CN103995803B - A kind of fine granularity text sentiment analysis method - Google Patents
A kind of fine granularity text sentiment analysis method Download PDFInfo
- Publication number
- CN103995803B CN103995803B CN201410178056.6A CN201410178056A CN103995803B CN 103995803 B CN103995803 B CN 103995803B CN 201410178056 A CN201410178056 A CN 201410178056A CN 103995803 B CN103995803 B CN 103995803B
- Authority
- CN
- China
- Prior art keywords
- emotion
- sentence
- word
- fine granularity
- benchmark
- 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.)
- Expired - Fee Related
Links
Landscapes
- Machine Translation (AREA)
Abstract
The invention discloses a kind of fine granularity text sentiment analysis method, step one：Build fine granularity sentiment dictionary；Step 2：Sentence structure relation judges；Step 3：The emotion value evaluation of simple sentence.The present invention can extract more user feeling relevant informations that text is included, and can preferably portray the impression of user's heart, the application study for supporting correlation, such as user emotion state and the situation of change analysis based on health.
Description
Technical field
The invention belongs to English text sentiment analysis technical field, it is related to a kind of fine granularity text sentiment analysis method, has
Say body, be related to a kind of fine granularity sentiment analysis method for comment text.
Background technology
The emotion of the mankind is complicated and many.Relation due to the complexity of emotion and with other something outsides,
It belongs in psychology the most phenomenon of challenge.The traditional approach for understanding the current mood of people has various：Such as can be with
Its subjective feeling is seeked advice from, the change in its countenance or behavior, and its physiological change is observed.In fact, the feelings of people
Thread is complicated, can not be directly measured, and is only capable of being recognized by their external expressive form, is so just expedited the emergence of out various
Method for recognizing human emotion.In general, the method for one emotional reactions of people of most common identification substantially may be used
To be divided three classes：(1) self-report, (2) physiology method, (3) behavior observation.
Substantial amounts of man power and material is needed based on the personal emotional state of conventional method monitoring, it is difficult to obtain a large number of users long
The mood related data of time.With continuing to develop for online social networks, its number of users for possessing constantly increases, Ren Menkai
Begin to be accustomed to often feeling the finding of oneself to be shared with good friend on line, thus researcher can be obtained by the API of social network sites
A large number of users mood related data, the emotional state of user is extracted based on text emotion analytical technology.
Text emotion analysis is an emerging research topic, with very big researching value and application value.Patent
200910219161.9 estimate topic language model according to the Expression of language of different themes text, calculate pending text
The distance of language model and positive and negative emotion model, the Sentiment orientation of the nearest emotion model of selected distance assigns the text.Patent
The 200910083522.1 initial emotions point that test text is determined according to the label of training text, institute is utilized based on figure sort algorithm
The emotion for stating the initial emotion point iterative calculation test text of test text is divided and is normalized, cross-cutting to solve
Emotion tendentiousness of text problem analysis.Patent 201210088366.X is based on positive and negative sentiment dictionary and judges all comprising descriptor
The polarity of sentence, front sentence polarity sum and negative sentence polarity sum in result of calculation set, so as to draw whole piece microblogging
Emotion tendency.Patent 201310000734.5 is by constructing a kind of DCRF models with Two-Level (bilayer) structure
The emotion tendency for realizing entity level judges that patent 201310036034.1 is using the pass between object properties and emotion word
The statistics of the relational implementation fine granularity emotion strength quantifies between connection information and emotion word and qualifier and calculating.
The user feeling that text is included mainly is divided into two classes by current existing sentiment analysis technology：Positively and negatively, exist
The division aspect of emotional category belongs to the text emotion analysis of coarseness, lost a large number of users emotion relevant information.In order to fill
User comment information contained is separately won to obtain, the impression of user is preferably portrayed, the present invention does further fine granularity to comment text
Sentiment analysis, just negative tendency will further divide respectively, and such as negative emotions can be angry, or sadness etc..
The content of the invention
It is an object of the invention to the defect for overcoming above-mentioned technology to exist, there is provided a kind of fine granularity text sentiment analysis side
Method, the method can more fully obtain emotion information contained by user comment, can preferably support the application study of correlation, for example
User emotion state and situation of change analysis based on health.
Its concrete technical scheme is：
Step one：Build fine granularity sentiment dictionary
International generally acknowledged benchmark emotional semantic classification is chosen as fine granularity emotional semantic classification, and using benchmark emotion word as all kinds of
Other seed emotion word, by wordNet, (by the psychologist of Princeton universities, linguist and Computer Engineer join
Close a kind of English dictionary based on cognitive linguistics of design) its TongYiCi CiLin is searched, and corresponding classification is put into, complete thin
The first step enlarging of granularity sentiment dictionary；
Word is divided into four classes by wordNet：Noun, verb, adverbial word and adjective；The name obtained by the extension of benchmark emotion word
Part of speech emotion set, and adjective in the same fashion according to benchmark emotion word, verb and adverbial word form, are built into respectively
The emotion set of its adjective, verb and adverbial word form；Generic emotion set, in addition to the difference of part of speech, not shadow
The calculating of emotion value is rung, then the emotion set under a classification is considered as a major class, so as to complete fine granularity sentiment dictionary
Second step is extended；
So far, the fine granularity sentiment dictionary of structure cannot also cover most emotion vocabulary；By remaining emotion word how
The problem of fine granularity emotional category is referred to, it is similar based on general knowledge on concept hierarchy to benchmark emotion word to be converted to analysis
Property, and assign it to the emotional category representated by similitude highest benchmark emotion word；Ultimate analysis categorization results, and it is complete
It is apt to defect that may be present；So far the enlarging of fine granularity sentiment dictionary is completed；
Step 2：Sentence structure relation judges
Judge whether there is conjunction in sentence, if, then it represents that the sentence is compound sentence, is somebody's turn to do according to relation rule between sentence
Sentence structure relation and the computation rule of sentence emotion value that conjunction is represented；If it is not, the sentence is simple sentence；
Step 3：The emotion value evaluation of simple sentence
If compound sentence, then it is split as two subordinate sentences and is processed；If simple sentence, then its emotion is directly calculated
Value；Now, it is considered to the emotion value assessment method of simple sentence, comment affection computation to consider theme word correlation, with descriptor without
The emotion word of pass can bring interference to affection computation；And theme is mainly embodied by the theme (subject and object) of sentence, then only
The nominal and Adjective affective characteristic words related to theme need to be considered；Passed judgement on according to sentence structure, dependence, emotion word
Degree and associated adverbs modification intensity, calculate the emotion value of simple sentence；
For the imperfect short sentence of the frequent appearance of comment text, using word part of speech, the accurate of dependence judgement is improved
Property；When the adjunctival before emotion word is long, sentence structure, word part of speech and dependence are combined, specific algorithm
It is as follows：Dependence is first detected, descriptor is found out, and then finds out the modified relationship for depending on descriptor, according to sentence structure point
The result of analysis obtains the nominal phrase that descriptor is constituted with its adjunctival, then analyzes the structure of this nominal phrase and repaiies
The part of speech of the word that decorations phrase is included, draws correct modified relationship；
Step 4：Comment text fine granularity affection computation
Sentence emotion value is obtained with reference to relation between sentence pattern and sentence；The emotion sum of all sentences is the overall feelings of comment text
Inductance value.
Compared with prior art, the beneficial effects of the invention are as follows：More user's feelings that text is included can be extracted
Sense relevant information, can preferably portray the impression of user's heart, the application study for supporting correlation, such as use based on health
Family emotional state and situation of change are analyzed.
Brief description of the drawings
Fig. 1 is fine granularity sentiment dictionary construction method flow chart of the invention；
Fig. 2 is text fine granularity sentiment analysis method flow diagram of the invention；
Fig. 3 is the example sentence sentence structure figure in present example.
Specific embodiment
Technical scheme is described in more detail with specific embodiment below in conjunction with the accompanying drawings.
Fine granularity sentiment dictionary construction method of the invention realizes flow as shown in figure 1, detailed step is as follows：
Step 101：Setting benchmark emotional category and seed emotion word.
So far, neither one the recognized standard is gone back in the division of psychology bound pair emotion, and the present invention is with scholar Ekman's
As a example by famous 6 benchmark emotion, specifically include：Happiness (happiness), sadness (sadness), anger (anger), fear is (probably
Fear), surprise (pleasantly surprised) and disgust (detest).First according to using 6 benchmark emotion words as seed emotion of all categories
Word, its TongYiCi CiLin is searched by wordNet, and is put into corresponding classification, and the first step for completing fine granularity sentiment dictionary expands
Build.
Step 102：According to synonym expanding sentiment dictionary.
Word is divided into four classes by wordNet：Noun, verb, adverbial word and adjective.We have been obtained by 6 benchmark emotions
The nominal emotion set that word extension is obtained, by adjective in the same fashion according to 6 benchmark emotion words, verb and adverbial word
Form, is built into the emotion set of its adjective, verb and adverbial word form respectively.For example, " joy (joy) " and " joyful is (fast
It is happy) " " happiness (happiness) " emotional category is belonged to, but belong to two emotion set, joy belongs to nominal set,
Joyful belongs to Adjective set.Generic emotion set, in addition to the difference of part of speech, has no effect on the meter of emotion value
Calculate, then the emotion set under a classification is considered as a major class by the present invention, so as to complete the second step of fine granularity sentiment dictionary
Enlarging.
Step 103：Based on general general knowledge storehouse expanding sentiment dictionary.
So far fine granularity sentiment dictionary includes more than 1000 word altogether, and this is obvious for analysis text emotion tendency
Deficiency, the word of also substantial amounts of expression people's emotion is uncovered.For example give expression to the verb cry (crying) of obvious emotion, its nothing
Method is added to sentiment dictionary by way of emotion set seed word synonym above extends.Based on general knowledge, we are usual
Will be considered that cry (crying) is larger with associating for sad (sadness) and angry (anger), it is sad or indignation that it often gives expression to main body
Mood.When the classification of fine granularity emotion word is carried out for cry, we are converted into analysis cry (crying) and six emotional semantic classifications
Representative emotion similitude based on general knowledge on concept hierarchy, similitude is higher, it is believed that main body is intended by out should
The probability of class emotion is bigger, then cry (crying) is added in the sentiment dictionary of such emotion.Emotion word is calculated to exist with emotional semantic classification
Similitude based on commonsense reasoning on concept hierarchy, general general knowledge storehouse is optimal selection, reflects the common sense knowledge of emotion and is
The a subset of general knowledge in these general knowledge storehouses.The present invention calculates emotion word and emotional semantic classification benchmark word using general general knowledge storehouse
Similitude, and assigned to similitude highest emotional category.
Step 104：Modification method based on the sentiment dictionary set up.
So far, the sentiment dictionary of structure can cover most of emotion word, but according to the analysis to matrix of consequence,
Also there is obvious defect, it has been found that for a number of adjective, although for its prototype with benchmark emotion word general
Read the correlation result on level and substantially conform to general knowledge, but the similarity result that its comparative degree and the superlative degree are obtained but all is 0.
Be to solve problems, we are that conventional adjective sets up comparative degree and highest vocabulary, its comparative degree and it is highest inherit and
Prototype identical classification and similarity.So far the enlarging of fine granularity sentiment dictionary is completed.
Text fine granularity sentiment analysis method of the invention realizes flow as shown in Fig. 2 detailed step is as follows：
Step 201：The structure of fine granularity sentiment dictionary.
By setting benchmark emotional category and seed emotion word, according to synonym expanding sentiment dictionary, based on general general knowledge
Storehouse expanding sentiment dictionary, defect correction based on the sentiment dictionary set up etc. four is partially completed the structure of sentiment dictionary, in detail
Step is referring to step 101~104.
Step 202：Sentence structure relation judges.
Judge whether there is conjunction in sentence, if, then it represents that the sentence is compound sentence, is somebody's turn to do according to relation rule between sentence
Sentence structure relation and the computation rule of sentence emotion value that conjunction is represented.If it is not, the sentence is simple sentence.
In one English sentence, often there is a conjunction for expressing different relations, such as but (turnover), if (assuming that), so (because
Really), and (arranged side by side), moreover (progressive).Different conjunctions can produce different influences to the emotion of sentence, and the present invention should
Calculated with following rules：
(1) conjunction rule in sentence：
1) the grammer custom commonly used based on people, if without but, a general sentence expresses a tendency.It is right
In comprising table side by side, the sentence of the conjunction of progressive relationship, then the rule in application document assign different weights.
