CN110222184A - A kind of emotion information recognition methods of text and relevant apparatus - Google Patents
A kind of emotion information recognition methods of text and relevant apparatus Download PDFInfo
- Publication number
- CN110222184A CN110222184A CN201910511441.0A CN201910511441A CN110222184A CN 110222184 A CN110222184 A CN 110222184A CN 201910511441 A CN201910511441 A CN 201910511441A CN 110222184 A CN110222184 A CN 110222184A
- Authority
- CN
- China
- Prior art keywords
- text
- obtains
- feature
- trained
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Machine Translation (AREA)
Abstract
This application discloses a kind of emotion information recognition methods of text, comprising: is pre-processed using the Word2vec model trained to text to be predicted, obtains the input vector of text to be predicted;Feature identification is carried out to input vector using the BiLSTM model trained, obtains context timing text information feature;Feature extraction is carried out to context timing text information feature by the CNN model trained, obtains target signature;Target signature is identified using the neural network trained, obtains emotion recognition result.The temporal aspect for going out the context of text to be predicted by the Word2vec model, BiLSTM model and CNN model extraction trained, gets the local feature and sequence information of text, improves the precision and accuracy rate of sentiment analysis.Disclosed herein as well is a kind of emotion information identification equipment, natural language analysis device and computer readable storage mediums, have the above beneficial effect.
Description
Technical field
This application involves natural language analysis technical field, in particular to the emotion information recognition methods of a kind of text, feelings
Feel information identification equipment, natural language analysis device and computer readable storage medium.
Background technique
Sentiment analysis is another analytic angle of natural language processing field, also known as proneness analysis, opinion extraction, opinion
Excavation, emotion excavation, subjective analysis etc., mainly to emotional color subjective texts analyzed, handle, conclude and
The process of reasoning, analyzes evaluation of the user to film such as from film comment, and user is analyzed from comment on commodity text to commodity
The attributes such as " price, size, weight, ease for use " Sentiment orientation.
And be exactly to analyze the text with emotion for the main task of the sentiment analysis of comment on commodity, locate
Reason is concluded and is judged.Conventional machines learning algorithm needs the data characteristics using a large amount of artificial selection, expends a large amount of manpowers and moves
Shifting property is not strong, and whether has supervision or unsupervised learning method, belongs to shallow-layer study, no calligraphy learning to text
Deeper information, in the case where limited text data and computing unit, machine learning for challenge processing and
It will receive a degree of limitation in the realization of sophisticated functions.
Therefore, the prior art is to overcome the defect of conventional machines learning algorithm, handled using deep learning algorithm from
Right language processing tasks, wherein CNN (Convolutional Neural Network convolutional neural networks) and RNN
(Recurrent Neural Networks recurrent neural network) is most popular network mould in text emotion analysis task
Type.But since word each in text or sentence have different decisive actions to the feeling polarities of entire text, and more than
In two kinds of neural networks, the former can ignore the context semanteme of word and a large amount of characteristic informations can be lost in the operation of maximum pondization
It loses, and there is gradient disappearance and gradient disperse in the latter.Specifically, CNN is not fully suitable for learning time sequence,
Therefore various complementary processing be may require that, and effect is also different sets.And RNN only considered unidirectional sequence problem, not fill
Divide and combine context of co-text, and will appear gradient disappearance and gradient explosion issues.Both the above situation can all cause pair
The accuracy of the sentiment analysis of text declines, and reduces the precision of sentiment analysis.
Therefore, the accuracy and precision for how improving the sentiment analysis in text are the emphasis of those skilled in the art's concern
Problem.
Summary of the invention
The purpose of the application is to provide a kind of emotion information recognition methods of text, emotion information identification equipment, natural language
It says analytical equipment and computer readable storage medium, passes through the Word2vec model, BiLSTM model and CNN mould trained
Type extracts the temporal aspect of the context of text to be predicted, fully gets the local feature and sequence information of text, mentions
The precision and accuracy rate of high touch analysis.
In order to solve the above technical problems, the application provides a kind of emotion information recognition methods of text, comprising:
Text to be predicted is pre-processed using the Word2vec model trained, obtains the defeated of the text to be predicted
Incoming vector;
Feature identification is carried out to the input vector using the BiLSTM model trained, obtains context timing text envelope
Cease feature;
Feature extraction is carried out to the context timing text information feature by the CNN model trained, obtains target
Feature;
The target signature is identified using the neural network trained, obtains emotion recognition result.
Optionally, the Word2vec model that use has been trained pre-processes text to be predicted, obtains described to be predicted
The input vector of text, comprising:
Stop words processing is carried out to the text to be predicted, obtains identification corpus;
The identification corpus is handled using the Word2vec model trained, obtains the input vector.
Optionally, feature extraction is carried out to the context timing text information feature by the CNN model trained, obtained
To target signature, comprising:
Feature is carried out respectively to each context timing text information feature by the CNN model trained to mention
It takes, obtains multiple sub-goal features;
All sub-goal features are connected, the target signature is obtained.
Optionally, the neural network that use has been trained identifies the target signature, obtains emotion recognition as a result, packet
It includes:
The target signature is input to the neural network trained, obtains processing result;
The processing result is input to sigmoid function category layer to classify, obtains the emotion recognition result.
The application also provides a kind of emotion information identification equipment of text, comprising:
Vector matrix obtains module, for being pre-processed using the Word2vec model trained to text to be predicted,
Obtain the input vector of the text to be predicted;
Sequence relation obtains module, for carrying out feature knowledge to the input vector using the BiLSTM model trained
Not, context timing text information feature is obtained;
Characteristic extracting module carries out the context timing text information feature for the CNN model by having trained
Feature extraction obtains target signature;
Emotion recognition module obtains emotion for identifying using the neural network trained to the target signature
Recognition result.
Optionally, the vector matrix obtains module, comprising:
Stop words processing unit obtains identification corpus for carrying out stop words processing to the text to be predicted;
Corpus recognition unit, for being handled using the Word2vec model trained the identification corpus,
Obtain the input vector.
Optionally, the characteristic extracting module, comprising:
Sub-goal feature acquiring unit, preface when for by the CNN model trained to each context
This information characteristics carries out feature extraction respectively, obtains multiple sub-goal features;
Target signature acquiring unit obtains the target signature for connecting all sub-goal features.
Optionally, the emotion recognition module, comprising:
Target signature recognition unit obtains everywhere for the target signature to be input to the neural network trained
Manage result;
Emotional semantic classification unit classifies for the processing result to be input to sigmoid function category layer, obtains institute
State emotion recognition result.
The application also provides a kind of natural language analysis device, comprising:
Memory, for storing computer program;
Processor, the step of emotion information recognition methods as described above is realized when for executing the computer program.
The application also provides a kind of computer readable storage medium, and calculating is stored on the computer readable storage medium
The step of machine program, the computer program realizes emotion information recognition methods as described above when being executed by processor.
The emotion information recognition methods of a kind of text provided herein, comprising: using the Word2vec mould trained
Type pre-processes text to be predicted, obtains the input vector of the text to be predicted;Using the BiLSTM model trained
Feature identification is carried out to the input vector, obtains context timing text information feature;By the CNN model trained to institute
It states context timing text information feature and carries out feature extraction, obtain target signature;Using the neural network trained to described
Target signature is identified, emotion recognition result is obtained.
Text to be predicted is handled by the Word2vec model, BiLSTM model and CNN model trained, by
It is the equal of that feature extraction is carried out to text to be predicted after which is linked up use in the characteristic of each model,
The characteristic of three models is respectively adopted in the process further extracted, and extracts the relationship being related in text between context,
And context timing text information feature, and then context timing text information feature is done further using CNN model again
Extraction operation, extraction obtain target signature, the temporal aspect of context are demonstrated by this feature, can fully extract text
Local feature and sequence information, and then can identify the important information feature of text, it can during carrying out sentiment analysis
To improve the precision and accuracy rate of sentiment analysis.
The application also provides the emotion information identification equipment of text a kind of, natural language analysis device and computer-readable
Storage medium has the above beneficial effect, and therefore not to repeat here.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of application for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is a kind of flow chart of the emotion information recognition methods of text provided by the embodiment of the present application;
Fig. 2 is a kind of structural schematic diagram of the emotion information identification equipment of text provided by the embodiment of the present application.
Specific embodiment
The core of the application is to provide a kind of emotion information recognition methods of text, emotion information identification equipment, natural language
It says analytical equipment and computer readable storage medium, passes through the Word2vec model, BiLSTM model and CNN mould trained
Type extracts the temporal aspect of the context of text to be predicted, fully gets the local feature and sequence information of text, mentions
The precision and accuracy rate of high touch analysis.
To keep the purposes, technical schemes and advantages of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application
In attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is
Some embodiments of the present application, instead of all the embodiments.Based on the embodiment in the application, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall in the protection scope of this application.
The prior art handles natural language using deep learning algorithm to overcome the defect of conventional machines learning algorithm
Processing task, wherein CNN (Convolutional Neural Network convolutional neural networks) and RNN (Recurrent
Neural Networks recurrent neural network) it is most popular network model in text emotion analysis task.
But since word each in text or sentence have different decisive actions to the feeling polarities of entire text, and
In both the above neural network, the former can ignore the context semanteme of word and a large amount of characteristic information meetings in the operation of maximum pondization
It loses, and there is gradient disappearance and gradient disperse in the latter.Specifically, CNN is not fully suitable for learning time sequence
Column, therefore may require that various complementary processing, and effect is also different sets.And RNN only considered unidirectional sequence problem, not have
Context of co-text is sufficiently combined, and will appear gradient disappearance and gradient explosion issues.Both the above situation can all cause
Accuracy decline to the sentiment analysis of text, reduces the precision of sentiment analysis.
Therefore, the application provides a kind of emotion information recognition methods of text, by the Word2vec model trained,
BiLSTM model and CNN model handle text to be predicted, and due to the characteristic of each model, which is connected
Get up after use, is the equal of that feature extraction is carried out to text to be predicted, three moulds are respectively adopted in the process further extracted
The characteristic of type extracts the relationship being related in text between context and context timing text information feature, and then adopts again
Further extraction operation is done to context timing text information feature with CNN model, extracts and obtains target signature, in this feature
It is demonstrated by the temporal aspect of context, can fully extract the local feature and sequence information of text, and then can be identified
The precision and accuracy rate of sentiment analysis can be improved in the important information feature of text during carrying out sentiment analysis.
Referring to FIG. 1, Fig. 1 is a kind of process of the emotion information recognition methods of text provided by the embodiment of the present application
Figure.
In the present embodiment, this method may include:
S101, the Word2vec model that use has been trained pre-process text to be predicted, obtain text to be predicted
Input vector;
This step is intended to first pre-process text to be predicted using the Word2vec model trained, and obtains to be predicted
The input vector of text.Wherein, Word2vec model is a series of for generating the correlation model of term vector.These models are shallow
And double-deck neural network, it is used to training with the word text of construction linguistics again.The sequence of word is not heavy in Word2vec
It wants.After training is completed, Word2vec model can be used to map each word to a vector, can be used to indicate word between word
Relationship.
During natural language processing, when identifying the substantially meaning of some word, for word word
For its contextual information be it is vital yes.Therefore, using Word2vec model, some word can be checked
Word contextual information neighbouring in a sentence carries out building word word vectors, for concrete details without place
Reason, the word word with the context in the similar sentence of degree is mapped in adjacent vector space.Word2vec mould
Type handle obtained embeded matrix can cover the word of each different classes of word word in training corpus to
Amount, and the word vector that embeded matrix can include can be more than 3,000,000.In the available training corpus of Word2Vec model
Each sentence, then slided by the window that a size is fixed as N, and predict the word window that those are provided
Centre word carries out the training of model.Whole process be reached for using loss function and optimization process corpus it is all only
One word word generates word term vector.
The acquisition methods for any one corpus that can be provided using the prior art in this step, do not do specific limit herein
It is fixed.Corpus is obtained by the acquisition methods of corpus, the unique word of each of vocabulary of corpus has uniquely
Its corresponding term vector is indexed, term vector matrix is finally obtained.For non-dictionary word, we must carry out generation using a special number
For such as 0.A string of words are in short obtained after word segmentation processing, such as " my dog is very cute and smart " passes through
Word segmentation processing becomes " my " later, " dog ", " is ", " very ", " cute ", " and ", " smart ", each word can have only
One index is corresponding, then obtains vector [20 196 122 160 422 20,216 41].It is then converted to uniform length
Form, if length is 10, the words contains 7 words, mends 0 in remaining 3 vacancy, becomes [20 196 122 160 422
20216 41 00 0], final term vector [batch_size=1, max_len=is then obtained by term vector matrix again
10, Word2vec_dimension=50], i.e. the input vector of word.
Optionally, this step may include:
Stop words processing is carried out to text to be predicted, obtains identification corpus;
Identification corpus is handled using the Word2vec model trained, obtains input vector.
As it can be seen that being to carry out stop words processing to text to be predicted first in this step, identification corpus is obtained, is then used again
Word2vec model handles identification corpus, obtains input vector.Wherein, stop words processing refers to removal text to be predicted
In stop words, avoid stop words nonsensical in text from interfering the prediction process of text, improve the accurate of prediction
Rate and precision.
S102, the BiLSTM model that use has been trained carry out feature identification to input vector, obtain context timing text
Information characteristics;
S103 carries out feature extraction to context timing text information feature by the CNN model trained, obtains target
Feature;
On the basis of S101, step S102 and S103 are intended to carry out further feature to the input vector got to mention
It takes, that is, input vector is handled using BiLSTM-CNN model.Specifically, the BiLSTM-CNN model in this step
Using multiple term vectors after reception Word2vec model treatment as input.Then it outputs it and is pooled to a lesser ruler
It is very little, it is then input to CNN layers.Bi-LSTM layers will enable the sequence with feature to understand the text sequence of input.By two-way
Shot and long term memory network obtains the output of each word, obtains output vector, and convolutional neural networks are exactly to each word
Carry out feature extraction finally export extraction feature, obtain the target signature, connect be input in neural network finally by
Sigmoid function is exported.
Optionally, step S103 may include:
By the CNN model trained feature extraction is carried out to each context timing text information feature respectively, obtained
Multiple sub-goal features;All sub-goal features are connected, target signature is obtained.
It is principally obtaining the corresponding sub-goal feature of each word in this optinal plan after CNN model treatment, and neural network
It is then to handle one section of continuous sentence.Therefore, it connects all sub-goal features to obtain target signature in this step.
S104, the neural network that use has been trained identify target signature, obtain emotion recognition result.
On the basis of S103, this step is intended to identify target signature using the neural network trained, and obtains
Emotion recognition result.
Specifically, this step may include:
Target signature is input to the neural network trained, obtains processing result;
Processing result is input to sigmoid function category layer to classify, obtains emotion recognition result.
Only to carry out passive and positive classification to emotion, classified by sigmoid function category layer, obtained knot
Only there are two types of as a result, 1 or 0 for fruit.Wherein, it 1 indicates actively, 0 mark is passive.
To sum up, the present embodiment passes through Word2vec model, BiLSTM model and the CNN model trained to text to be predicted
This is handled, and is the equal of to text to be predicted after which is linked up use due to the characteristic of each model
Feature extraction is carried out, the characteristic of three models is respectively adopted in the process further extracted, and extracts and is related in text up and down
Relationship and context timing text information feature between text, so it is special to context timing text information using CNN model again
Sign does further extraction operation, and extraction obtains target signature, the temporal aspect of context is demonstrated by this feature, can be abundant
Ground extracts the local feature and sequence information of text, and then can identify the important information feature of text, is carrying out emotion point
The precision and accuracy rate of sentiment analysis can be improved during analysis.
A kind of emotion information identification equipment of text provided by the embodiments of the present application is introduced below, it is described below
A kind of emotion information identification equipment of text can correspond to each other ginseng with a kind of above-described emotion information recognition methods of text
According to.
Referring to FIG. 2, Fig. 2 identifies that the structure of equipment is shown for a kind of emotion information of text provided by the embodiment of the present application
It is intended to.
In the present embodiment, which may include:
Vector matrix obtains module 100, for being located in advance using the Word2vec model trained to text to be predicted
Reason, obtains the input vector of text to be predicted;
Sequence relation obtains module 200, for carrying out feature identification to input vector using the BiLSTM model trained,
Obtain context timing text information feature;
Characteristic extracting module 300 carries out context timing text information feature for the CNN model by having trained special
Sign is extracted, and target signature is obtained;
Emotion recognition module 400 obtains emotion knowledge for identifying using the neural network trained to target signature
Other result.
Optionally, vector matrix acquisition module 100 may include:
Stop words processing unit obtains identification corpus for carrying out stop words processing to text to be predicted;
Corpus recognition unit is inputted for being handled using the Word2vec model trained identification corpus
Vector.
Optionally, this feature extraction module 300 may include:
Sub-goal feature acquiring unit, it is special to each context timing text information for the CNN model by having trained
Sign carries out feature extraction respectively, obtains multiple sub-goal features;
Target signature acquiring unit obtains target signature for connecting all sub-goal features.
Optionally, which may include:
Target signature recognition unit obtains processing result for target signature to be input to the neural network trained;
Emotional semantic classification unit is classified for processing result to be input to sigmoid function category layer, obtains emotion knowledge
Other result.
The embodiment of the present application also provides a kind of natural language analysis device, comprising:
Memory, for storing computer program;
Processor, the step of emotion information recognition methods as above is realized when for executing computer program.
The embodiment of the present application also provides a kind of computer readable storage medium, and meter is stored on computer readable storage medium
Calculation machine program, the step of emotion information recognition methods as above is realized when computer program is executed by processor.
The computer readable storage medium may include: USB flash disk, mobile hard disk, read-only memory (Read-Only
Memory, ROM), random access memory (Random Access Memory, RAM), magnetic or disk etc. is various to deposit
Store up the medium of program code.
Each embodiment is described in a progressive manner in specification, the highlights of each of the examples are with other realities
The difference of example is applied, the same or similar parts in each embodiment may refer to each other.For device disclosed in embodiment
Speech, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is referring to method part illustration
?.
Professional further appreciates that, unit described in conjunction with the examples disclosed in the embodiments of the present disclosure
And algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and
The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These
Function is implemented in hardware or software actually, the specific application and design constraint depending on technical solution.Profession
Technical staff can use different methods to achieve the described function each specific application, but this realization is not answered
Think beyond scope of the present application.
The step of method described in conjunction with the examples disclosed in this document or algorithm, can directly be held with hardware, processor
The combination of capable software module or the two is implemented.Software module can be placed in random access memory (RAM), memory, read-only deposit
Reservoir (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology
In any other form of storage medium well known in field.
Equipment, nature are identified to a kind of emotion information recognition methods of text provided herein, emotion information above
Language analysis device and computer readable storage medium are described in detail.Specific case used herein is to the application
Principle and embodiment be expounded, the present processes that the above embodiments are only used to help understand and its core
Thought is thought.It should be pointed out that for those skilled in the art, under the premise of not departing from the application principle,
Can also to the application, some improvement and modification can also be carried out, these improvement and modification also fall into the protection scope of the claim of this application
It is interior.
Claims (10)
1. a kind of emotion information recognition methods of text characterized by comprising
Text to be predicted is pre-processed using the Word2vec model trained, obtain the input of the text to be predicted to
Amount;
Feature identification is carried out to the input vector using the BiLSTM model trained, it is special to obtain context timing text information
Sign;
Feature extraction is carried out to the context timing text information feature by the CNN model trained, obtains target signature;
The target signature is identified using the neural network trained, obtains emotion recognition result.
2. emotion information recognition methods according to claim 1, which is characterized in that using the Word2vec model trained
Text to be predicted is pre-processed, the input vector of the text to be predicted is obtained, comprising:
Stop words processing is carried out to the text to be predicted, obtains identification corpus;
The identification corpus is handled using the Word2vec model trained, obtains the input vector.
3. emotion information recognition methods according to claim 1, which is characterized in that by the CNN model trained to institute
It states context timing text information feature and carries out feature extraction, obtain target signature, comprising:
Feature extraction is carried out respectively to each context timing text information feature by the CNN model trained,
Obtain multiple sub-goal features;
All sub-goal features are connected, the target signature is obtained.
4. emotion information recognition methods according to claim 1, which is characterized in that using the neural network trained to institute
It states target signature to be identified, obtains emotion recognition result, comprising:
The target signature is input to the neural network trained, obtains processing result;
The processing result is input to sigmoid function category layer to classify, obtains the emotion recognition result.
5. a kind of emotion information of text identifies equipment characterized by comprising
Vector matrix obtains module, for being pre-processed using the Word2vec model trained to text to be predicted, obtains
The input vector of the text to be predicted;
Sequence relation obtains module, for carrying out feature identification to the input vector using the BiLSTM model trained, obtains
To context timing text information feature;
Characteristic extracting module carries out feature to the context timing text information feature for the CNN model by having trained
It extracts, obtains target signature;
Emotion recognition module obtains emotion recognition for identifying using the neural network trained to the target signature
As a result.
6. emotion information according to claim 5 identifies equipment, which is characterized in that the vector matrix obtains module, packet
It includes:
Stop words processing unit obtains identification corpus for carrying out stop words processing to the text to be predicted;
Corpus recognition unit is obtained for being handled using the Word2vec model trained the identification corpus
The input vector.
7. emotion information according to claim 5 identifies equipment, which is characterized in that the characteristic extracting module, comprising:
Sub-goal feature acquiring unit, the CNN model for having trained described in is to each context timing text envelope
Breath feature carries out feature extraction respectively, obtains multiple sub-goal features;
Target signature acquiring unit obtains the target signature for connecting all sub-goal features.
8. emotion information according to claim 5 identifies equipment, which is characterized in that the emotion recognition module, comprising:
Target signature recognition unit obtains processing knot for the target signature to be input to the neural network trained
Fruit;
Emotional semantic classification unit classifies for the processing result to be input to sigmoid function category layer, obtains the feelings
Feel recognition result.
9. a kind of natural language analysis device characterized by comprising
Memory, for storing computer program;
Processor is realized when for executing the computer program such as the described in any item emotion information identifications of Claims 1-4
The step of method.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
Program realizes such as Claims 1-4 described in any item emotion information identification sides when the computer program is executed by processor
The step of method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910511441.0A CN110222184A (en) | 2019-06-13 | 2019-06-13 | A kind of emotion information recognition methods of text and relevant apparatus |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910511441.0A CN110222184A (en) | 2019-06-13 | 2019-06-13 | A kind of emotion information recognition methods of text and relevant apparatus |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110222184A true CN110222184A (en) | 2019-09-10 |
Family
ID=67816878
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910511441.0A Pending CN110222184A (en) | 2019-06-13 | 2019-06-13 | A kind of emotion information recognition methods of text and relevant apparatus |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110222184A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111191459A (en) * | 2019-12-25 | 2020-05-22 | 医渡云(北京)技术有限公司 | Text processing method and device, readable medium and electronic equipment |
CN111414475A (en) * | 2020-03-03 | 2020-07-14 | 北京明略软件系统有限公司 | Text emotion information identification method and device |
CN111428118A (en) * | 2019-11-08 | 2020-07-17 | 华东理工大学 | Method for detecting event reliability and electronic equipment |
CN111507421A (en) * | 2020-04-22 | 2020-08-07 | 上海极链网络科技有限公司 | Video-based emotion recognition method and device |
CN111723572A (en) * | 2020-06-12 | 2020-09-29 | 广西师范大学 | Chinese short text correlation measurement method based on CNN convolutional layer and BilSTM |
CN112233698A (en) * | 2020-10-09 | 2021-01-15 | 中国平安人寿保险股份有限公司 | Character emotion recognition method and device, terminal device and storage medium |
CN112270168A (en) * | 2020-10-14 | 2021-01-26 | 北京百度网讯科技有限公司 | Dialogue emotion style prediction method and device, electronic equipment and storage medium |
CN112528657A (en) * | 2020-12-23 | 2021-03-19 | 中移(杭州)信息技术有限公司 | Text intention recognition method and device based on bidirectional LSTM, server and medium |
CN112612878A (en) * | 2020-12-17 | 2021-04-06 | 大唐融合通信股份有限公司 | Customer service information providing method, electronic equipment and device |
CN116108859A (en) * | 2023-03-17 | 2023-05-12 | 美云智数科技有限公司 | Emotional tendency determination, sample construction and model training methods, devices and equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160099010A1 (en) * | 2014-10-03 | 2016-04-07 | Google Inc. | Convolutional, long short-term memory, fully connected deep neural networks |
CN106599933A (en) * | 2016-12-26 | 2017-04-26 | 哈尔滨工业大学 | Text emotion classification method based on the joint deep learning model |
CN108647219A (en) * | 2018-03-15 | 2018-10-12 | 中山大学 | A kind of convolutional neural networks text emotion analysis method of combination sentiment dictionary |
CN108829818A (en) * | 2018-06-12 | 2018-11-16 | 中国科学院计算技术研究所 | A kind of file classification method |
-
2019
- 2019-06-13 CN CN201910511441.0A patent/CN110222184A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160099010A1 (en) * | 2014-10-03 | 2016-04-07 | Google Inc. | Convolutional, long short-term memory, fully connected deep neural networks |
CN106599933A (en) * | 2016-12-26 | 2017-04-26 | 哈尔滨工业大学 | Text emotion classification method based on the joint deep learning model |
CN108647219A (en) * | 2018-03-15 | 2018-10-12 | 中山大学 | A kind of convolutional neural networks text emotion analysis method of combination sentiment dictionary |
CN108829818A (en) * | 2018-06-12 | 2018-11-16 | 中国科学院计算技术研究所 | A kind of file classification method |
Non-Patent Citations (1)
Title |
---|
李洋等: "《基于CNN 和BiLSTM 网络特征融合的文本情感分析》", 《计算机应用》 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111428118A (en) * | 2019-11-08 | 2020-07-17 | 华东理工大学 | Method for detecting event reliability and electronic equipment |
CN111428118B (en) * | 2019-11-08 | 2023-04-11 | 华东理工大学 | Method for detecting event reliability and electronic equipment |
CN111191459A (en) * | 2019-12-25 | 2020-05-22 | 医渡云(北京)技术有限公司 | Text processing method and device, readable medium and electronic equipment |
CN111191459B (en) * | 2019-12-25 | 2023-12-12 | 医渡云(北京)技术有限公司 | Text processing method and device, readable medium and electronic equipment |
CN111414475A (en) * | 2020-03-03 | 2020-07-14 | 北京明略软件系统有限公司 | Text emotion information identification method and device |
CN111507421A (en) * | 2020-04-22 | 2020-08-07 | 上海极链网络科技有限公司 | Video-based emotion recognition method and device |
CN111723572B (en) * | 2020-06-12 | 2021-11-19 | 广西师范大学 | Chinese short text correlation measurement method based on CNN convolutional layer and BilSTM |
CN111723572A (en) * | 2020-06-12 | 2020-09-29 | 广西师范大学 | Chinese short text correlation measurement method based on CNN convolutional layer and BilSTM |
CN112233698A (en) * | 2020-10-09 | 2021-01-15 | 中国平安人寿保险股份有限公司 | Character emotion recognition method and device, terminal device and storage medium |
CN112233698B (en) * | 2020-10-09 | 2023-07-25 | 中国平安人寿保险股份有限公司 | Character emotion recognition method, device, terminal equipment and storage medium |
CN112270168A (en) * | 2020-10-14 | 2021-01-26 | 北京百度网讯科技有限公司 | Dialogue emotion style prediction method and device, electronic equipment and storage medium |
CN112270168B (en) * | 2020-10-14 | 2023-11-24 | 北京百度网讯科技有限公司 | Method and device for predicting emotion style of dialogue, electronic equipment and storage medium |
CN112612878A (en) * | 2020-12-17 | 2021-04-06 | 大唐融合通信股份有限公司 | Customer service information providing method, electronic equipment and device |
CN112528657A (en) * | 2020-12-23 | 2021-03-19 | 中移(杭州)信息技术有限公司 | Text intention recognition method and device based on bidirectional LSTM, server and medium |
CN116108859A (en) * | 2023-03-17 | 2023-05-12 | 美云智数科技有限公司 | Emotional tendency determination, sample construction and model training methods, devices and equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110222184A (en) | A kind of emotion information recognition methods of text and relevant apparatus | |
Salur et al. | A novel hybrid deep learning model for sentiment classification | |
CN110134946B (en) | Machine reading understanding method for complex data | |
CN106156003B (en) | A kind of question sentence understanding method in question answering system | |
CN111709242B (en) | Chinese punctuation mark adding method based on named entity recognition | |
CN113095415B (en) | Cross-modal hashing method and system based on multi-modal attention mechanism | |
CN110750635B (en) | French recommendation method based on joint deep learning model | |
CN105938485A (en) | Image description method based on convolution cyclic hybrid model | |
CN108549658A (en) | A kind of deep learning video answering method and system based on the upper attention mechanism of syntactic analysis tree | |
CN109918681B (en) | Chinese character-pinyin-based fusion problem semantic matching method | |
CN109117470B (en) | Evaluation relation extraction method and device for evaluating text information | |
CN107357785A (en) | Theme feature word abstracting method and system, feeling polarities determination methods and system | |
CN111368544A (en) | Named entity identification method and device | |
CN110188926A (en) | A kind of order information forecasting system and method | |
CN115545041B (en) | Model construction method and system for enhancing semantic vector representation of medical statement | |
CN117236338B (en) | Named entity recognition model of dense entity text and training method thereof | |
CN111145914B (en) | Method and device for determining text entity of lung cancer clinical disease seed bank | |
CN112507124A (en) | Chapter-level event causal relationship extraction method based on graph model | |
CN113486174A (en) | Model training, reading understanding method and device, electronic equipment and storage medium | |
CN110909174B (en) | Knowledge graph-based method for improving entity link in simple question answering | |
CN109117471A (en) | A kind of calculation method and terminal of the word degree of correlation | |
CN110362734A (en) | Text recognition method, device, equipment and computer readable storage medium | |
CN110287396A (en) | Text matching technique and device | |
CN111813927A (en) | Sentence similarity calculation method based on topic model and LSTM | |
CN115130475A (en) | Extensible universal end-to-end named entity identification method |
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: 20190910 |
|
RJ01 | Rejection of invention patent application after publication |