CN102194001A - Internet public opinion crisis early-warning method - Google Patents
Internet public opinion crisis early-warning method Download PDFInfo
- Publication number
- CN102194001A CN102194001A CN2011101275099A CN201110127509A CN102194001A CN 102194001 A CN102194001 A CN 102194001A CN 2011101275099 A CN2011101275099 A CN 2011101275099A CN 201110127509 A CN201110127509 A CN 201110127509A CN 102194001 A CN102194001 A CN 102194001A
- Authority
- CN
- China
- Prior art keywords
- opinion
- topic
- talked
- much
- network public
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention relates to an Internet public opinion crisis early-warning method. The method comprises the following steps: getting a non-thematic word library and trend mining rules; extracting Internet public opinion hotspot words; detecting Internet public opinion hotspot topics; calculating similarity of the hotspot topics and automatically tracking; performing trend mining on the hotspot topics; and performing automatic early-warning on an Internet public opinion crisis. The method is divided into two parts, namely an off-line technology and an on-line technology. The off-line technology adopts a statistical analysis and machine learning method for getting the non-thematic word library and a trend mining rule library. The on-line part comprises the following steps: firstly adopting a two-stage filtration method for fast extracting the public opinion hotspot words; then adopting a word clustering method based on co-word analysis for getting the public opinion hotspot topics; further calculating the similarity of the hotspot topics in a continuous time period, quantifying changes of the hotspot topics as time goes on, and realizing the automatic tracking of the hotspot topics; and finally adopting a fuzzy reasoning technology to mine Internet public opinion trend knowledge and realizing the automatic early-warning of the public opinion crisis. By adopting the method, a public opinion worker can timely make a treatment decision of the public opinion crisis.
Description
Technical field
The invention belongs to the Intelligent Information Processing field, relate to a kind of real-time network public-opinion crisis alert method.
Background technology
Current internet has goed deep into huge numbers of families, it is wide to have use, propagate wide, propagate rapidly, be not subjected to features such as space-time restriction, can informational influence power be amplified at double as magnifier, especially some negative focus incidents, sensitive subjects, an appearance may be widely known between the sunset on the internet, cause the trust disappearance to relative subject, brand is broken one's promise, image is impaired, adverse effects such as common people's opposition, therefore need develop network public-opinion technology for automatically treating efficiently, help the public sentiment worker in time to take measures effectively to dredge netizen's mood, trend guides public opinion.We think; for realizing above target; tupe to network public-opinion should change " crisis alert in advance " into by " crisis processing afterwards ", and has only realization could really realize the automatic early-warning of public sentiment crisis to the correct judgement of online much-talked-about topic development trend.
Aspect the public sentiment crisis alert; the prior art means are also few; the common practices of existing public sentiment disposal system is that the seniority among brothers and sisters and the situation of change of much-talked-about topic or sensitive subjects are submitted to the public sentiment worker in modes such as daily paper, weekly or wall bulletins; judge by manually doing early warning then, think internet information monitoring analysis system etc. as Goonie network public-opinion monitoring analysis system, TRS internet public feelings management system, upright intelligence.At the automatic early-warning technical elements, Li Bicheng etc. are in paper " based on the network public-opinion method for early warning of intuitionistic fuzzy reasoning " (computer utility research, 2010 the 9th phases) propose to use for reference situation of battlefield in and analyze thought, chosen suitable computer implemented seven network public-opinion situation analytical models advanced warning grade is judged.But a plurality of parameters that this method needs are difficult for obtaining, and need manual intervention, can't realize by full automation.Ding Juling etc. are at paper " the flexible mining model of the viewpoint of a kind of network-oriented public sentiment crisis alert " (information magazine; 2009 the 10th phases) tentatively inquired into the flexible mining model of structure viewpoint in; the viewpoint evolutionary process is predicted the viewpoint evolution trend in the simulation crisis, to realize the feasibility of public sentiment crisis alert.Zhang Jue proposes to adopt the forecast model of economic field in Master's thesis " research of network public-opinion forecast model and platform " (Beijing Jiaotong University, 2009), as ARIMA model and BP neural network model, come network public-opinion is predicted.Above several method also is in the exploratory stage, also is not applied in the network public-opinion disposal system of reality.
Summary of the invention
The objective of the invention is the deficiency at existing public sentiment treatment technology, the network public-opinion development trend that proposes a kind of public sentiment content Network Based and network public-opinion evolution rule is excavated and the crisis automatic early warning method.
Characteristics such as that network public-opinion has is sudden, destructive, urgency, in the face of the network public sentiment information of magnanimity, how in the shortest time, to make analysis, study and judge with decision-making be vital.Existing focus speech extracting method based on statistics needs a large amount of in line computation, and too high based on the much-talked-about topic discover method computation complexity of text classification/cluster, directly causes system real time to be difficult to guarantee.
Basic thought of the present invention is to reduce on-line calculation as far as possible, and adopts new much-talked-about topic detection, tracking and trend method for digging, realizes the real-time early warning of network public-opinion crisis.
Technical scheme of the present invention is divided into two of off-line part and online parts.The off-line part is a support with a large scale network corpus (the 2000000 pieces of Sina's web page news in 6 years of time span) and a public sentiment corpus (gathering arrangement from ends of the earth forum), adopt statistical study and machine learning method, obtain non-theme dictionary and trend mining rule storehouse respectively, for network focus speech extracts and the excavation of much-talked-about topic trend provides support.The public feelings information that online part at first will be gathered in real time carries out pre-service, extracts plain text information, under non-theme dictionary is supported, adopts two-stage filtration method rapid extraction to go out the focus speech then.After the focus speech extracts, adopt based on the term clustering method of speech analysis altogether and obtain much-talked-about topic, then calculate the similarity between the continuous time much-talked-about topic, quantize that much-talked-about topic is passed in time and the situation of change that produces, realize much-talked-about topic from motion tracking.Under supporting in trend mining rule storehouse at last; introduce Fuzzy Inference much-talked-about topic is carried out the trend dredge operation; obtain the tendency information of reflection public sentiment future thrust, if satisfying early-warning conditions then reports to the police automatically, auxiliary public sentiment worker makes the public sentiment crisis and disposes decision-making.
Technical scheme of the present invention can solve existing focus speech acquisition methods not high problem of efficient in the face of the mass network text time based on online statistics, can solve the existing too high problem of much-talked-about topic discover method computation complexity based on text classification/cluster, can realize simultaneously much-talked-about topic from motion tracking and automatic early-warning, to alleviate public sentiment worker's workload.
The method that the present invention proposes is proved to be effectively reliable through overtesting, finishes detection and trend excavation to much-talked-about topic in can 3 minutes after the public feelings information collection is downloaded, and provides the public sentiment information warning.
Description of drawings
Fig. 1 is a FB(flow block) of the present invention.
Fig. 2 extracts process flow diagram for the focus speech.
Fig. 3 finds process flow diagram for much-talked-about topic.
Fig. 4 changes language value subordinate function figure for the much-talked-about topic temperature.
Fig. 5 is degree of confidence language value subordinate function figure.
Fig. 6 follows the tracks of with trend for much-talked-about topic and excavates process flow diagram.
Fig. 7 a is the much-talked-about topic trend map.
Fig. 7 b is public sentiment crisis alert figure.
Embodiment
For better understanding technical scheme of the present invention, describe embodiments of the present invention in detail below in conjunction with accompanying drawing, the total FB(flow block) of the present invention is seen Fig. 1.
(1) non-theme dictionary and the trend mining rule storehouse based on statistical study and machine learning obtains.
(2) based on the network public-opinion focus speech rapid extraction of two-stage filtration.
(3) based on the network public-opinion much-talked-about topic detection of speech analysis and term clustering altogether.
(4) much-talked-about topic of calculating based on continuous time much-talked-about topic similarity is from motion tracking.
(5) based on the network public-opinion trend knowledge excavation of fuzzy reasoning.
(6) based on the network public-opinion crisis automatic early-warning of multimedia show technology.
1. to obtain be the basic steps that public feelings information is handled to public sentiment focus speech, and existing public sentiment focus speech acquisition methods mainly is the method for identification, and the present invention adopts two-stage filtration method (seeing accompanying drawing 2), is divided into following step:
(1) the public sentiment text is carried out obtaining public sentiment text word sequence after pre-service, participle and the part-of-speech tagging;
(2) stop words filters, and the stop words that is about in the public sentiment text word sequence is deleted.Inactive vocabulary adopts following method to make up: large scale network corpus of off-line analysis (the 2000000 pieces of Sina's web page news in 6 years of time span), carry out participle and part-of-speech tagging, and extract coordinating conjunction, conjunction, interjection, preposition, personal pronoun, place interrogative pronoun, time interrogative pronoun, the predicate interrogative pronoun, interrogative pronoun, place demonstrative pronoun, the time demonstrative pronoun, predicate demonstrative pronoun, demonstrative pronoun, pronoun, auxiliary word, modal particle obtains a vocabulary.Get the inactive vocabulary (containing 767 words and symbol) of Harbin Institute of Technology information retrieval experiment chamber and the inactive vocabulary (containing 1176 words and symbol) of union conduct of above-mentioned vocabulary then.
(3) non-descriptor is filtered, and the non-descriptor that is about in the public sentiment text word sequence is deleted.Non-theme dictionary adopts following method to make up: a large scale network corpus (the 2000000 pieces of Sina's web page news in 6 years of time span) monthly is divided into 72 subclass, extract the notional word (verb, noun, adjective) in each subclass and calculate the value of TF*IDF, use formula (1) to calculate the variance of each notional word then.What variance embodied is the distribution characteristics of notional word, and variance is more little, and the distribution of this speech is stable more, the impossible more focus speech that becomes, can be with variance less than the notional word of a threshold value (0.07314133) as non-descriptor.
Wherein
Be the TF*IDF value of certain n period of notional word,
Be the mean value of this notional word TF*IDF on all the period of time.
2. public sentiment much-talked-about topic detection (seeing accompanying drawing 3) divides following step:
(1) makes up focus speech co-occurrence matrix
M M Ij Be right to the focus speech
W i With
W j A tolerance that occurs situation in same public sentiment text jointly, the big more co-occurrence degree that shows of value is high more.In the quantifying of reality was analyzed, the right co-occurrence frequency of speech was an absolute value, is difficult to reflect the degree that interdepends real between speech and the speech, needed the processing of word frequency rate containing.The method of containingization processing has a lot, and the present invention adopts the processing of mutual containing Y-factor method Y containing, sees formula (2).
（2）
Wherein,
H Ij Represent speech right
W i With
W j Co-occurrence number of times in text collection,
H i Represent the focus speech
W i Frequency of occurrence in text collection,
H j Represent the focus speech
W j Frequency of occurrence in text collection,
E Ij Value between 0～1.This method has symmetry, need not consider the focus speech
W i With
W j Appearance order.
(2) based on focus speech co-occurrence matrix, will
1/M Ij Right as speech
W i With
W j Semantic distance, adopt the ant colony clustering algorithm that the focus speech is carried out cluster.In iterative process, pick up the weights that object is searched for again by formula (3) assessment ant, can make ant general time loss as much as possible in the object point that does not find correct position like this, quicken cluster process convergent purpose thereby reach.
Distinctiveness ratio adopts centroid distance formula (4) between class between class.
Wherein
V NUM Be meant class Chinese version number,
V i The expression text vector,
V Center The representation class center.
Wherein
C i The cluster centre vector of representation class i.Setting the interior tight ness rating threshold value of class is
, work as so
Tight ness rating has all reached requirement in the time explanation class.The distinctiveness ratio threshold value is between setting class and the class
, when
The time, illustrate that the distinctiveness ratio between class and the class has reached requirement.When the both reaches the threshold value requirement is that clustering algorithm can select to stop to continue iteration.Each focus part of speech group represented a much-talked-about topic after cluster was finished.The data volume that term clustering need be handled with respect to text cluster significantly reduces, and counting yield significantly improves.
(3) the focus speech is as first vegetarian refreshments, the frequency that the focus speech occurs in language material is as degree of membership, focus part of speech group is converted into a fuzzy set, expresses as the formalization of much-talked-about topic, and with the frequency of word in the focus part of speech group temperature index as much-talked-about topic.
3. trend mining rule storehouse obtains the learn-by-example method of employing based on decision tree, divides following step:
(1) compiles typical public feelings information, make up the public sentiment corpus (the present invention adopts ends of the earth forum language material) of a sequential;
(2) be that unit is divided into subclass with the public sentiment corpus with the sky;
(3) to the method in each language material subclass employing step 2, obtain much-talked-about topic and temperature value thereof;
(4) with preceding n-1 days language material as predicted condition, n days language materials constitute sample data as predicting the outcome;
(5) employing is obtained fuzzy inference rule based on the learn-by-example method of decision tree, and rule is " IF-THEN+degree of confidence " form.Through study, obtain 316 of inference rules altogether, the inference rule that wherein predicts the outcome to ' trend=fast rise ' sees the following form.
Article one, the meaning of rule is: if certain much-talked-about topic is ' fast rise ' in the actual temperature variation of period i and period i-1, then it is changed to the possibility ' greatly ' of ' fast rise ' in the temperature of period i-2.The meaning of Else Rule can the rest may be inferred.
In these inference rules, temperature changes with degree of confidence all gets the language value, and wherein temperature variation language value set is: fast rise, fast rising slowly rise, and steadily, slowly decline, very fast decline descends fast }; Degree of confidence language value set is: and greatly, very big, in, very little, minimum }.Each language value defines with a fuzzy subset, sees formula (6)-(17), and the membership function curve map is seen Fig. 4, Fig. 5.
（7）
（12）
4. the public sentiment much-talked-about topic is followed the tracks of with trend excavation (seeing accompanying drawing 4) and is divided following step:
(1) data in continuous acquisition public feelings information source, to every day language material adopt method in the step 2 with real-time language material, obtain much-talked-about topic and temperature value thereof;
(2) use
MinkowskiDistance (seeing formula (18)) is calculated the temperature variation that much-talked-about topic takes place on continuous time, and draws the temperature change curve of much-talked-about topic, realizes the tracking of much-talked-about topic.
(3) under trend mining rule storehouse (fuzzy inference rule that is obtained by step 3 constitutes) supported, with historical public sentiment focus and temperature change information thereof as predicted condition (reasoning former piece), adopt fuzzy logic to carry out reasoning, obtain the trend knowledge that the future development of reflection public sentiment changes.
5. public sentiment crisis automatic early-warning, main points are:
Be predicted as ' fast rise ' and possibility for the temperature of excavating and adopt directly perceived, eye-catching multimedia means to send information warning for the much-talked-about topic of ' greatly ' or ' very big ' in step 4, specifically mode comprises:
(1) voice suggestion;
(2) the highlighted demonstration of figure;
(3) animated show.
The present invention proposes a kind of real-time network public-opinion crisis alert method, this method defiber technology and at line technology two parts.Off-line partly adopts statistical study and machine learning method, obtains non-theme dictionary and trend mining rule storehouse.Online part realizes detection, tracking, the trend excavation of network public-opinion much-talked-about topic, finally realizes the automatic early-warning of public sentiment crisis.Novelty of the present invention is mainly reflected in:
(1) off-line learning with combine in line computation, can realize the real-time early warning of public sentiment crisis;
(2) extract based on the focus speech of two-stage filtration and detect based on the much-talked-about topic of term clustering, computing velocity is fast;
(3) based on the much-talked-about topic similarity calculating of fuzzy semantics distance calculation with from motion tracking, algorithm is robust more;
(4) excavate based on the trend of machine learning and fuzzy reasoning, it is more credible to excavate the result;
(5) with the evolutionary process of multimedia show Technical Follow-Up much-talked-about topic, the early warning mode is visual in image.
Embodiment: network public-opinion monitoring and early warning
Hangzhou net forum ' Hangzhoupro netizen's sound ' version: http://bbs.hangzhou.com.cn/forum-88-1.html
Period: 2011.3.1---2011.3.5, wherein 3.1---3.4 is tracking results, and 3.5 for predicting the outcome.
Much-talked-about topic trend changes sees Fig. 7 a, and public sentiment crisis alert example is seen Fig. 7 b.
Much-talked-about topic has following 5:
1, Premier Wen is about the question of morality of house developer.(Wen Jiabao, developer, morals)
2, the retirement age is postponed problem.(retirement, age postpone)
3, Zhejiang University changes the secretary.(Zhejiang University, the secretary, Jin Deshui)
4, house property limit in Hangzhou is purchased policy.(Hangzhou, limit is purchased, house property)
5, thatch is hit pupil's incident to rise the teacher of experimental primary school.(Mao Yisheng hits, the student)
6, the regulation and control problem discussion of room rate.(room rate, regulation and control).
Claims (10)
1. network public-opinion crisis alert method is characterized in that this method comprises the steps:
(1) non-theme dictionary and the trend mining rule storehouse based on statistical study and machine learning obtains;
(2) based on the network public-opinion focus speech rapid extraction of two-stage filtration;
(3) based on the network public-opinion much-talked-about topic detection of speech analysis and term clustering altogether;
(4) much-talked-about topic of calculating based on continuous time much-talked-about topic similarity is from motion tracking;
(5) based on the network public-opinion trend knowledge excavation of fuzzy reasoning;
(6) based on the network public-opinion crisis automatic early-warning of multimedia show technology.
2. network public-opinion crisis alert method as claimed in claim 1 is characterized in that described non-theme dictionary obtains to comprise the steps:
2-1. the distribution characteristics of each notional word in the calculating corpus;
2-2. with the notional word of distributional stability as non-descriptor.
3. network public-opinion crisis alert method as claimed in claim 2, the distribution characteristics that it is characterized in that described notional word adopts the TF * variance of IDF value on all the period of time of word, and distributional stability is meant that variance is less than 0.07314133.
4. network public-opinion crisis alert method as claimed in claim 1 is characterized in that described trend mining rule storehouse obtains may further comprise the steps:
4-1. as predicted condition, n time window constitutes sample data as predicting the outcome with n-1 time window before the sequential public sentiment corpus;
4-2. adopt learn-by-example algorithm, obtain the trend mining rule based on decision tree.
5. network public-opinion crisis alert method as claimed in claim 4 is characterized in that described trend mining rule adopts following form, and the value of degree of confidence adopts the language value;
Trend mining rule form: IF-THEN+degree of confidence.
6. network public-opinion crisis alert method as claimed in claim 1 is characterized in that described network public-opinion focus speech rapid extraction based on two-stage filtration comprises the steps:
6-1. stop words filters;
6-2. non-descriptor is filtered.
7. network public-opinion crisis alert method as claimed in claim 1 is characterized in that described the detection based on the network public-opinion much-talked-about topic that is total to speech analysis and term clustering comprises the steps:
7-1. make up focus speech co-occurrence matrix;
Realize the focus term clustering 7-2. adopt the ant colony clustering algorithm;
7-3. use based on the fuzzy set of focus part of speech group and express much-talked-about topic.
8. network public-opinion crisis alert method as claimed in claim 1 is characterized in that described automatic tracking of much-talked-about topic of calculating based on continuous time much-talked-about topic similarity comprises the steps:
8-1. adopt the similarity between the fuzzy set semantic distance computing method calculating much-talked-about topic, the fuzzy set semantic distance calculates and adopts
MinkowskiDistance, and semantic distance is mapped to a language value set;
8-2. draw the temperature change curve of each much-talked-about topic, much-talked-about topic followed the tracks of.
9. network public-opinion crisis alert method as claimed in claim 1 is characterized in that described network public-opinion trend knowledge excavation based on fuzzy reasoning comprises the steps:
9-1. gather public feelings information in real time, extract much-talked-about topic;
9-2. adopt fuzzy logic to carry out reasoning, obtain much-talked-about topic trend knowledge.
10. network public-opinion crisis alert method as claimed in claim 1; it is characterized in that described network public-opinion crisis automatic early-warning based on the multimedia show technology is: be predicted as the topic that increases rapidly for temperature, adopt voice, figure, animation mode to provide information warning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101275099A CN102194001A (en) | 2011-05-17 | 2011-05-17 | Internet public opinion crisis early-warning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101275099A CN102194001A (en) | 2011-05-17 | 2011-05-17 | Internet public opinion crisis early-warning method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102194001A true CN102194001A (en) | 2011-09-21 |
Family
ID=44602070
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011101275099A Pending CN102194001A (en) | 2011-05-17 | 2011-05-17 | Internet public opinion crisis early-warning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102194001A (en) |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177090A (en) * | 2013-03-08 | 2013-06-26 | 亿赞普(北京)科技有限公司 | Topic detection method and device based on big data |
| CN103631901A (en) * | 2013-11-20 | 2014-03-12 | 清华大学 | Rumor control method based on maximum spanning tree of user-trusted network |
| CN104111999A (en) * | 2014-07-02 | 2014-10-22 | 烽火通信科技股份有限公司 | Search engine hot word analytical algorithm |
| CN104636408A (en) * | 2014-08-21 | 2015-05-20 | 中国科学院计算技术研究所 | News authentication early warning method and system based on user generated content |
| CN104809751A (en) * | 2015-04-30 | 2015-07-29 | 百度在线网络技术（北京）有限公司 | Method and device for generating event group evolution diagram |
| CN104866561A (en) * | 2015-05-19 | 2015-08-26 | 国家计算机网络与信息安全管理中心 | Method for mining microblog topic tendency initiator |
| CN105447081A (en) * | 2015-11-04 | 2016-03-30 | 国云科技股份有限公司 | Cloud platform-oriented government affair and public opinion monitoring method |
| CN106156192A (en) * | 2015-04-21 | 2016-11-23 | 北大方正集团有限公司 | Public sentiment data clustering method and public sentiment data clustering system |
| CN106531151A (en) * | 2016-11-16 | 2017-03-22 | 北京云知声信息技术有限公司 | Voice recognition method and voice recognition device |
| CN107391712A (en) * | 2017-07-28 | 2017-11-24 | 王亚迪 | A kind of network public opinion trend prediction analysis method |
| CN107491499A (en) * | 2017-07-27 | 2017-12-19 | 杭州中奥科技有限公司 | A kind of public sentiment method for early warning based on unstructured data |
| CN108614813A (en) * | 2017-12-19 | 2018-10-02 | 武汉楚鼎信息技术有限公司 | A kind of stock market's subject matter public sentiment temperature calculating method and system device |
| WO2018184306A1 (en) * | 2017-04-07 | 2018-10-11 | 平安科技（深圳）有限公司 | Topic alarm method, device, computer apparatus, and storage medium |
| CN108733791A (en) * | 2018-05-11 | 2018-11-02 | 北京科技大学 | network event detection method |
| CN109255026A (en) * | 2018-08-23 | 2019-01-22 | 云南师范大学 | A method of it is analyzed based on Co-word analysis and the learning demand of clustering |
| CN109299374A (en) * | 2018-10-24 | 2019-02-01 | 重庆理工大学 | Online social network information based on artificial neural network is propagated and the positive construction method of public sentiment evolution and system |
| CN109635283A (en) * | 2018-11-26 | 2019-04-16 | 汉纳森（厦门）数据股份有限公司 | It is a kind of based on the occurred events of public safety method for early warning in advance for excavating citizen and complaining text |
| CN110096652A (en) * | 2019-05-06 | 2019-08-06 | 上海汽车集团股份有限公司 | Public sentiment wind vane index calculation method and device, readable storage medium storing program for executing |
| CN111190996A (en) * | 2020-04-10 | 2020-05-22 | 支付宝(杭州)信息技术有限公司 | Topic mining method and device |
| CN111914152A (en) * | 2020-06-30 | 2020-11-10 | 中国科学院计算技术研究所 | Network event early warning method and system |
| CN112732904A (en) * | 2020-10-15 | 2021-04-30 | 中科曙光南京研究院有限公司 | Abnormal emergency detection method and system based on text processing |
| US11301629B2 (en) | 2019-08-21 | 2022-04-12 | International Business Machines Corporation | Interleaved conversation concept flow enhancement |
-
2011
- 2011-05-17 CN CN2011101275099A patent/CN102194001A/en active Pending
Cited By (37)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177090B (en) * | 2013-03-08 | 2016-11-23 | 亿赞普(北京)科技有限公司 | A kind of topic detection method and device based on big data |
| CN103177090A (en) * | 2013-03-08 | 2013-06-26 | 亿赞普(北京)科技有限公司 | Topic detection method and device based on big data |
| CN103631901A (en) * | 2013-11-20 | 2014-03-12 | 清华大学 | Rumor control method based on maximum spanning tree of user-trusted network |
| CN103631901B (en) * | 2013-11-20 | 2017-01-18 | 清华大学 | Rumor control method based on maximum spanning tree of user-trusted network |
| CN104111999A (en) * | 2014-07-02 | 2014-10-22 | 烽火通信科技股份有限公司 | Search engine hot word analytical algorithm |
| CN104111999B (en) * | 2014-07-02 | 2017-12-12 | 南京烽火星空通信发展有限公司 | A kind of search engine focus word parser |
| CN104636408A (en) * | 2014-08-21 | 2015-05-20 | 中国科学院计算技术研究所 | News authentication early warning method and system based on user generated content |
| CN104636408B (en) * | 2014-08-21 | 2017-08-08 | 中国科学院计算技术研究所 | News certification method for early warning and system based on user-generated content |
| CN106156192A (en) * | 2015-04-21 | 2016-11-23 | 北大方正集团有限公司 | Public sentiment data clustering method and public sentiment data clustering system |
| CN104809751A (en) * | 2015-04-30 | 2015-07-29 | 百度在线网络技术（北京）有限公司 | Method and device for generating event group evolution diagram |
| CN104809751B (en) * | 2015-04-30 | 2017-11-24 | 百度在线网络技术（北京）有限公司 | The method and apparatus for generating event group evolution diagram |
| CN104866561A (en) * | 2015-05-19 | 2015-08-26 | 国家计算机网络与信息安全管理中心 | Method for mining microblog topic tendency initiator |
| CN104866561B (en) * | 2015-05-19 | 2018-09-07 | 国家计算机网络与信息安全管理中心 | A method of excavating microblog topic trend promoter |
| CN105447081A (en) * | 2015-11-04 | 2016-03-30 | 国云科技股份有限公司 | Cloud platform-oriented government affair and public opinion monitoring method |
| CN106531151A (en) * | 2016-11-16 | 2017-03-22 | 北京云知声信息技术有限公司 | Voice recognition method and voice recognition device |
| CN106531151B (en) * | 2016-11-16 | 2019-10-11 | 北京云知声信息技术有限公司 | Audio recognition method and device |
| WO2018184306A1 (en) * | 2017-04-07 | 2018-10-11 | 平安科技（深圳）有限公司 | Topic alarm method, device, computer apparatus, and storage medium |
| US11205046B2 (en) | 2017-04-07 | 2021-12-21 | Ping An Technology (Shenzhen) Co., Ltd. | Topic monitoring for early warning with extended keyword similarity |
| CN107491499A (en) * | 2017-07-27 | 2017-12-19 | 杭州中奥科技有限公司 | A kind of public sentiment method for early warning based on unstructured data |
| CN107491499B (en) * | 2017-07-27 | 2018-09-04 | 杭州中奥科技有限公司 | A kind of public sentiment method for early warning based on unstructured data |
| CN107391712A (en) * | 2017-07-28 | 2017-11-24 | 王亚迪 | A kind of network public opinion trend prediction analysis method |
| CN108614813A (en) * | 2017-12-19 | 2018-10-02 | 武汉楚鼎信息技术有限公司 | A kind of stock market's subject matter public sentiment temperature calculating method and system device |
| CN108733791A (en) * | 2018-05-11 | 2018-11-02 | 北京科技大学 | network event detection method |
| CN108733791B (en) * | 2018-05-11 | 2020-11-20 | 北京科技大学 | Network event detection method |
| CN109255026B (en) * | 2018-08-23 | 2021-06-25 | 云南师范大学 | Learning demand analysis method based on common word analysis and cluster analysis |
| CN109255026A (en) * | 2018-08-23 | 2019-01-22 | 云南师范大学 | A method of it is analyzed based on Co-word analysis and the learning demand of clustering |
| CN109299374B (en) * | 2018-10-24 | 2019-05-21 | 重庆理工大学 | Online social network information based on artificial neural network is propagated and the positive construction method of public sentiment evolution and system |
| CN109299374A (en) * | 2018-10-24 | 2019-02-01 | 重庆理工大学 | Online social network information based on artificial neural network is propagated and the positive construction method of public sentiment evolution and system |
| CN109635283A (en) * | 2018-11-26 | 2019-04-16 | 汉纳森（厦门）数据股份有限公司 | It is a kind of based on the occurred events of public safety method for early warning in advance for excavating citizen and complaining text |
| CN109635283B (en) * | 2018-11-26 | 2022-10-14 | 汉纳森（厦门）数据股份有限公司 | Public safety event pre-warning method based on mining citizen complaint text |
| CN110096652A (en) * | 2019-05-06 | 2019-08-06 | 上海汽车集团股份有限公司 | Public sentiment wind vane index calculation method and device, readable storage medium storing program for executing |
| US11301629B2 (en) | 2019-08-21 | 2022-04-12 | International Business Machines Corporation | Interleaved conversation concept flow enhancement |
| US11757812B2 (en) | 2019-08-21 | 2023-09-12 | International Business Machines Corporation | Interleaved conversation concept flow enhancement |
| CN111190996A (en) * | 2020-04-10 | 2020-05-22 | 支付宝(杭州)信息技术有限公司 | Topic mining method and device |
| CN111914152A (en) * | 2020-06-30 | 2020-11-10 | 中国科学院计算技术研究所 | Network event early warning method and system |
| CN111914152B (en) * | 2020-06-30 | 2023-05-12 | 中国科学院计算技术研究所 | Network event early warning method and system |
| CN112732904A (en) * | 2020-10-15 | 2021-04-30 | 中科曙光南京研究院有限公司 | Abnormal emergency detection method and system based on text processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102194001A (en) | Internet public opinion crisis early-warning method | |
| Qiu et al. | Geoscience keyphrase extraction algorithm using enhanced word embedding | |
| CN104484815B (en) | Based on fuzzy ontology towards the sentiment analysis method and system in terms of product | |
| Hidayatullah et al. | Road traffic topic modeling on Twitter using latent dirichlet allocation | |
| CN111143576A (en) | Event-oriented dynamic knowledge graph construction method and device | |
| CN108536801A (en) | A kind of civil aviaton's microblogging security public sentiment sentiment analysis method based on deep learning | |
| CN113254667A (en) | Scientific and technological figure knowledge graph construction method and device based on deep learning model and terminal | |
| Sayfullina et al. | Learning representations for soft skill matching | |
| CN109710769A (en) | A kind of waterborne troops's comment detection system and method based on capsule network | |
| Fan et al. | A system analytics framework for detecting infrastructure-related topics in disasters using social sensing | |
| Zhang et al. | Covid-19 public opinion and emotion monitoring system based on time series thermal new word mining | |
| CN105760499A (en) | Method for analyzing and predicting online public opinion based on LDA topic models | |
| CN106874419B (en) | A kind of real-time hot spot polymerization of more granularities | |
| CN108470061A (en) | A kind of emotional semantic classification system for visual angle grade text | |
| Zhuang et al. | Analysis of public opinion evolution of COVID-19 based on LDA-ARMA hybrid model | |
| Bhatia et al. | Opinion target extraction with sentiment analysis | |
| Xu et al. | BERT gated multi-window attention network for relation extraction | |
| CN108536781A (en) | A kind of method for digging and system of social networks mood focus | |
| Gupta et al. | A novel classification approach based on context connotative network (CCNet): A case of construction site accidents | |
| Wilson et al. | A developed framework for multi-document summarization using softmax regression and spider monkey optimization methods | |
| Li et al. | Power system transient stability prediction algorithm based on ReliefF and LSTM | |
| Das et al. | Subjectivity detection using genetic algorithm | |
| Lisjana et al. | Classifying complaint reports using rnn and handling imbalanced dataset | |
| Wu et al. | Typical opinions mining based on Douban film comments in animated movies | |
| Qureshi et al. | Detecting social polarization and radicalization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20110921 |