For example：The camera takes great pictures and has a long battery life. this sentence
Mean：This camera photographic effect is good, and battery life is long.Generally, if we know that great is positive, then long mono-
As be also positive
(2) conjunction rule between sentence：
1) based on the assumption that, people between sentence also often expression identical tendency.Unless there are but (but), however
(but) etc..
2) for the conjunction comprising different relations, we assign its guiding clause different weights, generally progressive pass
It is that the subordinate clause that conjunction is guided has the tendency of the affectional change of subordinate clause for strengthening Sentiment orientation, cause and effect and the guiding of coordination conjunction
It is not very big, and assumes that the subordinate clause of relation conjunction guiding is often the imagination to realistic situation, its precondition is in language performance
In serve prior effect, then generally require reduction hypothetical sentence rear half point.For the subordinate clause that other conjunctions are guided it
Sentiment orientation weighing factor setting, present invention application general rule.
3) for other universal relation conjunctions, such as represent that precedence conjunction first (first) does not interfere with it and draws substantially
The Sentiment orientation and intensity of the subordinate clause led；For the conjunction of table time, as when (when ... when) if do not consider its emotion
Tendency.
The conjunction of table turning relation, by but (but) as a example by, no matter as conjunction between conjunction or sentence in sentence, its guiding
Subordinate clause contrast previous contents have turnover semantic, but the positive and negative tendency of its emotion is but not necessarily conversely, i.e. analysis text emotion inclines
Xiang Shi, but (but) guiding subordinate clause Sentiment orientation, not necessarily with its before sentence Sentiment orientation conversely, through analysis deposit
In three kinds of situations：There is no obvious Sentiment orientation and the sentence before it to be inclined to phase with the sentence consistent, subordinate clause of tendency before it
Instead.It is exemplified below respectively：
1st, the first situation：I′m sure the wines are amazing too but I didn′t regret
getting a pint ofthe IPA.
Sentence justice：I determines that those wine are also very good, but I does not regret have selected one pint of this beverage.
2nd, second situation：Cool to see but don′t get too close.
Sentence justice：But this seems very cruel should not lean on too near.
3rd, the third situation：It′s not that cheap as they say.But good for who are
crazy with brands！
Sentence justice：What they did not said is so cheap.But it is fine for the people for liking ardently brand.
For the treatment rule of three cases above：
If but (but) guiding subordinate clause, consistent with the short sentence tendency before it, then its subordinate clause has progressive implication, then
Aggravate the weight of its subordinate clause；If but (but) guiding subordinate clause there is no obvious Sentiment orientation, may only as explanation or point out
Deng then only considering the Sentiment orientation of its previous sentence；If but (but) guiding subordinate clause with its before subordinate clause tendency conversely,
Then there is emotion reversion, but (but) before the subordinate sentence meaning can weaken, and the subordinate sentence after prominent turnover.
Step 203：The emotion value evaluation of simple sentence.
In a sentence, the Sentiment orientation and intensity of emotion phrase by comprising emotion word and depend on modification thereon
Composition is determined.The present invention extracts each participle unit in sentence inside using Stanford Parser (parser of Stamford)
Between dependence pair.The prevailing relationship pair that the present invention is applied to, as shown in table 1.
Table 1 relies primarily on relation pair
| Title | Relation |
| nsubj | Subject and predicate |
| dobj | Object and predicate |
| amod | Adjective is modified |
| advmod | Adverbial word is modified |
| comod | Coordination |
| neg | Negative modification |
Base sheet in sentiment dictionary is noun, verb, adverbial word and adjective, and adverbial word is generally used for repairing in English text
Decorations verb and adjective, then emotion word only possibly be present at noun, in verb and adjective these three words, inquire about sentiment dictionary
Its correspondence feeling polarities can then be found out.For adverbial word, we build degree adverb table, include conventional adverbial word and its corresponding repair
Decorations intensity.
Rule-based text emotion analysis, the Sentiment orientation of its sentence is the Sentiment orientation by calculating emotion word in sentence
Get, but if occurring negative word before emotion word, then can shine into influence to the Sentiment orientation of sentence or intensity, for example " I do
Not like it. (I does not like it) ", like (liking) itself express positive mood, but because have not before it
(no) modify, then the Sentiment orientation of whole sentence there occurs polarity inversion, be transformed into negative mood.Through analysis, uncertainty relation can go out
On present verb and Adjective emotion word, to its Sentiment orientation there may be the influence in two kinds of degree：1st, Sentiment orientation hair
Raw polarity inversion, intensity is substantially unaffected.2nd, Sentiment orientation is constant, but remitted its fury.
