deeper
Semantic relation, and preferable classifying quality can be obtained.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of Tibetan language sentiment analysis method based on interacting depth study, which comprises the steps of:
S1 obtains Tibetan language sentence, and pre-processes to the Tibetan language sentence, obtains pretreatment Tibetan language；
S2 carries out Tibetan language participle to the pretreatment Tibetan language, the Tibetan language after being segmented；
S3 carries out preliminary treatment to the Tibetan language after the participle, obtains term vector, and establish dimension according to the term vector and be
The term vector model of 256 Tibetan language sentence；
The term vector model is introduced into three-layer coil product neural network and is iterated processing to extract the first term vector by S4
Feature, the dimension of the first term vector feature are 128；
The first term vector feature is introduced into two layers of shot and long term memory network and is iterated processing to extract second by S5
Term vector feature, the dimension of the second term vector feature are 64；
S6 is classified after being compiled the second term vector feature by classification function to obtain sentiment analysis result.
2. the Tibetan language sentiment analysis method according to claim 1 based on interacting depth study, which is characterized in that the S1
Specific implementation are as follows:
The Tibetan language sentence in network is crawled by web crawlers, and filters out the expression in the Tibetan language sentence, symbol, Chinese
Sentence obtains pretreatment Tibetan language.
3. the Tibetan language sentiment analysis method according to claim 1 based on interacting depth study, which is characterized in that the S3
Specific implementation are as follows:
Tibetan language after the participle is introduced into word2vec term vector tool and is handled, term vector model is obtained.
4. the Tibetan language sentiment analysis method according to claim 1-3 based on interacting depth study, feature exist
In, the S5 specifically includes the following steps:
The activation primitive in two layers of shot and long term memory network is set Softmax by S51；
The parameter of Dropout in two layers of shot and long term memory network is set as 0.5 by S52.
5. the Tibetan language sentiment analysis method according to claim 1-3 based on interacting depth study, feature exist
In, the S6 specifically includes the following steps:
S61 is compiled the second term vector feature by compile function, obtains compiling secondary vector feature, wherein institute
State the loss function in compile function and be set as binary_crossentropy, majorized function be set as Adam (lr=0.01,
Beta_1=0.9, beta_2=0.999, epsilon=1e-08)；
S62 classifies to obtain sentiment analysis result to the compiling secondary vector feature by softsign classification function.
6. a kind of Tibetan language sentiment analysis system based on interacting depth study characterized by comprising
Tibetan language preprocessing module is pre-processed for obtaining Tibetan language sentence, and to the Tibetan language sentence, obtains pretreatment hiding
Text；
Word segmentation module, for carrying out Tibetan language participle to the pretreatment Tibetan language, the Tibetan language after being segmented；
Term vector model building module obtains term vector, and according to institute for carrying out preliminary treatment to the Tibetan language after the participle
Predicate vector establishes the term vector model for the Tibetan language sentence that dimension is 256；
First iteration module is iterated processing for the term vector model to be introduced into three-layer coil product neural network to mention
The first term vector feature is taken out, the dimension of the first term vector feature is 128；
Secondary iteration module is iterated place for the first term vector feature to be introduced into two layers of shot and long term memory network
For reason to extract the second term vector feature, the dimension of the second term vector feature is 64；
Compiling and categorization module are classified for passing through classification function after being compiled the second term vector feature to obtain
To sentiment analysis result.
7. the Tibetan language sentiment analysis system according to claim 6 based on interacting depth study, which is characterized in that the hiding
Literary preprocessing module is specifically used for: crawling the Tibetan language sentence in network by web crawlers, and filters out the Tibetan language sentence
In expression, symbol, Chinese sentence, obtain pretreatment Tibetan language.
8. the Tibetan language sentiment analysis system according to claim 6 based on interacting depth study, which is characterized in that institute's predicate
Vector model is established module and is specifically used for: the Tibetan language after the participle is introduced into word2vec term vector tool
Reason, obtains term vector model.
9. according to the described in any item Tibetan language sentiment analysis systems based on interacting depth study of claim 6-8, feature exists
In the secondary iteration module further include:
Activation primitive setting unit, for setting Softmax for the activation primitive in two layers of shot and long term memory network；
Parameter set unit, for the parameter of the Dropout in two layers of shot and long term memory network to be set as 0.5.
10. according to the described in any item Tibetan language sentiment analysis systems based on interacting depth study of claim 6-8, feature exists
In the compiling and categorization module include:
It is special to obtain compiling secondary vector for being compiled by compile function to the second term vector feature for compilation unit
Sign, wherein the loss function in the compile function is set as binary_crossentropy, and majorized function is set as Adam
(lr=0.01, beta_1=0.9, beta_2=0.999, epsilon=1e-08)；
Taxon classifies to obtain sentiment analysis to the compiling secondary vector feature by softsign classification function
As a result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811045667.8A CN109408805A (en) | 2018-09-07 | 2018-09-07 | A kind of Tibetan language sentiment analysis method and system based on interacting depth study |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811045667.8A CN109408805A (en) | 2018-09-07 | 2018-09-07 | A kind of Tibetan language sentiment analysis method and system based on interacting depth study |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109408805A true CN109408805A (en) | 2019-03-01 |
Family
ID=65464630
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811045667.8A Pending CN109408805A (en) | 2018-09-07 | 2018-09-07 | A kind of Tibetan language sentiment analysis method and system based on interacting depth study |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109408805A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110083716A (en) * | 2019-05-07 | 2019-08-02 | 青海大学 | Multi-modal affection computation method and system based on Tibetan language |
| CN110321563A (en) * | 2019-06-28 | 2019-10-11 | 浙江大学 | Text emotion analysis method based on mixing monitor model |
| CN111008186A (en) * | 2019-06-11 | 2020-04-14 | 中央民族大学 | Expression method of Tibetan knowledge base |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105468744A (en) * | 2015-11-25 | 2016-04-06 | 浪潮软件集团有限公司 | Big data platform for realizing tax public opinion analysis and full text retrieval |
| CN108415972A (en) * | 2018-02-08 | 2018-08-17 | 合肥工业大学 | text emotion processing method |
| CN108427670A (en) * | 2018-04-08 | 2018-08-21 | 重庆邮电大学 | A kind of sentiment analysis method based on context word vector sum deep learning |
-
2018
- 2018-09-07 CN CN201811045667.8A patent/CN109408805A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105468744A (en) * | 2015-11-25 | 2016-04-06 | 浪潮软件集团有限公司 | Big data platform for realizing tax public opinion analysis and full text retrieval |
| CN108415972A (en) * | 2018-02-08 | 2018-08-17 | 合肥工业大学 | text emotion processing method |
| CN108427670A (en) * | 2018-04-08 | 2018-08-21 | 重庆邮电大学 | A kind of sentiment analysis method based on context word vector sum deep learning |
Non-Patent Citations (4)
| Title |
|---|
| BENWANG SUN 等: "Tibetan Micro-Blog Sentiment Analysis Based on Mixed Deep Learning", 《 2018 INTERNATIONAL CONFERENCE ON AUDIO, LANGUAGE AND IMAGE PROCESSING》 * |
| QIANZI SHEN 等: "Sentiment Analysis of Movie Reviews Based on CNN-BLSTM", 《INTERNATIONAL CONFERENCE ON INTELLIGENCE SCIENCE 2017》 * |
| 刘伟 等: "基于注意模型深度学习的文本情感倾向性研究", 《第19届中国系统仿真技术及其应用学术年会论文集》 * |
| 孙本旺 等: "基于深度学习算法的藏文微博情感计算研究", 《计算机技术与发展》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110083716A (en) * | 2019-05-07 | 2019-08-02 | 青海大学 | Multi-modal affection computation method and system based on Tibetan language |
| CN111008186A (en) * | 2019-06-11 | 2020-04-14 | 中央民族大学 | Expression method of Tibetan knowledge base |
| CN110321563A (en) * | 2019-06-28 | 2019-10-11 | 浙江大学 | Text emotion analysis method based on mixing monitor model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112487203B (en) | Relation extraction system integrated with dynamic word vector | |
| Mou et al. | Natural language inference by tree-based convolution and heuristic matching | |
| CN109255119A (en) | A kind of sentence trunk analysis method and system based on the multitask deep neural network for segmenting and naming Entity recognition | |
| CN111143576A (en) | Event-oriented dynamic knowledge graph construction method and device | |
| CN110532554A (en) | A kind of Chinese abstraction generating method, system and storage medium | |
| CN110502753A (en) | A kind of deep learning sentiment analysis model and its analysis method based on semantically enhancement | |
| CN109408805A (en) | A kind of Tibetan language sentiment analysis method and system based on interacting depth study | |
| CN102194001A (en) | Internet public opinion crisis early-warning method | |
| CN112307473A (en) | Malicious JavaScript code detection model based on Bi-LSTM network and attention mechanism | |
| CN111386524A (en) | Facilitating domain and client specific application program interface recommendations | |
| CN101751385B (en) | Multilingual information extraction method adopting hierarchical pipeline filter system structure | |
| CN105912524B (en) | The article topic keyword extracting method and device decomposed based on low-rank matrix | |
| CN111310476A (en) | Public opinion monitoring method and system using aspect-based emotion analysis method | |
| CN110502634A (en) | A kind of case by judgement and grasping means and its system | |
| CN111966827A (en) | Conversation emotion analysis method based on heterogeneous bipartite graph | |
| CN111090811A (en) | Method and system for extracting massive news hot topics | |
| CN113742396A (en) | Mining method and device for object learning behavior pattern | |
| CN106021225B (en) | A kind of Chinese Maximal noun phrase recognition methods based on the simple noun phrase of Chinese | |
| CN114764463A (en) | Internet public opinion event automatic early warning system based on event propagation characteristics | |
| Kim et al. | Predicting emotion in movie scripts using deep learning | |
| Shaw et al. | Structural processing of waveforms as trees | |
| CN112131879A (en) | Relationship extraction system, method and device | |
| Farhan et al. | Ensemble of gated recurrent unit and convolutional neural network for sarcasm detection in bangla | |
| Ma et al. | Aspect-Based Attention LSTM for Aspect-Level Sentiment Analysis | |
| Gao et al. | A hybrid GCN and RNN structure based on attention mechanism for text classification |
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: 20190301 |
|
| RJ01 | Rejection of invention patent application after publication |