CN109446423A - A kind of Judgment by emotion system and method for news and text - Google Patents
A kind of Judgment by emotion system and method for news and text Download PDFInfo
- Publication number
- CN109446423A CN109446423A CN201811257151.XA CN201811257151A CN109446423A CN 109446423 A CN109446423 A CN 109446423A CN 201811257151 A CN201811257151 A CN 201811257151A CN 109446423 A CN109446423 A CN 109446423A
- Authority
- CN
- China
- Prior art keywords
- emotion
- judgment
- word
- text
- negative
- 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
- 230000008451 emotion Effects 0.000 title claims abstract description 136
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000013459 approach Methods 0.000 claims abstract description 11
- 238000004364 calculation method Methods 0.000 claims abstract description 10
- 230000005484 gravity Effects 0.000 claims abstract description 5
- 230000008569 process Effects 0.000 claims abstract description 5
- 238000002474 experimental method Methods 0.000 claims description 14
- 230000011218 segmentation Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 5
- 230000002996 emotional effect Effects 0.000 claims description 4
- 230000007935 neutral effect Effects 0.000 claims description 4
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 230000001568 sexual effect Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 2
- 230000014509 gene expression Effects 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 239000004744 fabric Substances 0.000 description 2
- 241000287828 Gallus gallus Species 0.000 description 1
- 206010028916 Neologism Diseases 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000036651 mood Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Abstract
The invention discloses a kind of news and the Judgment by emotion methods of text, comprising the following steps: pre-processes to the newsletter archive that network crawls, removes crawler web page tag, and remove the stop words in newsletter archive；Using deep approach of learning to the preliminary Judgment by emotion of newsletter archive；Using SVM method to the secondary Judgment by emotion of newsletter archive；Expression front or the negative emotion word summarized in newsletter archive are collected, and is matched with just negative affection data library, the specific gravity of front or negative emotion word in newsletter archive is calculated, carries out Judgment by emotion three times；By preliminary Judgment by emotion result, secondary Judgment by emotion result and three times Judgment by emotion result carry out weight calculation, the emotion of comprehensive descision newsletter archive；The Judgment by emotion method of three of the above method is carried out weight calculation by this programme, to carry out comprehensive Judgment by emotion to news and text, improves the accuracy rate to the Judgment by emotion of news and text.
Description
Technical field
The present invention relates to artificial intelligence and natural language processing technique field, the emotion of specially a kind of news and text
Judge system and method.
Background technique
With network technology, the fast development of the network media, news, User Perspective, user's evaluation, society in network
The massive informations such as public sentiment have sharply increased.In these information much can with the information of subjective emotion, have positive emotion,
There is negative emotion.
It is all the emotion for manually judging news and text in the past.This needs a large amount of manpower to go to judge network
The emotion of news and text.Manually come if judging emotion, as today of network information explosion, artificial judgment
The Sentiment orientation of news tonality and text is extremely backward.
Therefore, how from the information of magnanimity, using it is unartificial, high speed, accurately judge the subjectivities of these information
Emotion becomes urgent, the important technical task of government, business unit, public institution etc..
Wherein in the invention that this number of patent application is 201710463295.X, discloses a kind of Internet news and obtain simultaneously
The system for predicting text emotion.This system is using the newsletter archive that network crawls as training set, using SVM method, to network
The emotion of news is labeled and is trained, and later, carries out the judgement of Internet news emotion.
In this invention above, there is some problems, that is, a kind of Judgment by emotion mode are only used, in Judgment by emotion
Accuracy on be tool it is limited, be difficult have better performance in accuracy.
Summary of the invention
In order to overcome the shortcomings of prior art, the present invention provide a kind of news and text Judgment by emotion system and
Method can effectively solve the problem of background technique proposes.