Situation one：Try it, you won ' t regret.
Sentence justice：Try, what you will not regret.
Situation two：Not the best restaurant but always have something to eat.
Sentence justice：It is not best restaurant, but still has the thing can to eat.
Comment is compared with general article sentence, the characteristics of have one prominent：Often there is imperfect brief comment,
I.e. sentence structure is imperfect, lacks one or several, such as sentence in subject, predicate or object ' Highly
Recommended. (it is strongly recommended that) ' lack subject and object.Incomplete sentence structure, the detection for increasing dependence is difficult
Degree, causes to detect the decline of accuracy, by previous sentence ' Highly recommended. (it is strongly recommended that) ' as a example by, due to this
Sentence without complete sentence structure, syntactic analysis its be difficult to distinguish that Highly's (strong) and recommended (recommendation) is correct
Dependence, it is a noun to take for Highly, is nsubj (subject-predicate relation) by the judgement of dependence mistake, causes meter
The emotion value of recommended is only considered during calculation, and does not consider booster actions of the Highly to its emotion intensity.For this
Incomplete short sentence, the present invention takes the part of speech for first obtaining word, the method for recycling dependence.
When the adjunctival before an emotion word is long, the difficulty for detecting dependence is caused to increase, result of detection
Accuracy reduction.By sentence ' as a example by There is a completely unexpected great hotel. ', the sheet of sentence
Meaning is to say：Here there is fabulous (great) hotel of complete (completely) (unexpected) beyond imagination.
Completely is used to modify unexpected, Completely unexpected modifications great, completely
Unexpected great modify hotel, because the level modified is deeper, cause the detection of dependence mistake, mistake occur
Detection modify hotel for completely unexpected, great individually modifies hotel, have impact on emotion Strength co-mputation
Accuracy.But a completely unexpected can be clearly seen that by the sentence structure (as shown in Figure 3) of this
Great hotel are the nominal phrases based on descriptor hotel, and completely and unexpected collectively form adverbial word
Phrase modifies great.
Finally, according to the dependence for detecting, the descriptor of sentence is extracted, according to sentence structure, dependence, feelings
Sense word degree of passing judgement on and associated adverbs modification intensity, calculating depends on the emotion value of descriptor, so as to draw the emotion value of simple sentence.
Step 204：Comment text fine granularity affection computation
English is divided into four types by purposes：Declarative sentence, imperative sentence, exclamative sentence, interrogative sentence.Imperative sentence is typicallyed represent please
Ask, order, seldom occur in comment text, the present invention does not consider the Sentiment orientation of this sentence pattern.For usually going out in comment
Existing declarative sentence, exclamative sentence and interrogative sentence, are detected, and assign different weights using punctuation mark and sentence pattern keyword.
Sentence emotion value is obtained with reference to relation between sentence pattern and sentence.The emotion sum of all sentences is the overall feelings of comment text
Inductance value.
The above, preferably specific embodiment only of the invention, protection scope of the present invention not limited to this are any ripe
Those skilled in the art are known in the technical scope of present disclosure, the letter of the technical scheme that can be become apparent to
Altered or equivalence replacement are each fallen within protection scope of the present invention.
Claims (1)
1. a kind of fine granularity text sentiment analysis method, it is characterised in that comprise the following steps：
Step one：Build fine granularity sentiment dictionary
International generally acknowledged benchmark emotional semantic classification is chosen as fine granularity emotional semantic classification, and using benchmark emotion word as of all categories
Seed emotion word, its TongYiCi CiLin is searched by wordNet, and is put into corresponding classification, completes fine granularity sentiment dictionary
The first step is extended；
Word is divided into four classes by wordNet：Noun, verb, adverbial word and adjective；Obtain nominal by the extension of benchmark emotion word
Emotion set, adjective in the same fashion according to benchmark emotion word, verb and adverbial word form, is built into it and describes respectively
The emotion set of word, verb and adverbial word form；Generic emotion set, in addition to the difference of part of speech, has no effect on emotion
The calculating of value, then be considered as a major class by the emotion set under a classification, so as to complete the second step of fine granularity sentiment dictionary
Enlarging；
So far, the fine granularity sentiment dictionary of structure also has the most emotion vocabulary cannot to cover；How remaining emotion word is returned
, to the problem of fine granularity emotional category, it is similar based on general knowledge on concept hierarchy to benchmark emotion word to be converted to analysis for class
Property, and assign it to the emotional category representated by similitude highest benchmark emotion word；Ultimate analysis categorization results, and it is complete
It is apt to defect that may be present；So far the enlarging of fine granularity sentiment dictionary is completed；
Step 2：Sentence structure relation judges
Judge whether there is conjunction in sentence, if, then it represents that the sentence is compound sentence, and the conjunction is obtained according to relation rule between sentence
The sentence structure relation of expression and the computation rule of sentence emotion value；If it is not, the sentence is simple sentence；
Step 3：Emotion value is evaluated
If compound sentence, then it is split as two subordinate sentences and is processed；If simple sentence, then its emotion value is directly calculated；Comment
The correlation that consider with descriptor is calculated by emotion value, the emotion word unrelated with descriptor can be calculated to emotion value brings dry
Disturb；And descriptor is mainly embodied by the subject of sentence and object, then need to only consider it is related to subject and object nominal and
Adjective emotion word；Intensity is modified according to sentence structure, dependence, emotion word degree of passing judgement on and associated adverbs, letter is calculated
The emotion value of simple sentence；
For the imperfect short sentence for commenting on frequent appearance, using word part of speech, the accuracy that dependence judges is improved；Work as emotion
When adjunctival before word is long, sentence structure, word part of speech and dependence are combined, specific method is as follows：First visit
Dependence is surveyed, descriptor is found out, and then finds out the adjunctival for depending on descriptor, the result according to sentence structure analysis is obtained
To the nominal phrase that descriptor and its adjunctival are constituted, the structure and adjunctival for then analyzing this nominal phrase are wrapped
The part of speech of the word for containing, draws correct dependence；
Step 4：Comment text fine granularity affection computation
Sentence emotion value is obtained with reference to relation rule between sentence pattern and sentence；The emotion value sum of all sentences is the entirety of comment text
Emotion value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410178056.6A CN103995803B (en) | 2014-04-25 | 2014-04-25 | A kind of fine granularity text sentiment analysis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410178056.6A CN103995803B (en) | 2014-04-25 | 2014-04-25 | A kind of fine granularity text sentiment analysis method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103995803A CN103995803A (en) | 2014-08-20 |
| CN103995803B true CN103995803B (en) | 2017-07-04 |
Family
ID=51309972
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410178056.6A Expired - Fee Related CN103995803B (en) | 2014-04-25 | 2014-04-25 | A kind of fine granularity text sentiment analysis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103995803B (en) |
Families Citing this family (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104881402B (en) * | 2015-06-02 | 2018-01-09 | 北京京东尚科信息技术有限公司 | The method and device of Chinese network topics comment text semantic tendency analysis |
| CN105335352A (en) * | 2015-11-30 | 2016-02-17 | 武汉大学 | Entity identification method based on Weibo emotion |
| CN108268439B (en) * | 2016-12-30 | 2021-09-07 | 北京国双科技有限公司 | Text emotion processing method and device |
| CN107291689A (en) * | 2017-05-31 | 2017-10-24 | 温州市鹿城区中津先进科技研究院 | A kind of analysis method based on the Chinese network comments sentence theme semantic tendency of big data |
| CN107515856A (en) * | 2017-08-30 | 2017-12-26 | 哈尔滨工业大学 | A kind of fine granularity Emotion element abstracting method represented based on local message |
| CN107704996B (en) * | 2017-09-12 | 2021-07-02 | 青岛大学 | Teacher evaluation system based on emotion analysis |
| CN107729374A (en) * | 2017-09-13 | 2018-02-23 | 厦门快商通科技股份有限公司 | A kind of extending method of sentiment dictionary and text emotion recognition methods |
| CN107807920A (en) * | 2017-11-17 | 2018-03-16 | 新华网股份有限公司 | Construction method, device and the server of mood dictionary based on big data |
| CN107862087B (en) * | 2017-12-01 | 2022-02-18 | 深圳爱数云科技有限公司 | Emotion analysis method and device based on big data and deep learning and storage medium |
| CN108460015A (en) * | 2018-02-08 | 2018-08-28 | 合肥工业大学 | Text emotion grouped data enhances analysis method |
| CN109271623A (en) * | 2018-08-16 | 2019-01-25 | 龙马智芯（珠海横琴）科技有限公司 | Text emotion denoising method and system |
| US11132511B2 (en) | 2019-02-05 | 2021-09-28 | International Business Machines Corporation | System for fine-grained affective states understanding and prediction |
| CN110083684B (en) * | 2019-04-24 | 2021-11-19 | 吉林大学 | Interpretable recommendation model for fine-grained emotion |
| CN110765769B (en) * | 2019-08-27 | 2023-05-02 | 电子科技大学 | Clause feature-based entity attribute dependency emotion analysis method |
| CN112784011B (en) * | 2021-01-04 | 2023-06-30 | 南威软件股份有限公司 | Emotion problem processing method, device and medium based on CNN and LSTM |
| CN112905736B (en) * | 2021-01-27 | 2023-09-19 | 郑州轻工业大学 | Quantum theory-based unsupervised text emotion analysis method |
| CN115794988A (en) * | 2022-09-13 | 2023-03-14 | 广东美云智数科技有限公司 | Method, apparatus, and computer storage medium for extracting viewpoint of text |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678278A (en) * | 2013-12-16 | 2014-03-26 | 中国科学院计算机网络信息中心 | Chinese text emotion recognition method |
-
2014
- 2014-04-25 CN CN201410178056.6A patent/CN103995803B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678278A (en) * | 2013-12-16 | 2014-03-26 | 中国科学院计算机网络信息中心 | Chinese text emotion recognition method |
Non-Patent Citations (3)
| Title |
|---|
| David B. Bracewell等.Determining the emotion of news articles.《Computational Intelligence》.2006,第918-923页. * |
| David B. Bracewell等.Semi-automatic creation of an emotion dictionary using wordnet and its evaluation.《Cybernetics and Intelligent Systems, 2008 IEEE Conference on》.2008,第1385-1389页. * |
| 基于情感本体的主题网络舆情倾向性分析;王兰成等;《信息与控制》;20130228;第43卷(第1期);第46-52页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103995803A (en) | 2014-08-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103995803B (en) | A kind of fine granularity text sentiment analysis method | |
| Ahmed et al. | SentiCR: A customized sentiment analysis tool for code review interactions | |
| Kunneman et al. | Signaling sarcasm: From hyperbole to hashtag | |
| Murphy | Linguistic representation and processing of copredication | |
| González-Ibánez et al. | Identifying sarcasm in twitter: a closer look | |
| Croft | Typology and universals | |
| Shivhare et al. | Emotion detection from text | |
| Chaski | Empirical evaluations of language-based author identification techniques | |
| US20160357731A1 (en) | Method for Automatically Detecting Meaning and Measuring the Univocality of Text | |
| Flekova et al. | Personality profiling of fictional characters using sense-level links between lexical resources | |
| Fonteyn | Categoriality in language change: the case of the English gerund | |
| Koptjevskaja-Tamm et al. | Typological approaches to lexical semantics | |
| Dawdy-Hesterberg et al. | Learnability and generalisation of Arabic broken plural nouns | |
| Gelbukh et al. | Semantic analysis of verbal collocations with lexical functions | |
| Akita | Toward a frame-semantic definition of sound-symbolic words: A collocational analysis of Japanese mimetics | |
| Johnson et al. | The FrameNet project: Tools for lexicon building | |
| Wanner | Towards automatic fine-grained semantic classification of verb-noun collocations | |
| Gärdenfors | A semantic theory of word classes | |
| Chen | Computational generation of Chinese noun phrases | |
| Neviarouskaya et al. | Attitude sensing in text based on a compositional linguistic approach | |
| O'Gorman | Bringing together computational and linguistic models of implicit role interpretation | |
| Odijka et al. | MWE-finder: Querying for multiword expressions in large Dutch text corpora | |
| Gawryjolek | Automated annotation and visualization of rhetorical figures | |
| Kotowski | Adjectival modification and order restrictions: The influence of temporariness on prenominal word order | |
| Neto | Subject expression in Brazilian Portuguese: Construction and frequency effects |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170704 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |