CN103631961A - Method for identifying relationship between sentiment words and evaluation objects - Google Patents
Method for identifying relationship between sentiment words and evaluation objects Download PDFInfo
- Publication number
- CN103631961A CN103631961A CN201310693087.0A CN201310693087A CN103631961A CN 103631961 A CN103631961 A CN 103631961A CN 201310693087 A CN201310693087 A CN 201310693087A CN 103631961 A CN103631961 A CN 103631961A
- Authority
- CN
- China
- Prior art keywords
- evaluation object
- emotion word
- feature
- word
- language material
- 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.)
- Granted
Links
Images
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
- 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
- G06F16/31—Indexing; Data structures therefor; Storage structures
- G06F16/313—Selection or weighting of terms for indexing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Machine Translation (AREA)
Abstract
The invention relates to a method for identifying the relationship between sentiment words and evaluation objects. The method comprises the main steps of making linguistic data, training a condition random field model, extracting the sentiment words and the evaluation objects, forming candidate sets of the sentiment words and the evaluation objects, training a maximum entropy classifier, testing the maximum entropy classifier and carrying out practical application. According to the method, the relationship between the sentiment words and the evaluation objects is fully considered, the maximum entropy classifier is adopted, and whether the candidate sets of the sentiment words and the evaluation objects which are extracted by the condition random field model have corresponding relationship or not can be indentified by combining multiple characteristics, so that good identification effect can be achieved. After the method for identifying the relationship between the sentiment words and the evaluation objects is adopted, experiments prove that better effect can be achieved; the method is suitable for being applied to the practical problem.
Description
Technical field
The present invention relates to natural language processing technique field and area of pattern recognition, particularly a kind of relation recognition method of emotion word and evaluation object.
Background technology
Since 21 century, develop rapidly along with internet, people express oneself viewpoint and emotion more and more on network, and this class text often exists with the form of comment on commodity, forum's comment, blog, and most of text can well reflect people's view and suggestion.And the emotion information of analyzing in these mass texts by artificial method need to expend a large amount of time, man power and material, under this background, sentiment analysis technology is arisen at the historic moment, and in natural language processing research field, obtained numerous researchers' concern, there is very large using value.
So-called text emotion analysis, analyzes speaker's attitude (or claiming viewpoint, emotion) exactly, namely the subjectivity information in text is analyzed.The research of sentiment analysis has been carried out for many years, and the main task of research concentrates on above emotional semantic classification subtask.But along with application problem in the urgent need to, the fine-grained emotion information such as word rank or phrase rank extracts and analytical technology starts to be gradually subject to numerous researchers' concern.For example, in following several concrete application, need fine-grained emotion information: a) viewpoint question and answer: about some problems of evaluation object entity, as " user likes which aspect of product X? " B) commending system: system will be clear that recommends those to obtain the good entity of evaluating at certain aspect concrete; C) suggestion summary: the viewpoint of all positive/negative of the Y aspect about entity X is done to a summary, and they are correspondingly cut apart.The common ground of above-mentioned task be system must be able to identify the viewpoint that single sentence comprises be specifically to evaluate which object, for the viewpoint of this object be commendation or derogatory sense, need to know what " evaluation object " be, and corresponding polarity.Therefore,, in building a concrete application system, except knowing the feeling polarities of text representation, also need other relevant informations that text emotion is expressed understand and analyze.
At present, the research of emotion information extraction task mainly concentrates on following three aspects: viewpoint holder (Opinion Holder), evaluation word (Polarity Word) and evaluation object (Opinion Target).Viewpoint holder's extraction is mainly towards newsletter archive, and identifying object is the person of being subordinate to of viewpoint/comment.Abstracting method about viewpoint holder is mainly the heuristic rule based on non-supervisory, roughly has the following aspects: a) abstracting method based on named entity recognition: the b) abstracting method based on semantic character labeling.Evaluating word is emotion word, refers to the word with emotion color, in emotion information extracts, plays very important effect.The abstracting method of evaluating word is mainly divided into based on two kinds of corpus and dictionaries.Evaluation object extract be in extracting comment text emotional expression towards object, this task is that emotion information extracts a task study task the most widely, has in succession occurred that a large amount of abstracting methods roughly can be divided into two parts: based on non-supervisory and abstracting method supervised learning.Wherein the relation recognition of emotion word and evaluation object is also a basic problem during emotion information extracts.
Summary of the invention
In view of this, the object of the invention is to propose a kind of relation recognition method of emotion word and evaluation object, the method can extract emotion word and the evaluation object in sentence, and whether identification exists corresponding relation between the two.
A kind of emotion word proposing according to this goal of the invention and the relation recognition method of evaluation object, comprising:
A) make language material: extract one section of Chinese text as language material, carry out word segmentation processing and part-of-speech tagging, obtain the part of speech of each word, in the Chinese text after participle, manually mark out emotion word and evaluation object corresponding in sentence;
B) training condition random field models: the good language material of above-mentioned mark is divided into three parts, chooses wherein first's language material and, as training set, train and obtain a conditional random field models;
C) extract emotion word and evaluation object: use described conditional random field models, extract respectively emotion word and evaluation object in second portion language material and third part language material, obtain destination file S1 and S2;
D) form emotion word and evaluation object candidate set: by the emotion word extracting in each sentence in described S1 and S2 and evaluation object combination of two, form emotion word-evaluation object candidate set, the relation recognition of emotion word and evaluation object is modeled as to binary classification problems;
E) training maximum entropy classifiers: the latent structure that extracts emotion word-evaluation object candidate set in described S1 is trained and obtained a maximum entropy classifiers as training set;
F) test maximum entropy classifiers: use emotion word and evaluation object in described S2 to test above-mentioned maximum entropy classifiers;
G) practical application: use described conditional random field models to extract and treat emotion word and the evaluation object in target Chinese text, use described maximum entropy classifiers to judge the emotion word and the evaluation object that extract, identify described emotion word and the relation of evaluation object.
Preferably, described step adopts Stamford parser to realize the participle of language material and part-of-speech tagging in a).
While preferably, described step b) using a conditional random field models of described training set training, this training set, by first's language material, selects word feature and part of speech to be characterized as extraction feature, is processed into the form of corpus.
Preferably, described step c) service condition random field models in, when the emotion word in extraction second portion language material and third part language material and evaluation object, first described second portion language material to be become to the form of corpus according to word feature and part of speech characteristic processing with third part language material.
Preferably, in described step e), first from emotion word-evaluation object candidate set of described S1 gained, choose the correct group of positive training examples as described maximum entropy classifiers that has corresponding relation, extract the positive example training sample of the feature composition and classification device of this emotion word and evaluation object; Then from emotion word-evaluation object candidate set of described S1 gained, choose and do not exist the candidate set of corresponding relation as the negative training examples of sorter, extract the negative routine training sample of the feature composition and classification device of emotion word and evaluation object; Finally, by training, obtain described maximum entropy classifiers.
Preferably, the feature of described emotion word and evaluation object is as follows:
Feature 1: the assemblage characteristic that the emotion word word feature in candidate set and the word feature of evaluation object form;
Feature 2: the assemblage characteristic that the emotion word word feature in candidate set and the part of speech feature of evaluation object form;
Feature 3: the assemblage characteristic that the emotion word part of speech feature in candidate set and evaluation object word feature form;
Feature 4: the assemblage characteristic that the emotion word part of speech feature in candidate set and evaluation object part of speech feature form;
Feature 5: whether emotion word and evaluation object in candidate set are arranged in same clause;
Feature 6: the emotion word in candidate set and the distance feature of evaluation object.
Preferably, step f) in, for same these 6 features that extract described emotion word and evaluation object of described S2, test, identify emotion word and evaluation object candidate set in S2 and whether have corresponding relation.
Preferably, described first language material, second portion language material and third part language material respectively account for 40%, 40% and 20%.
With respect to prior art, the embodiment of the present invention can take into full account the relation of emotion word and evaluation object, use maximum entropy classifiers and combined various features and gone emotion word and evaluation object candidate set that condition for identification random field models extracts whether to have corresponding relation, obtained good recognition effect.This is due to the feature that the embodiment of the present invention provides, to can be good at reacting the corresponding relation of emotion word and evaluation object.The embodiment of the present invention is considered as binary classification problems by the relation of emotion word and evaluation object, and a kind of relation recognition method based on maximum entropy classifiers is provided, and has obtained good effect.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, to the accompanying drawing of required use in embodiment or description of the Prior Art be briefly described below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skills, do not paying under the prerequisite of creative work, can also obtain according to these accompanying drawings other accompanying drawing.
The schematic flow sheet of a kind of emotion word that Fig. 1 provides for the embodiment of the present invention and the relation recognition method of evaluation object.
Fig. 2 is the experimental performance that extracts emotion word and evaluation object in the embodiment of the present invention.
Fig. 3 is the relation recognition result of emotion word and evaluation object in the embodiment of the present invention.
Embodiment
The embodiment of the present invention provides a kind of relation recognition method of emotion word and evaluation object, taken into full account the relation between emotion word and evaluation object, realization utilizes maximum entropy classifiers identification whether to have corresponding relation, disclose 3 category features and learnt for maximum entropy classifiers, obtained good recognition effect.
Below with reference to accompanying drawing, describe according to a preferred embodiment of the invention.
Fig. 1 is emotion word of the present invention and evaluation object relation recognition method flow diagram, and concrete steps are described below:
101, make language material: extract one section of Chinese text as language material, carry out word segmentation processing and part-of-speech tagging, obtain the part of speech of each word, in the Chinese text after participle, manually mark out emotion word and evaluation object corresponding in sentence.
Chinese word segmentation refers to a Chinese character sequence is cut into independent one by one word, exactly continuous word sequence is reassembled into the process of word sequence according to certain standard.As become " I like tourism " after " I like tourism " participle.Part-of-speech tagging refers to the part of speech (as: noun, verb, adjective etc.) of indicating word.Participle and part-of-speech tagging are used existing participle instrument, as most probable number method, maximum matching method, condition random field method etc.The present invention adopts Stamford parser (Stanford Parser instrument) to realize evaluating participle and the part-of-speech tagging of language material.Wherein Stanford Parser is an existing public instrument.Certainly, the present invention also can adopt other participle and part-of-speech tagging instrument.
In text after participle, manually mark out in sentence corresponding evaluation object and emotion word, in mark process, ignore the situation that lacks emotion word in sentence or lack evaluation object.
102, training condition random field models: the good language material of above-mentioned mark is divided into three parts, chooses wherein first's language material and, as training set, train and obtain a conditional random field models.
This wherein, first from existing language material, choosing at random 40% language material is that language material ，Gai first of first language material is as the training set of conditional random field models; Then Dui Gai first language material extracts the form that characteristic processing becomes corpus, the feature is here selected word feature and part of speech feature (wherein in embodiments of the present invention, word feature refers to the morphology of current word, morphology feature often can determine whether a word is evaluation object, or emotion word, and part of speech feature refers to the part of speech of current word, because emotion word is often adjective, evaluation object is noun or noun phrase often, so these two features can identify emotion word and the evaluation object in sentence preferably); Finally train, obtaining extraction model M1(is conditional random field models).
In this step, the machine learning method that the embodiment of the present invention is selected is conditional random field models, below this model is summarized:
Conditional random field models (Conditional Random Fields, CRFs) be a kind of sequence labelling model based on statistics by propositions such as John Lafferty, it is under the condition of given input node, for calculating the non-directed graph model of the conditional probability of output node.Be often used as cutting apart and marking of sequence data.CRFs has been applied to a plurality of fields of natural language processing, as Chinese word segmentation, named entity recognition etc.
Lafferty is subject to the impact of maximum entropy very large on the selection of condition random field potential function, and the formal definition of potential function is as follows:
In the equation above, y|c represents a stochastic variable, the node in corresponding c the group of this variable, f
kbe a fundamental function, it is a Boolean type, and p (y|x) is:
Wherein Z (x) represents normalized factor,
At present, the realization of conventional condition random field probability model has CRF++, CRF Mallet toolkit etc., the sequence labelling model that the embodiment of the present invention selects CRF++ to extract as evaluation object, and the parameter adopting is acquiescence.
103, extract emotion word and evaluation object: use described conditional random field models, extract respectively emotion word and evaluation object in second portion language material and third part language material, obtain destination file S1 and S2;
Word feature and part of speech feature described in same optional step 102, using in residue language material, account for equally whole language materials 40% as second portion language material, be processed into the form of corpus; In recycling 102, resulting extraction model M1 tests, and extracts emotion word and evaluation object in this second portion language material, and this part result, using the training examples source as follow-up sorter, for the ease of narration, is called S1; Last 20% language material, as third part language material, utilizes resulting model M 1 in 102 to test equally, extracts emotion word and evaluation object in text, and this part language material, using the test sample source as follow-up sorter, is called S2.Fig. 2 is for extracting performance, and wherein, P (Precision) represents that accuracy rate, R (Recall) represent that recall rate, F (F1-Measure) are F1 value.As can be seen from Figure, in the situation that only making word feature and part of speech feature, obtained good result, for subsequent step is laid a good foundation.
104, form emotion word and evaluation object candidate set: by the emotion word extracting in each sentence in described S1 and S2 and evaluation object combination of two, form emotion word-evaluation object candidate set, the relation recognition of emotion word and evaluation object is modeled as to binary classification problems.
In this step, first the destination file S1 to two parts of gained in 103 steps, S2 processes, by the emotion word extracting in each sentence and evaluation object combination of two, form emotion word-evaluation object candidate set, the relation recognition of emotion word and evaluation object is modeled as to binary classification problems.
105, training maximum entropy classifiers: the latent structure that extracts emotion word-evaluation object candidate set in described S1 is trained and obtained a maximum entropy classifiers as training set.
First from choose the correct group of positive training examples as sorter that has corresponding relation step 104 in emotion word-evaluation object candidate set of S1 gained, extract the positive example training sample T2 of the feature composition and classification device of emotion word and evaluation object; Then from the candidate set by S1 gained, choose and do not exist the candidate set of corresponding relation as the negative training examples of sorter, extract the negative routine training sample T3 of the feature composition and classification device of emotion word and evaluation object; Finally, by training, obtain a maximum entropy classifiers M2.
This wherein, we chosen 3 groups totally six features carry out training classifier, details are as follows for feature:
For the ease of understanding, candidate set " good-looking "-" book " of take is example, and the part of speech of " good-looking " is VA, and the part of speech of " book " is NN,
Feature 1: the assemblage characteristic that the emotion word word feature in candidate set and the word feature of evaluation object form, is about to " good-looking _ book " as feature;
Feature 2: the assemblage characteristic that the emotion word word feature in candidate set and the part of speech feature of evaluation object form, i.e. " good-looking _ NN ";
Feature 3: the assemblage characteristic that the emotion word part of speech feature in candidate set and evaluation object word feature form, i.e. " VA_ book ";
Feature 4: the assemblage characteristic that the emotion word part of speech feature in candidate set and evaluation object part of speech feature form, i.e. " VA_NN ";
Feature 5: whether emotion word and evaluation object in candidate set are arranged in same clause, this is boolean's feature, if emotion word and evaluation object are arranged in same clause, eigenwert is " 1 ", otherwise be " 0 ", wherein clause refers to the part after sentence is cut apart with punctuate;
Feature 6: the emotion word in candidate set and the distance feature of evaluation object, the word counting of this feature after with participle, eigenwert is the word quantity of being separated by between emotion word and evaluation object.
In this step, the machine learning method that the embodiment of the present invention is selected is maximum entropy sorting technique, below this model is summarized:
Maximum entropy sorting technique is based on maximum entropy information theory, and its basic thought is to set up model for all known factors, and the factor of all the unknowns is foreclosed.To find a kind of probability distribution, meet all known facts, but allow the randomization of unknown factor.With respect to naive Bayesian method, it is independent that the feature of the method maximum is exactly the condition that does not need to meet between feature and feature.Therefore, the method is applicable to the various different features of statistics, and without the impact of considering between them.
Under maximum entropy model, the formula of predicted condition probability P (c|D) is as follows:
Wherein Z (D) is normalized factor.F
k,cbe fundamental function, be defined as:
106, test maximum entropy classifiers: use emotion word and evaluation object in described S2 to test above-mentioned maximum entropy classifiers.
First from choose the correct group of just test sample as sorter that has corresponding relation step 104 in emotion word-evaluation object candidate set of S2 gained, extract the positive example test sample book L3 of the feature composition and classification device of emotion word and evaluation object; Then from the candidate set by S2 gained, choose and do not exist the candidate set of corresponding relation as the negative testing sample of sorter, extract the negative routine test sample book L4 of the feature composition and classification device of emotion word and evaluation object; Finally, by the sorter M2 test sample book of gained in 105, whether emotion word and the evaluation object of judging candidate set there is corresponding relation.
107, practical application: after obtaining above-mentioned conditional random field models and maximum entropy classifiers, apply these two models each sentence in Chinese text to be identified is analyzed and judged, thereby identify the corresponding relation of emotion word and evaluation object in the text.During concrete operations, first the described conditional random field models of this use extracts and treats emotion word and the evaluation object in target Chinese text, then use described maximum entropy classifiers to judge the emotion word and the evaluation object that extract, identify described emotion word and the relation of evaluation object.
With text, " I like football, but I dislike basketball " be example, emotion word in the text comprises " liking " and " disliking " two, evaluation object in the text comprises " football " and " basketball " two, when applying method of the present invention it being identified, conditional random field models is by after training, can extract above-mentioned two emotion words and two evaluation objects, and make it to become " liking-football ", " like-basketball ", " disagreeable-football ", the candidate set of " disagreeable-basketball " four groups of emotion word-evaluation objects, then utilize maximum entropy classifiers, every group of candidate set judged, identify wherein correct combination, thereby obtain the emotion information in the text.
Fig. 3 is that the embodiment of the present invention adds the results of property after each stack features, and as seen from the figure, the embodiment of the present invention can judge between emotion word and evaluation object, whether there is corresponding relation comparatively accurately.
Above-mentioned explanation to the disclosed embodiments, makes professional and technical personnel in the field can realize or use the present invention.To the multiple modification of these embodiment, will be apparent for those skilled in the art, General Principle as defined herein can, in the situation that not departing from the spirit or scope of the present invention, realize in other embodiments.Therefore, the present invention will can not be restricted to embodiment illustrated herein, but will meet the widest scope consistent with principle disclosed herein and features of novelty.
Claims (8)
1. a relation recognition method for emotion word and evaluation object, is characterized in that, comprising:
A) make language material: extract one section of Chinese text as language material, carry out word segmentation processing and part-of-speech tagging, obtain the part of speech of each word, in the Chinese text after participle, manually mark out emotion word and evaluation object corresponding in sentence;
B) training condition random field models: the good language material of above-mentioned mark is divided into three parts, chooses wherein first's language material and, as training set, train and obtain a conditional random field models;
C) extract emotion word and evaluation object: use described conditional random field models, extract respectively emotion word and evaluation object in second portion language material and third part language material, obtain destination file S1 and S2;
D) form emotion word and evaluation object candidate set: by the emotion word extracting in each sentence in described S1 and S2 and evaluation object combination of two, form emotion word-evaluation object candidate set, the relation recognition of emotion word and evaluation object is modeled as to binary classification problems;
E) training maximum entropy classifiers: the latent structure that extracts emotion word-evaluation object candidate set in described S1 is trained and obtained a maximum entropy classifiers as training set;
F) test maximum entropy classifiers: use emotion word and evaluation object in described S2 to test above-mentioned maximum entropy classifiers;
G) practical application: use described conditional random field models to extract and treat emotion word and the evaluation object in target Chinese text, use described maximum entropy classifiers to judge the emotion word and the evaluation object that extract, identify described emotion word and the relation of evaluation object.
2. recognition methods according to claim 1, is characterized in that: described step a) the middle Stamford parser that adopts realizes the participle of language material and part-of-speech tagging.
3. recognition methods according to claim 1, it is characterized in that: while described step b) using a conditional random field models of described training set training, this training set, by first's language material, selects word feature and part of speech to be characterized as extraction feature, is processed into the form of corpus.
4. recognition methods according to claim 1, it is characterized in that: service condition random field models described step c), when the emotion word in extraction second portion language material and third part language material and evaluation object, first described second portion language material to be become to the form of corpus according to word feature and part of speech characteristic processing with third part language material.
5. recognition methods according to claim 1, it is characterized in that: in described step e), first from emotion word-evaluation object candidate set of described S1 gained, choose the correct group of positive training examples as described maximum entropy classifiers that has corresponding relation, extract the positive example training sample of the feature composition and classification device of this emotion word and evaluation object; Then from emotion word-evaluation object candidate set of described S1 gained, choose and do not exist the candidate set of corresponding relation as the negative training examples of sorter, extract the negative routine training sample of the feature composition and classification device of emotion word and evaluation object; Finally, by training, obtain described maximum entropy classifiers.
6. recognition methods according to claim 5, is characterized in that: the feature of described emotion word and evaluation object is as follows:
Feature 1: the assemblage characteristic that the emotion word word feature in candidate set and the word feature of evaluation object form;
Feature 2: the assemblage characteristic that the emotion word word feature in candidate set and the part of speech feature of evaluation object form;
Feature 3: the assemblage characteristic that the emotion word part of speech feature in candidate set and evaluation object word feature form;
Feature 4: the assemblage characteristic that the emotion word part of speech feature in candidate set and evaluation object part of speech feature form;
Feature 5: whether emotion word and evaluation object in candidate set are arranged in same clause;
Feature 6: the emotion word in candidate set and the distance feature of evaluation object.
7. recognition methods according to claim 6, it is characterized in that: step f), for same these 6 features that extract described emotion word and evaluation object of described S2, test, identify emotion word and evaluation object candidate set in S2 and whether have corresponding relation.
8. recognition methods according to claim 1, is characterized in that: described first language material, second portion language material and third part language material respectively account for 40%, 40% and 20%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310693087.0A CN103631961B (en) | 2013-12-17 | 2013-12-17 | Method for identifying relationship between sentiment words and evaluation objects |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310693087.0A CN103631961B (en) | 2013-12-17 | 2013-12-17 | Method for identifying relationship between sentiment words and evaluation objects |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103631961A true CN103631961A (en) | 2014-03-12 |
| CN103631961B CN103631961B (en) | 2017-01-18 |
Family
ID=50213002
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310693087.0A Active CN103631961B (en) | 2013-12-17 | 2013-12-17 | Method for identifying relationship between sentiment words and evaluation objects |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103631961B (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104268160A (en) * | 2014-09-05 | 2015-01-07 | 北京理工大学 | Evaluation object extraction method based on domain dictionary and semantic roles |
| CN104298665A (en) * | 2014-10-16 | 2015-01-21 | 苏州大学 | Identification method and device of evaluation objects of Chinese texts |
| CN105095302A (en) * | 2014-05-15 | 2015-11-25 | 财团法人工业技术研究院 | Public praise-oriented analysis and inspection system, device and method |
| CN105786929A (en) * | 2014-12-26 | 2016-07-20 | 阿里巴巴集团控股有限公司 | Information monitoring method and device |
| CN106339368A (en) * | 2016-08-24 | 2017-01-18 | 乐视控股（北京）有限公司 | Text emotional tendency acquiring method and device |
| CN106897439A (en) * | 2017-02-28 | 2017-06-27 | 百度在线网络技术（北京）有限公司 | The emotion identification method of text, device, server and storage medium |
| CN107102980A (en) * | 2016-02-19 | 2017-08-29 | 北京国双科技有限公司 | The extracting method and device of emotion information |
| CN107220238A (en) * | 2017-05-24 | 2017-09-29 | 电子科技大学 | A kind of text object abstracting method based on Mixed Weibull distribution |
| CN107526721A (en) * | 2017-06-21 | 2017-12-29 | 深圳美云智数科技有限公司 | A kind of disambiguation method and device to electric business product review vocabulary |
| CN108255803A (en) * | 2016-12-29 | 2018-07-06 | 北京国双科技有限公司 | The judgment method and device of document emotion |
| CN108776652A (en) * | 2018-05-21 | 2018-11-09 | 众安信息技术服务有限公司 | A kind of forecast for market tendency method based on news corpus |
| CN109086340A (en) * | 2018-07-10 | 2018-12-25 | 太原理工大学 | Evaluation object recognition methods based on semantic feature |
| CN109271627A (en) * | 2018-09-03 | 2019-01-25 | 深圳市腾讯网络信息技术有限公司 | Text analyzing method, apparatus, computer equipment and storage medium |
| CN109299400A (en) * | 2018-09-06 | 2019-02-01 | 北京奇艺世纪科技有限公司 | A kind of viewpoint abstracting method, device and equipment |
| CN109325238A (en) * | 2018-10-31 | 2019-02-12 | 北京国信云服科技有限公司 | A kind of method of multiple entity sentiment analysis in long text |
| CN109472026A (en) * | 2018-10-31 | 2019-03-15 | 北京国信云服科技有限公司 | Accurate emotion information extracting methods a kind of while for multiple name entities |
| CN110765769A (en) * | 2019-08-27 | 2020-02-07 | 电子科技大学 | Entity attribute dependency emotion analysis method based on clause characteristics |
| CN111126046A (en) * | 2019-12-06 | 2020-05-08 | 腾讯云计算（北京）有限责任公司 | Statement feature processing method and device and storage medium |
| CN111339752A (en) * | 2020-02-18 | 2020-06-26 | 哈尔滨工业大学 | Evaluation object-oriented emotion analysis method for multi-task joint learning |
| CN111753058A (en) * | 2020-06-30 | 2020-10-09 | 北京信息科技大学 | Text viewpoint mining method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010250814A (en) * | 2009-04-14 | 2010-11-04 | Nec (China) Co Ltd | Part-of-speech tagging system, training device and method of part-of-speech tagging model |
| CN102200969A (en) * | 2010-03-25 | 2011-09-28 | 日电（中国）有限公司 | Text sentiment polarity classification system and method based on sentence sequence |
| CN102682124A (en) * | 2012-05-16 | 2012-09-19 | 苏州大学 | Emotion classifying method and device for text |
| CN103034626A (en) * | 2012-12-26 | 2013-04-10 | 上海交通大学 | Emotion analyzing system and method |
-
2013
- 2013-12-17 CN CN201310693087.0A patent/CN103631961B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010250814A (en) * | 2009-04-14 | 2010-11-04 | Nec (China) Co Ltd | Part-of-speech tagging system, training device and method of part-of-speech tagging model |
| CN102200969A (en) * | 2010-03-25 | 2011-09-28 | 日电（中国）有限公司 | Text sentiment polarity classification system and method based on sentence sequence |
| CN102682124A (en) * | 2012-05-16 | 2012-09-19 | 苏州大学 | Emotion classifying method and device for text |
| CN103034626A (en) * | 2012-12-26 | 2013-04-10 | 上海交通大学 | Emotion analyzing system and method |
Non-Patent Citations (1)
| Title |
|---|
| 张慧等: "基于评价对象类型的跨领域情感分类方法研究", 《计算机科学》 * |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105095302A (en) * | 2014-05-15 | 2015-11-25 | 财团法人工业技术研究院 | Public praise-oriented analysis and inspection system, device and method |
| CN105095302B (en) * | 2014-05-15 | 2019-05-17 | 财团法人工业技术研究院 | Public praise-oriented analysis and inspection system, device and method |
| CN104268160A (en) * | 2014-09-05 | 2015-01-07 | 北京理工大学 | Evaluation object extraction method based on domain dictionary and semantic roles |
| CN104268160B (en) * | 2014-09-05 | 2017-06-06 | 北京理工大学 | A kind of OpinionTargetsExtraction Identification method based on domain lexicon and semantic role |
| CN104298665A (en) * | 2014-10-16 | 2015-01-21 | 苏州大学 | Identification method and device of evaluation objects of Chinese texts |
| CN105786929A (en) * | 2014-12-26 | 2016-07-20 | 阿里巴巴集团控股有限公司 | Information monitoring method and device |
| CN105786929B (en) * | 2014-12-26 | 2019-09-03 | 阿里巴巴集团控股有限公司 | A kind of information monitoring method and device |
| CN107102980A (en) * | 2016-02-19 | 2017-08-29 | 北京国双科技有限公司 | The extracting method and device of emotion information |
| CN106339368A (en) * | 2016-08-24 | 2017-01-18 | 乐视控股（北京）有限公司 | Text emotional tendency acquiring method and device |
| CN108255803A (en) * | 2016-12-29 | 2018-07-06 | 北京国双科技有限公司 | The judgment method and device of document emotion |
| CN106897439B (en) * | 2017-02-28 | 2020-04-14 | 百度在线网络技术（北京）有限公司 | Text emotion recognition method, device, server and storage medium |
| CN106897439A (en) * | 2017-02-28 | 2017-06-27 | 百度在线网络技术（北京）有限公司 | The emotion identification method of text, device, server and storage medium |
| CN107220238A (en) * | 2017-05-24 | 2017-09-29 | 电子科技大学 | A kind of text object abstracting method based on Mixed Weibull distribution |
| CN107526721A (en) * | 2017-06-21 | 2017-12-29 | 深圳美云智数科技有限公司 | A kind of disambiguation method and device to electric business product review vocabulary |
| CN107526721B (en) * | 2017-06-21 | 2020-07-10 | 深圳美云智数科技有限公司 | Ambiguity elimination method and device for comment vocabularies of e-commerce products |
| CN108776652A (en) * | 2018-05-21 | 2018-11-09 | 众安信息技术服务有限公司 | A kind of forecast for market tendency method based on news corpus |
| CN108776652B (en) * | 2018-05-21 | 2022-04-01 | 众安信息技术服务有限公司 | Market forecasting method based on news corpus |
| CN109086340A (en) * | 2018-07-10 | 2018-12-25 | 太原理工大学 | Evaluation object recognition methods based on semantic feature |
| CN109271627A (en) * | 2018-09-03 | 2019-01-25 | 深圳市腾讯网络信息技术有限公司 | Text analyzing method, apparatus, computer equipment and storage medium |
| CN109271627B (en) * | 2018-09-03 | 2023-09-05 | 深圳市腾讯网络信息技术有限公司 | Text analysis method, apparatus, computer device and storage medium |
| CN109299400A (en) * | 2018-09-06 | 2019-02-01 | 北京奇艺世纪科技有限公司 | A kind of viewpoint abstracting method, device and equipment |
| CN109472026A (en) * | 2018-10-31 | 2019-03-15 | 北京国信云服科技有限公司 | Accurate emotion information extracting methods a kind of while for multiple name entities |
| CN109325238A (en) * | 2018-10-31 | 2019-02-12 | 北京国信云服科技有限公司 | A kind of method of multiple entity sentiment analysis in long text |
| CN110765769A (en) * | 2019-08-27 | 2020-02-07 | 电子科技大学 | Entity attribute dependency emotion analysis method based on clause characteristics |
| CN110765769B (en) * | 2019-08-27 | 2023-05-02 | 电子科技大学 | Clause feature-based entity attribute dependency emotion analysis method |
| CN111126046A (en) * | 2019-12-06 | 2020-05-08 | 腾讯云计算（北京）有限责任公司 | Statement feature processing method and device and storage medium |
| CN111126046B (en) * | 2019-12-06 | 2023-07-14 | 腾讯云计算（北京）有限责任公司 | Sentence characteristic processing method and device and storage medium |
| CN111339752A (en) * | 2020-02-18 | 2020-06-26 | 哈尔滨工业大学 | Evaluation object-oriented emotion analysis method for multi-task joint learning |
| CN111753058A (en) * | 2020-06-30 | 2020-10-09 | 北京信息科技大学 | Text viewpoint mining method and system |
| CN111753058B (en) * | 2020-06-30 | 2023-06-02 | 北京信息科技大学 | Text viewpoint mining method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103631961B (en) | 2017-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103631961B (en) | Method for identifying relationship between sentiment words and evaluation objects | |
| Haque et al. | Sentiment analysis on large scale Amazon product reviews | |
| CN104794212B (en) | Context sensibility classification method and categorizing system based on user comment text | |
| CN102682124B (en) | Emotion classifying method and device for text | |
| CN105205124B (en) | A kind of semi-supervised text sentiment classification method based on random character subspace | |
| CN107491531A (en) | Chinese network comment sensibility classification method based on integrated study framework | |
| Shirsat et al. | Document level sentiment analysis from news articles | |
| Valakunde et al. | Multi-aspect and multi-class based document sentiment analysis of educational data catering accreditation process | |
| Tromp et al. | Senticorr: Multilingual sentiment analysis of personal correspondence | |
| Khatri et al. | Using sentimental analysis in prediction of stock market investment | |
| CN102682130A (en) | Text sentiment classification method and system | |
| CN110134934A (en) | Text emotion analysis method and device | |
| CN112287197B (en) | Method for detecting sarcasm of case-related microblog comments described by dynamic memory cases | |
| CN103020249A (en) | Classifier construction method and device as well as Chinese text sentiment classification method and system | |
| CN105740382A (en) | Aspect classification method for short comment texts | |
| CN104346326A (en) | Method and device for determining emotional characteristics of emotional texts | |
| Pramodh et al. | Automatic personality recognition of authors using big five factor model | |
| Jayasanka et al. | Sentiment analysis for social media | |
| Haque et al. | Opinion mining from bangla and phonetic bangla reviews using vectorization methods | |
| Das et al. | Sentiment analysis of movie reviews using POS tags and term frequencies | |
| Reddy et al. | N-gram approach for gender prediction | |
| CN105760502A (en) | Commercial quality emotional dictionary construction system based on big data text mining | |
| Addepalli et al. | A proposed framework for measuring customer satisfaction and product recommendation for ecommerce | |
| Kim et al. | Comparing machine learning classifiers for movie WOM opinion mining | |
| Hu et al. | Improving automated patent claim parsing: dataset, system, and experiments |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20170914 Address after: Suzhou City, Jiangsu province 215123 Xiangcheng District Ji Road No. 8 Patentee after: Soochow University Address before: Zhangjiagang mayor Jingyang Road Suzhou City, Jiangsu province 215600 No. 10 Patentee before: Zhangjiagang Institute of Industrial Technologies Soochow University |
|
| TR01 | Transfer of patent right |