The technical solution adopted by the present invention to solve the technical problems is:
A kind of Judgment by emotion method of news and text, which comprises the following steps:
Step 100 pre-processes the newsletter archive that network crawls, and the title of newsletter archive and content is merged, removal is climbed
Stop words in worm web page tag and text；
Step 200, using deep approach of learning to the preliminary Judgment by emotion of newsletter archive；
Step 300, using SVM method to the secondary Judgment by emotion of newsletter archive；
Step 400 collects to summarize and indicates front or negative emotion word in newsletter archive, and by emotion word and feelings
Sense database is matched, and the specific gravity of front or negative emotion word in newsletter archive is calculated, and carries out Judgment by emotion three times；
Step 500, by preliminary Judgment by emotion result, secondary Judgment by emotion result and three times Judgment by emotion result carry out weight
It calculates, the emotion of comprehensive descision newsletter archive.
Further, in step 200, the specific steps of preliminary Judgment by emotion are carried out using deep approach of learning are as follows:
Step 201, will remove stop words newsletter archive carry out word segmentation processing, obtain unitary word, binary word, ternary word and
The sequence of terms of polynary word；
Step 202 compares the participle element of sequence of terms in step 201 with affection data library respectively, again will
Sequence of terms is rearranged according to negative word and degree word, generates extension word sequence；
All phrase quantity of Sentiment orientation are indicated in step 203, statistics extension word sequence in lemma element, i.e., statistics is new
The word total amount of front tendency is indicated in news or text, indicates the word total amount of negative tendency and the word total amount of middle sexual orientation；
The Sentiment orientation word extended in word sequence is input to Judgment by emotion model and is trained by step 204, obtains judgement
As a result.
Further, in the step 202, element and the comparison of affection data library will be being segmented, extension word sequence is generated
The step of specifically:
Firstly, news or text header classification are corresponded to corresponding subject fields；
Then, the affection data library in corresponding subject fields is selected；
It, will be with front or negative finally, respectively match the participle element in sequence of terms with affection data library
The adjacent degree word of Sentiment orientation word and negative word are merged together, and will be marked in affection data library with similar in participle element
Quasi- word replaces the lemma element in sequence of terms, reintegrates into extension word sequence.
Further, in step 300, secondary Judgment by emotion is carried out using SVM method method particularly includes:
Step 301 extracts Sentiment orientation Feature Words in newsletter archive in sequence, and by the emotion in newsletter archive
Tendency Feature Words are divided into front, neutrality, negative three classes；
Step 302, the IG algorithm using Sentiment orientation Feature Words, multiple Feature Words are integrated in feature lexicon；
Step 303 carries out tf/idf calculating to the Feature Words in feature lexicon, and SVM is added in the tf/idf value of Feature Words
It is trained in model, obtains front, neutrality, negative three classes Sentiment orientation value.
Further, the calculation formula of the IG algorithm specifically:
IG=∑ P (i) ln (P (i)/Q (i))；
Wherein IG is information gain, and P (i) is the probability distribution of ith feature word, and Q (i) is the probability point of emotional semantic classification
Cloth.
Further, the affection data library includes positive emotion dictionary, negative emotion dictionary, degree adverb dictionary and no
Determine dictionary.
Further, in the step 500, the formula of weight calculation is carried out to Judgment by emotion result specifically:
E (X)=∑ (p (x) * e (x))；
Wherein E (X) is that the Sentiment orientation statistical mathematics of above-mentioned three kinds of algorithms it is expected, p (x) is the weight of certain above-mentioned algorithm, e
It (x) is certain above-mentioned algorithm Sentiment orientation value.
Further, the weight circular of three kinds of Judgment by emotion algorithms:
The first, the experiment text of a variety of different fields and different themes is obtained；
The second, the Sentiment orientation of text manually is accurately identified to experiment text as a result, being front tendency to the theme, anti-
Face tendency or neutral tendency；
Third, successively according to three kinds of deep approach of learning, SVM method and affection data library method Judgment by emotion methods to above-mentioned reality
Text is tested to carry out Judgment by emotion, and records three kinds of Judgment by emotion methods respectively to the Judgment by emotion result of experiment text；
4th, successively by three kinds of Judgment by emotion methods to the judging result of all experiment texts and the emotion knot of artificial judgment
Fruit compares, and determines the accuracy rate of three kinds of Judgment by emotion methods respectively, and the accuracy rate is three kinds of Judgment by emotion methods
Weight.
Compared with prior art, the beneficial effects of the present invention are: the present invention is by deep approach of learning, SVM method, affection data
Library method is combined, and the Judgment by emotion method of three of the above method is carried out weight calculation, is carried out to news and text
Comprehensive Judgment by emotion achieves extraordinary effect, accuracy rate in practical applications, to the Judgment by emotion of news and text
It is very high.
Detailed description of the invention
Fig. 1 is judgement flow diagram of the invention；
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
As shown in Figure 1, the present invention provides a kind of news and the Judgment by emotion methods of text, comprising the following steps:
Step 100 pre-processes the newsletter archive that network crawls, and the title of newsletter archive and content is merged, removal is climbed
Stop words in worm web page tag and text is removed mixed and disorderly tag processes to the content got off is crawled, and mixed and disorderly label includes
Such as " ◆ ", " ▲ ", " ↓ " etc. symbol.And the html label of webpage is got rid of, stop words is specially conjunction and adverbial word,
Such as "and", " obtaining ", " ", " between " etc..
Step 200, using deep approach of learning to the preliminary Judgment by emotion of newsletter archive.
In this step, the specific steps of preliminary Judgment by emotion are carried out using deep approach of learning are as follows:
Step 201, will remove stop words newsletter archive carry out word segmentation processing, obtain unitary word, binary word, ternary word and
The sequence of terms of polynary word, if word segmentation processing here is to be divided into every words of newsletter archive according to common word for unit
Dry lemma element, to form the sequence of terms of several groups binary word, ternary word and polynary word.
Step 202, the participle element of sequence of terms in above-mentioned steps is compared with affection data library respectively, again will
Sequence of terms is rearranged according to negative word and degree word, generates extension word sequence, due to the fast development of present a networked society,
There are many neologisms to occur to express certain mood, to avoid omitting Sentiment orientation word, first by the lemma element and emotion of sequence of terms
Database compares, then without modification for the lemma element in existing affection data library, for transformable lemma element, then
It is plain that the lemma in term vector sequence is replaced with the close word that affection data library is summarized, such as can convert " useless " for " chicken ribs ",
And since a front tendency word of Chinese adds negative word, it is possible to translate into reverse side tendency word, and degree adverb
Addition, Sentiment orientation value can be improved, therefore in this step, need to word element carry out again be integrated into a new member
Element generates extension word sequence.
It is specific in the step of segmenting element and affection data library compares, and generation extends word sequence in the step 202
Are as follows:
Firstly, news or text header classification are corresponded to corresponding subject fields；
Then, the affection data library in corresponding subject fields is selected；
It, will be with front or negative finally, respectively match the participle element in sequence of terms with affection data library
The adjacent degree word of Sentiment orientation word and negative word are merged together, and will be marked in affection data library with similar in participle element
Quasi- word replaces the lemma element in sequence of terms, reintegrates into extension word sequence.
Some vocabulary are in different fields, and the Sentiment orientation difference indicated is very big, so first to the master of newsletter archive
Topic is clear, carries out front and back sides classification convenient for the participle element to Sentiment orientation, carries out front or negative emotion to participle element
Classification, finally again to term vector sequence according to front or negative emotion be divided into positive emotion sequence vector and negative emotion to
Measure sequence.
Step 203, all phrase quantity of Sentiment orientation are indicated in statistics extension word sequence in lemma element, i.e., statistics is new
The word total amount of front tendency is indicated in news or text, indicates the word total amount of negative tendency and the word total amount of middle sexual orientation, is being counted
Before Sentiment orientation word, first vector conversion is carried out to the lemma element in expansion word sequence vector using word embedded technology, generated
Term vector sequence word, word embedded technology are exactly to assign a vector to each element in expansion word sequence vector, and vector represents sky
Between in point, the close word of meaning, vector is also close to can be converted into the operation for vector for the operation of word in this way
, be otherwise known as tensor in deep learning.
It further illustrates, word and reverse side Sentiment orientation word is inclined to for positive emotion, to the vector after its assignment
Difference is very big, indicates that word is advantageous in that with vector: first, it can overcome the problems, such as that text length is uneven, because if each
Word has had corresponding term vector, then the text for being N for length, if vector representated by corresponding N number of word is chosen, and
It is come together by the sequencing of word in text, wherein the dimension of each term vector is the same；Second, word itself can not shape
At feature, but tensor is exactly abstract quantization, it is calculated by being abstracted layer by layer for multilayer neural network；Third,
Text is composed of words, and the feature of text can be combined by the tensor of word, and the tensor of text has contained the group between multiple words
Meaning is closed, this is considered the Feature Engineering of text, and then provides basis for machine learning text classification.
Step 204, the Sentiment orientation word extended in word sequence is input to Judgment by emotion model to be trained, obtains judgement
As a result, Judgment by emotion model carries out Judgment by emotion, the depth convolutional neural networks benefit using depth convolutional neural networks principle
Emotion word activation is carried out with ReLU line rectification function.
Affection data library in present embodiment include positive emotion dictionary, negative emotion dictionary, degree adverb dictionary and
It negate dictionary.
Step 300, using SVM method to the secondary Judgment by emotion of newsletter archive；
Secondary Judgment by emotion is carried out using SVM method method particularly includes:
(1) the Sentiment orientation Feature Words in newsletter archive are extracted in sequence, and by the Sentiment orientation in newsletter archive
Feature Words are divided into front, neutrality, negative three classes, before extracting newsletter archive, it is first determined theme of news, it can be according to unitary
Word, binary word, ternary word or polynary word carry out word segmentation processing, then will have Sentiment orientation Feature Words to carry out emotional semantic classification.
(2) the IG algorithm for using Sentiment orientation Feature Words, multiple Feature Words is integrated in feature lexicon, the IG algorithm
Calculation formula specifically:
IG=∑ P (i) ln (P (i)/Q (i))；
Wherein IG is information gain, and P (i) is the probability distribution of ith feature word, and Q (i) is the probability point of emotional semantic classification
Cloth.
It should be added that feature lexicon is a kind of variodenser model, and any type object can be stored, in this reality
It applies in mode, feature lexicon may include multiple elements, and wherein each element in feature lexicon includes Feature Words variable and right
Answer the information gain of Feature Words.
(3) tf/idf calculating is carried out to the Feature Words in feature lexicon, the tf/idf value of Feature Words is added in SVM model
It being trained, obtains front, neutrality, negative three classes Sentiment orientation value, tf/idf is used to calculate the weight calculation of Feature Words, wherein
Tf represents the word frequency of Feature Words, and the ability for describing document content to calculate the specific word, idf is inverse document frequency, based on
The ability that the specific word distinguishes document is calculated, tf/idf value is added in the information gain of Feature Words by nonlinear transformation, from
And obtain front, neutrality, negative three classes Sentiment orientation value.
Step 400 collects that expressions summarized in newsletter archive be positive or negative emotion word, and with positive negative emotion
Database is matched, and the specific gravity of front or negative emotion word in newsletter archive is calculated, and carries out Judgment by emotion three times.
In this step, it is necessary first to segment to obtain segmentation sequence to news or text, and to segmentation sequence according to
Adjacent degree word and negative word integration generate extension word sequence, and mark to the Sentiment orientation word in extension word sequence
Note classification just corresponds to news or text is made then when there is the word of corresponding negative emotion in news and text
It is otherwise non-negative press and text negatively to judge, finally the progress of multiple segmentation sequences is negatively judged and non-negative
The statistical stacking of judgement, relatively more negative judgement and the non-specific gravity negatively judged judge that the emotion of the newsletter archive tends to.
Step 500, by preliminary Judgment by emotion result, secondary Judgment by emotion result and three times Judgment by emotion result carry out weight
It calculates, the emotion of comprehensive descision newsletter archive.
In this step, the formula of weight calculation is carried out to Judgment by emotion result specifically:
E (X)=∑ (p (x) * e (x))；
Wherein E (X) is that the Sentiment orientation statistical mathematics of above-mentioned three kinds of algorithms it is expected, p (x) is the weight of certain above-mentioned algorithm, e
It (x) is certain above-mentioned algorithm Sentiment orientation value.
Three kinds of Judgment by emotion methods are weighted and averaged using the method for weighting, can be reduced shared by the low judgement algorithm of accuracy rate
Ratio, to improve the accuracy rate of Judgment by emotion.
It should be added that the weight circular of three kinds of Judgment by emotion algorithms:
The first, the experiment text of a variety of different fields and different themes is obtained；
The second, the Sentiment orientation of text manually is accurately identified to experiment text as a result, being front tendency to the theme, anti-
Face tendency or neutral tendency；
Third, successively according to three kinds of deep approach of learning, SVM method and affection data library method Judgment by emotion methods to above-mentioned reality
Text is tested to carry out Judgment by emotion, and records three kinds of Judgment by emotion methods respectively to the Judgment by emotion result of experiment text；
4th, successively by three kinds of Judgment by emotion methods to the judging result of all experiment texts and the emotion knot of artificial judgment
Fruit compares, and determines the accuracy rate of three kinds of Judgment by emotion methods respectively, and the accuracy rate is three kinds of Judgment by emotion methods
Weight.
That is, in present embodiment three kinds of Judgment by emotion methods weight, be the judgment experiment by a large amount of texts
It obtains, such mode also improves every kind of weight accuracy for judging emotion method as far as possible.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included within the present invention.Any reference signs in the claims should not be construed as limiting the involved claims.
Claims (8)
1. a kind of Judgment by emotion method of news and text, which comprises the following steps:
Step 100 pre-processes the newsletter archive that network crawls, and the title of newsletter archive and content are merged, crawler net is removed
Stop words in page label and text；
Step 200, using deep approach of learning to the preliminary Judgment by emotion of newsletter archive；
Step 300, using SVM method to the secondary Judgment by emotion of newsletter archive；
Step 400 collects to summarize and indicates front or negative emotion word in newsletter archive, and by emotion word and emotion number
It is matched according to library, calculates the specific gravity of front or negative emotion word in newsletter archive, carry out Judgment by emotion three times；
Step 500, by preliminary Judgment by emotion result, secondary Judgment by emotion result and three times Judgment by emotion result carry out weight meter
It calculates, the emotion of comprehensive descision newsletter archive.
2. the Judgment by emotion method of a kind of news according to claim 1 and text, which is characterized in that in step 200
In, the specific steps of preliminary Judgment by emotion are carried out using deep approach of learning are as follows:
Step 201, the newsletter archive that will remove stop words carry out word segmentation processing, obtain unitary word, binary word, ternary word and polynary
The sequence of terms of word；
Step 202 compares the participle element of sequence of terms in step 201 with affection data library respectively, again by word
Sequence is rearranged according to negative word and degree word, generates extension word sequence；
Indicate all phrase quantity of Sentiment orientation in step 203, statistics extension word sequence in lemma element, i.e., statistics in news or
The word total amount of front tendency is indicated in text, indicates the word total amount of negative tendency and the word total amount of middle sexual orientation；
The Sentiment orientation word extended in word sequence is input to Judgment by emotion model and is trained by step 204, obtains judgement knot
Fruit.
3. the Judgment by emotion method of a kind of news according to claim 2 and text, which is characterized in that in the step
In 202, in the step of segmenting element and affection data library compares, and generation extends word sequence specifically:
Firstly, news or text header classification are corresponded to corresponding subject fields；
Then, the affection data library in corresponding subject fields is selected；
Finally, respectively match the participle element in sequence of terms with affection data library, it will be with front or negative emotion
The adjacent degree word and negative word of tendency word is merged together, and by affection data library with standard words similar in participle element
Instead of the lemma element in sequence of terms, extension word sequence is reintegrated into.
4. the Judgment by emotion method of a kind of news according to claim 1 and text, which is characterized in that in step 300
In, secondary Judgment by emotion is carried out using SVM method method particularly includes:
Step 301 extracts Sentiment orientation Feature Words in newsletter archive in sequence, and by the Sentiment orientation in newsletter archive
Feature Words are divided into front, neutrality, negative three classes；
Step 302, the IG algorithm using Sentiment orientation Feature Words, multiple Feature Words are integrated in feature lexicon；
Step 303 carries out tf/idf calculating to the Feature Words in feature lexicon, and SVM model is added in the tf/idf value of Feature Words
In be trained, obtain front, neutral, negative three classes Sentiment orientation value.
5. the Judgment by emotion method of a kind of news according to claim 4 and text, which is characterized in that the IG algorithm
Calculation formula specifically:
IG=∑ P (i) ln (P (i)/Q (i))；
Wherein IG is information gain, and P (i) is the probability distribution of ith feature word, and Q (i) is the probability distribution of emotional semantic classification.
6. the Judgment by emotion method of a kind of news according to claim 1 and text, which is characterized in that the emotion number
It include positive emotion dictionary, negative emotion dictionary, degree adverb dictionary and negative dictionary according to library.
7. the Judgment by emotion method of a kind of news according to claim 1 and text, which is characterized in that in the step
In 500, the formula of weight calculation is carried out to Judgment by emotion result specifically:
E (X)=∑ (p (x) * e (x))；
Wherein E (X) is that the Sentiment orientation statistical mathematics of above-mentioned three kinds of algorithms it is expected, p (x) is the weight of certain above-mentioned algorithm, e (x)
For certain above-mentioned algorithm Sentiment orientation value.
8. the Judgment by emotion method of a kind of news according to claim 7 and text, which is characterized in that three kinds of emotions are sentenced
The weight circular of disconnected algorithm:
The first, the experiment text of a variety of different fields and different themes is obtained；
The second, the Sentiment orientation of text manually is accurately identified to experiment text as a result, being that front is inclined to, reverse side inclines to the theme
To or neutral tendency；
Third, successively according to three kinds of deep approach of learning, SVM method and affection data library method Judgment by emotion methods to above-mentioned experiment text
It is original to carry out Judgment by emotion, and three kinds of Judgment by emotion methods are recorded respectively to the Judgment by emotion result of experiment text；
4th, successively by three kinds of Judgment by emotion methods to it is all experiment texts judging results and artificial judgment emotion result into
Row comparison, determines the accuracy rate of three kinds of Judgment by emotion methods, and the accuracy rate is the power of three kinds of Judgment by emotion methods respectively
Weight.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811257151.XA CN109446423B (en) | 2018-10-26 | 2018-10-26 | System and method for judging sentiment of news and texts |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811257151.XA CN109446423B (en) | 2018-10-26 | 2018-10-26 | System and method for judging sentiment of news and texts |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109446423A true CN109446423A (en) | 2019-03-08 |
| CN109446423B CN109446423B (en) | 2022-07-01 |
Family
ID=65547648
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811257151.XA Expired - Fee Related CN109446423B (en) | 2018-10-26 | 2018-10-26 | System and method for judging sentiment of news and texts |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109446423B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111666412A (en) * | 2020-06-02 | 2020-09-15 | 国家计算机网络与信息安全管理中心 | Fraud log text analysis method and system based on SVM text analysis |
| CN111813937A (en) * | 2020-07-07 | 2020-10-23 | 新华智云科技有限公司 | Positive energy news classification method based on positive energy dictionary |
| CN112231470A (en) * | 2019-06-28 | 2021-01-15 | 上海智臻智能网络科技股份有限公司 | Topic mining method and device, storage medium and terminal |
| CN112364170A (en) * | 2021-01-13 | 2021-02-12 | 北京智慧星光信息技术有限公司 | Data emotion analysis method and device, electronic equipment and medium |
| CN117216419A (en) * | 2023-11-08 | 2023-12-12 | 江西为易科技有限公司 | Data analysis method based on AI technology |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107291696A (en) * | 2017-06-28 | 2017-10-24 | 达而观信息科技（上海）有限公司 | A kind of comment word sentiment analysis method and system based on deep learning |
| CN108376133A (en) * | 2018-03-21 | 2018-08-07 | 北京理工大学 | The short text sensibility classification method expanded based on emotion word |
| CN108388608A (en) * | 2018-02-06 | 2018-08-10 | 金蝶软件（中国）有限公司 | Emotion feedback method, device, computer equipment and storage medium based on text perception |
-
2018
- 2018-10-26 CN CN201811257151.XA patent/CN109446423B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107291696A (en) * | 2017-06-28 | 2017-10-24 | 达而观信息科技（上海）有限公司 | A kind of comment word sentiment analysis method and system based on deep learning |
| CN108388608A (en) * | 2018-02-06 | 2018-08-10 | 金蝶软件（中国）有限公司 | Emotion feedback method, device, computer equipment and storage medium based on text perception |
| CN108376133A (en) * | 2018-03-21 | 2018-08-07 | 北京理工大学 | The short text sensibility classification method expanded based on emotion word |
Non-Patent Citations (1)
| Title |
|---|
| 夏火松 等: "中文情感分类挖掘预处理关键技术比较研究", 《情报杂志》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112231470A (en) * | 2019-06-28 | 2021-01-15 | 上海智臻智能网络科技股份有限公司 | Topic mining method and device, storage medium and terminal |
| CN111666412A (en) * | 2020-06-02 | 2020-09-15 | 国家计算机网络与信息安全管理中心 | Fraud log text analysis method and system based on SVM text analysis |
| CN111813937A (en) * | 2020-07-07 | 2020-10-23 | 新华智云科技有限公司 | Positive energy news classification method based on positive energy dictionary |
| CN112364170A (en) * | 2021-01-13 | 2021-02-12 | 北京智慧星光信息技术有限公司 | Data emotion analysis method and device, electronic equipment and medium |
| CN117216419A (en) * | 2023-11-08 | 2023-12-12 | 江西为易科技有限公司 | Data analysis method based on AI technology |
| CN117216419B (en) * | 2023-11-08 | 2024-02-09 | 江西为易科技有限公司 | Data analysis method based on AI technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109446423B (en) | 2022-07-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109446423A (en) | A kind of Judgment by emotion system and method for news and text | |
| CN104268197B (en) | A kind of industry comment data fine granularity sentiment analysis method | |
| CN107085581B (en) | Short text classification method and device | |
| CN110032632A (en) | Intelligent customer service answering method, device and storage medium based on text similarity | |
| CN110321563B (en) | Text emotion analysis method based on hybrid supervision model | |
| CN110134946B (en) | Machine reading understanding method for complex data | |
| CN108388608B (en) | Emotion feedback method and device based on text perception, computer equipment and storage medium | |
| CN109885670A (en) | A kind of interaction attention coding sentiment analysis method towards topic text | |
| CN108829662A (en) | A kind of conversation activity recognition methods and system based on condition random field structuring attention network | |
| KR20190063978A (en) | Automatic classification method of unstructured data | |
| CN107688870B (en) | Text stream input-based hierarchical factor visualization analysis method and device for deep neural network | |
| CN108874896B (en) | Humor identification method based on neural network and humor characteristics | |
| CN112417127B (en) | Dialogue model training and dialogue generation methods, devices, equipment and media | |
| CN110750648A (en) | Text emotion classification method based on deep learning and feature fusion | |
| CN113343690B (en) | Text readability automatic evaluation method and device | |
| CN110134934A (en) | Text emotion analysis method and device | |
| CN110728144B (en) | Extraction type document automatic summarization method based on context semantic perception | |
| CN110569355B (en) | Viewpoint target extraction and target emotion classification combined method and system based on word blocks | |
| CN116028604A (en) | Answer selection method and system based on knowledge enhancement graph convolution network | |
| CN113220964B (en) | Viewpoint mining method based on short text in network message field | |
| CN113934835A (en) | Retrieval type reply dialogue method and system combining keywords and semantic understanding representation | |
| CN113486143A (en) | User portrait generation method based on multi-level text representation and model fusion | |
| CN111581379B (en) | Automatic composition scoring calculation method based on composition question-deducting degree | |
| CN110377753B (en) | Relation extraction method and device based on relation trigger word and GRU model | |
| CN116401368A (en) | Intention recognition method and system based on topic event analysis |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220701 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |