CN105740238B - A kind of event relation intensity map construction method merging sentence justice information - Google Patents
A kind of event relation intensity map construction method merging sentence justice information Download PDFInfo
- Publication number
- CN105740238B CN105740238B CN201610124157.4A CN201610124157A CN105740238B CN 105740238 B CN105740238 B CN 105740238B CN 201610124157 A CN201610124157 A CN 201610124157A CN 105740238 B CN105740238 B CN 105740238B
- Authority
- CN
- China
- Prior art keywords
- event
- word
- information
- sentence
- topic
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/258—Heading extraction; Automatic titling; Numbering
-
- 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
Abstract
The present invention relates to a kind of event relation intensity map construction methods for merging sentence justice information.It is primarily based on Chinese sentence meaning structural theory, extract sentence semantics information, the characteristic dimension of expansion event, and event vector is completed using improved TF-IDF method and is expressed, in conjunction with contextual information and core event Advance data quality event vector, finally using relationship strength between LDA method acquisition event, suitable relationship strength threshold value is set, constructs event relation intensity map.The present invention passes through fusion sentence justice information, provide a kind of event by isolated dispersion by it is a kind of it is mensurable in the form of the method that associates, and the relationship between event is intuitively shown by event relation intensity map, accurately positioning core event, the natural language processings applications such as the subsequent automatic abstract based on event relation of powerful support, public sentiment prediction.
Description
Technical field
The present invention relates to it is a kind of merge sentence justice information event relation intensity map construction method, belong to computer science and from
The information extraction field of right Language Processing.
Background technique
Event refer to some specific time and place generation, participated in by one or more roles, by one or more
Something of a movement composition.With the rapid development of development of Mobile Internet technology, after event occurs, especially emergency event,
Meeting rapid emergence goes out the texts such as a large amount of associated news report, microblogging comment and blog.Due to the event intrinsic time,
The attributes such as place, personage, cause, process, result are frequently not isolated, but can there are different from other events
The correlation of degree is usually measured with relationship strength.In order to be quickly and accurately positioned interrelated from a large amount of text
Event, need based between event relationship strength construct event relation intensity map, thus for automatic abstract, public sentiment predict etc. from
Right language application provides effectively support.
Currently, the research for event relation identification mainly has pattern matching method, element and position analysis method and rule
Rationalistic method.
Pattern matching method is one of the main method of event relation detection, mainly by affair character item instructional model
It establishes.Trigger word is the core of event, directly shows the generation of event, is the main feature for determining event category.According to thing
Relationship between part trigger word, Manual definition's template extract the event relation for meeting template in text.The mould of event relation detection
Formula matching process is studied often through between the relationship event trigger word, by the relationship between trigger word, is formulated corresponding
Mode, the identification of relationship between ancillary events.
Element and position analysis method, Event element are the important components of event, and Event element gives the ginseng of event
With person, the information such as time, place.Each event includes specific Event element information, and often shares certain between dependent event
One or certain several Event element.Event location can be shown that the context environmental that event occurs, and relevant event is in text
Successively occur with biggish probability.Therefore position and Event element play important role in the identification of event relation.Base
Event relation detection method in position and element, be using event location and element information as important clue identification events it
Between relationship method.
Rule-based reasoning method is mainly based upon " interval algebra " the algorithm building inference rule of Allen, and such as " if-then " is this kind of
Rule realizes event relation automated reasoning；Or rule is extended, the effective classifier of training carries out event relation
Classification.Such as the event-order serie relationship identification system of Mani and Tatu.
Existing method can be summarized as following two aspects: 1. not providing specific relationship type, detect around event relation
Task between event logical relation carry out whether there is or not judgement.But such methods only know event shallow-layer logical relation
Not, the semantic relation that cannot be deep into inside event, such as relationship type or connection tightness, from practical application, there are also certain differences
Away from.2. from a certain specific type relationship in classification relation, mainly based on causality and sequential relationship.But this
There are the following problems for class method: firstly, only studying a certain certain types of event relation, without universality and comprehensively
Property.Secondly, these methods have only carried out preliminary definition and category division to event relation in terms of event relation definition, but
Not yet form the unified definition of event relation.Meanwhile also not supported than more complete event relation mark corpus, this makes
Such method faces bigger difficulty in comprehensive identification events relationship.
Therefore, the effect is unsatisfactory for the existing event relation strength calculation method based on type identification, the event of building
Strength relationship figure is difficult to application.
Summary of the invention
The present invention is to solve the problems, such as that relationship type recognition effect is poor, event relation intensity map accuracy is not high, proposes one
The event relation intensity map construction method of kind fusion sentence justice information.Using Chinese semantics sentence justice structural theory excavate event it
Between inherent semantic relation, expand affair character dimension, establish event vector model, directly using vector calculating measurement event between
The power of incidence relation, and then event relation intensity map is constructed, realize the visualization of event correlation relationship.
Design principle of the invention are as follows: 1. are based on Chinese sentence meaning structural theory, extract sentence semantics information, expand event
Characteristic dimension；2. completing event vector using improved TF-IDF method to express；3. contextual information and core event is combined to believe
Breath completes the optimization of event vector；4. utilizing relationship strength between LDA method acquisition event；5. setting suitable relationship strength threshold
Value obtains event to set, constructs event relation intensity map.
It is described to be based on Chinese sentence meaning structural theory, sentence justice semantic information is extracted, the characteristic dimension for expanding event, which refers to, to be passed through
Text where analysis event extracts the topic of sentence using Chinese sentence meaning structural model, states topic, general term and elementary item.Wherein,
Topic and the division that topic is distich justice is stated, be described object using topic ingredient available sentence justice, stated and inscribe ingredient and can obtain
Take sentence justice to the description content of topic.Wherein, general term and elementary item are the marks to word, are obtained in sentence justice using general term
Ornamental equivalent, utilize the core of the available sentence of elementary item semantic.Using topic, state topic, general term and elementary item expansion
The available more semantic informations in affair character space.
The improved term frequency-inverse document frequency tf-idf method with the word in all texts it is characterized in that occur
Number summation and all texts in total word number TF value of the ratio as the word when, the short texts such as solution event to
Sparse Problems present in amount expression.
It is described to complete the optimization of event vector it is characterized in that the title for passing through introducing trigger word using core event information
Information Istit, location information Loc, frequency of occurrences information Freq, label word information Mark, word part-of-speech information Pos, topic
Topic states topic Comment, general term Comment Arg and elementary item Basic Arg and gives a mark to event trigger word.It chooses
The highest trigger word of score value substitutes former trigger word weight as core trigger word, and with marking weight, complete core event to
The optimization of amount.
The suitable relationship strength threshold value of setting is it is characterized in that retain the event that relationship strength value is greater than threshold value 0.2
It is right, and be attached with effective line segment, to obtain event relation figure.
The technical scheme is that be achieved by the steps of:
Step 1: being based on Chinese sentence meaning structural theory, extract sentence semantics information, expand the characteristic dimension of event；The sentence
Adopted information refers to that the sentence justice ingredient extracted in Chinese sentence meaning structural model, including predicate, topic state topic, elementary item and general term；
Wherein, topic is that sentence justice is described object, and stating topic is description content of the sentence justice to topic, elementary item is that the core of sentence is semantic
Information, general term are the ornamental equivalents in sentence justice；
Step 1.1, the corresponding text of event sets is segmented according to the semanteme of sentence justice structural model, part-of-speech tagging etc.
Processing obtains event sets and its corresponding sequence of terms；Noise word in sequence of terms is removed according to part of speech, including is described
Word and adverbial word, obtain Feature Words；
Step 1.2, it is based on Chinese sentence meaning structural theory, the predicate of extraction event place text, states topic, elementary item at topic
With general term as sentence justice feature, the trigger word and Event element of binding events itself obtain affair character space；
Step 2: on the basis of step 1 obtains affair character space, using improved TF-IDF method complete event to
Amount expression；The improved TF-IDF method, according to phrase semantic information, TF value calculating method is word in all texts
The ratio of total word number in the number summation of appearance and all texts, IDF value calculating method are text sum and word
The ratio of the text sum of appearance；Wherein, word TF value calculation formula are as follows:
Ntf(wi)=No/NV (1)
Ntf(wi) indicate word wiTF value, NoIndicate the number summation that the word occurs in all texts, NVIndicate institute
There is word number total in text.The calculation formula of word IDF value are as follows:
Nidf(wi)=Nm/Nm,o (2)
Nidf(wi) indicate word wiIDF value, NmThe sum for indicating text, if wiElementary item as sentence justice structural model
Occur, then Nm,oIndicate word wiAs the text number that elementary item occurs, if wiGeneral term as sentence justice structural model occurs,
Then Nm,oIndicate word wiThe text number occurred as general term.Word wiFinal TF-IDF value Ntfidf(wi) calculation formula are as follows:
Ntfidf(wi)=Ntf*Nidf (3)
Step 3, on the basis of step 2 obtains event vector expression, come in conjunction with core event information and contextual information
Optimize event vector；
Step 3.1, using the optimization of core event information core event vector, introduce trigger word heading message Istit,
Location information Loc, frequency of occurrences information Freq, label word information Mark, word part-of-speech information Pos, topic Topic, state topic
Comment, general term Comment Arg and elementary item Basic Arg give a mark to event trigger word, choose score value highest
Trigger word as core trigger word, corresponding event vector is core event vector, specific formula such as formula (4) institute that gives a mark
Show:
Former trigger word weight is substituted with marking weight, completes the optimization of core event vector；
Step 3.2, optimize event vector using contextual information, it is described to be referred to using contextual information optimization event vector
If there is same sentence in two events, the corresponding position of event vector for taking the weight of event trigger word below that filling is gone to be located at front
It sets, completes event vector optimization；
Step 4, strong using relationship between LDA method acquisition event on the basis of step 3 completes event vector optimization
Degree；
Step 4.1, the theme for obtaining word indicates as a result, its method particularly includes: using LDA model to obtaining word-thing
Part matrix is analyzed, and the theme for obtaining word indicates as a result, it is event number in text that theme number k, which takes N/2, N, at this time；
Step 4.2, the relationship strength between calculating event, method particularly includes: event is regarded as and is made of theme
Vector calculates the cosine angle value between event two-by-two, the relationship strength as event pair；Calculation formula are as follows:
Step 5, on the basis of step 4 obtains event relation two-by-two, suitable relationship strength threshold value is chosen, constructs event
Relationship strength figure, the suitable relationship strength threshold value are set as 0.2.
Beneficial effect
The present invention provides a kind of event relation intensity map construction method by fusion sentence justice information, can be by these orphans
The event of vertical dispersion associates in a kind of mensurable form, the problem for avoiding conventional event relation recognition effect poor, and
The incidence relation between event is intuitively shown by event relation intensity map, accurately positions core event, the subsequent base of powerful support
In natural language processings applications such as automatic abstract, the public sentiment predictions of event relation.
Detailed description of the invention
Fig. 1 is event relation intensity map construction method flow chart of the invention；
Fig. 2 is the corresponding event relation intensity map of the embodiment of the present invention.
Specific embodiment
The data that experiment uses come from Shanghai University's computer laboratory Chinese event corpus (CEC), select in the corpus
Fire class Emergent Public Events news report.4 paragraphs, 9 sentences are contained in the text of selection altogether.
Step 1, it is based on Chinese sentence meaning structural theory, extracting sentence semantics information, (including predicate, topic state topic, elementary item
And general term), expand the characteristic dimension of event；
Step 1.1, the corresponding text of event sets is segmented according to the semanteme of sentence justice structural model, part-of-speech tagging etc.
Processing obtains event sets and its corresponding sequence of terms；Noise word in sequence of terms is removed according to part of speech, including is described
Word and adverbial word, obtain Feature Words；
Step 1.2, it is based on Chinese sentence meaning structural theory, the predicate of extraction event place text, states topic, elementary item at topic
With general term as sentence justice feature, the trigger word and Event element of binding events itself obtain affair character space；
Step 2, on the basis of step 1 obtains affair character space, using improved TF-IDF method complete event to
Amount expression；Wherein, word TF value calculation formula are as follows:
Ntf(wi)=No/NV (1)
Ntf(wi) indicate word wiTF value, NoIndicate the number summation that the word occurs in all texts, NVIndicate institute
There is word number total in text.The calculation formula of word IDF value are as follows:
Nidf(wi)=Nm/Nm,o (2)
Nidf(wi) indicate word wiIDF value, NmThe sum for indicating text, if wiElementary item as sentence justice structural model
Occur, then Nm,oIndicate word wiAs the text number that elementary item occurs, if wiGeneral term as sentence justice structural model occurs,
Then Nm,oIndicate word wiThe text number occurred as general term；Word wiFinal TF-IDF value Ntfidf(wi) calculation formula are as follows:
Ntfidf(wi)=Ntf*Nidf (3)
Step 3, on the basis of step 2 obtains event vector expression, come in conjunction with core event information and contextual information
Optimize event vector；
Step 3.1, using the optimization of core event information core event vector, introduce trigger word heading message Istit,
Location information Loc, frequency of occurrences information Freq, label word information Mark, word part-of-speech information Pos, topic Topic, state topic
Comment, general term Comment Arg and elementary item Basic Arg give a mark to event trigger word, choose score value highest
Trigger word as core trigger word, corresponding event vector is core event vector, specific formula such as formula (4) institute that gives a mark
Show:
With the weight for serving as core trigger word in marking weight substitution core event vector, the excellent of core event vector is completed
Change.
Step 3.2, optimize event vector using contextual information.It is described to be referred to using contextual information optimization event vector
If there is same sentence in two events, the corresponding position of event vector for taking the weight of event trigger word below that filling is gone to be located at front
It sets, completes event vector optimization；
Step 4, strong using relationship between LDA method acquisition event on the basis of step 3 completes event vector optimization
Degree；
Step 4.1, obtained event-word matrix is analyzed using LDA model, the theme for obtaining word indicates knot
Fruit, it is event number in text that theme number k, which takes N/2, N, at this time；
Step 4.2, using the relationship strength between two events of vector cosine angle calcu-lation；Calculation formula are as follows:
Step 5, event relation intensity map is constructed.Retain the event pair that relationship strength value is greater than threshold value 0.2, finally obtains 16
A node, 51 pairs of events pair for meeting condition, using community network drawing tools NetDraw software by these events to scheme
Form is shown, as shown in Figure 2.
Claims (3)
1. a kind of event relation intensity map construction method for merging sentence justice information, described method includes following steps:
Step 1: being based on Chinese sentence meaning structural theory, extract sentence semantics information, expand the characteristic dimension of event；The sentence justice letter
Breath refers to that the sentence justice ingredient extracted in Chinese sentence meaning structural model, including predicate, topic state topic, elementary item and general term；Its
In, topic is that sentence justice is described object, and stating topic is description content of the sentence justice to topic, elementary item is the core semanteme letter of sentence
Breath, general term are the ornamental equivalents in sentence justice；
Step 1.1, the corresponding text of event sets is segmented according to the semanteme of sentence justice structural model, part-of-speech tagging, is obtained
Event sets and its corresponding sequence of terms；Noise word in sequence of terms, including adjective and adverbial word are removed according to part of speech,
Obtain Feature Words；
Step 1.2, it is based on Chinese sentence meaning structural theory, the predicate of extraction event place text, states topic, elementary item and one at topic
As item as sentence justice feature, the trigger word and Event element of binding events itself obtain affair character space；
Step 2: on the basis of step 1 obtains affair character space, completing event vector table using improved TF-IDF method
It reaches；The improved TF-IDF method, according to phrase semantic information, TF value calculating method is that word occurs in all texts
Number summation and all texts in total word number ratio, IDF value calculating method is that text sum and word occur
Text sum ratio；
Step 3: on the basis of step 2 obtains event vector expression, optimizing in conjunction with core event information and contextual information
Event vector；
Step 3.1, described to be referred to using core event Advance data quality event vector using core event Advance data quality event vector
Pass through the heading message Istit of text where introducing trigger word, location information Loc, frequency of occurrences information Freq, label word letter
Mark, word part-of-speech information Pos, topic Topic are ceased, topic Comment, general term Comment Arg and elementary item Basic are stated
Arg gives a mark to trigger word, chooses the highest trigger word of score value as core trigger word, corresponding event vector is
Core event vector, and former trigger word weight is substituted with marking weight, complete the optimization of event vector；
Step 3.2, optimize event vector using contextual information, if described refer to two using contextual information optimization event vector
There is same sentence in a event, the corresponding position of event vector for taking the weight of event trigger word below that filling is gone to be located at front, complete
At event vector optimization；
Step 4: on the basis of step 3 completes event vector optimization, utilizing relationship strength between LDA method acquisition event；
Step 4.1, obtain word theme indicate as a result, its method particularly includes: using LDA model to obtained event-word
Matrix is analyzed, and the theme for obtaining word indicates as a result, it is event number in text that theme number k, which takes N/2, N, at this time；
Step 4.2, the relationship strength between calculating event, method particularly includes: event is regarded as to the vector being made of theme,
Calculate the cosine angle value between event two-by-two, the relationship strength as event pair；
Step 5: on the basis of step 4 obtains event relation two-by-two, choosing suitable relationship strength threshold value, construct event relation
Intensity map, the suitable relationship strength threshold value are set as 0.2.
2. a kind of event relation intensity map construction method for merging sentence justice information according to claim 1, it is characterised in that:
In step 2, word TF value calculation formula are as follows:
Ntf(wi)=No/NV (1)
Ntf(wi) indicate word wiTF value, NoIndicate the number summation that the word occurs in all texts, NVIndicate all texts
Total word number, the calculation formula of word IDF value in this are as follows:
Nidf(wi)=Nm/Nm,o (2)
Nidf(wi) indicate word wiIDF value, NmThe sum for indicating text, if wiElementary item as sentence justice structural model occurs,
Then Nm,oIndicate word wiAs the text number that elementary item occurs, if wiGeneral term as sentence justice structural model occurs, then Nm,o
Indicate word wiThe text number occurred as general term；Word wiFinal TF-IDF value Ntfidf(wi) calculation formula is as follows:
Ntfidf(wi)=Ntf*Nidf (3)。
3. a kind of event relation intensity map construction method for merging sentence justice information according to claim 1, it is characterised in that:
The heading message Istit, location information Loc, frequency of occurrences information Freq, label word information of trigger word are introduced in step 3.1
Mark, word part-of-speech information Pos, topic Topic state topic Comment, general term Comment Arg and elementary item Basic Arg
It gives a mark to event trigger word, shown in specific formula such as formula (4):
The highest trigger word of score value is chosen as core trigger word, corresponding event vector is core event vector, is used
The marking weight that formula (4) is calculated substitutes former trigger word weight, completes the optimization of core event vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610124157.4A CN105740238B (en) | 2016-03-04 | 2016-03-04 | A kind of event relation intensity map construction method merging sentence justice information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610124157.4A CN105740238B (en) | 2016-03-04 | 2016-03-04 | A kind of event relation intensity map construction method merging sentence justice information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105740238A CN105740238A (en) | 2016-07-06 |
| CN105740238B true CN105740238B (en) | 2019-02-01 |
Family
ID=56249811
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610124157.4A Expired - Fee Related CN105740238B (en) | 2016-03-04 | 2016-03-04 | A kind of event relation intensity map construction method merging sentence justice information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105740238B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109543185B (en) * | 2018-11-22 | 2021-11-16 | 联想(北京)有限公司 | Statement topic acquisition method and device |
| CN110705255B (en) * | 2019-10-12 | 2021-05-25 | 京东数字科技控股有限公司 | Method and device for detecting association relation between sentences |
| CN115422948B (en) * | 2022-11-04 | 2023-01-24 | 文灵科技(北京)有限公司 | Event level network identification system and method based on semantic analysis |
| CN117435697B (en) * | 2023-12-21 | 2024-03-22 | 中科雨辰科技有限公司 | Data processing system for acquiring core event |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177089A (en) * | 2013-03-08 | 2013-06-26 | 北京理工大学 | Sentence meaning composition relationship lamination identification method based on central blocks |
| CN103544255A (en) * | 2013-10-15 | 2014-01-29 | 常州大学 | Text semantic relativity based network public opinion information analysis method |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9158755B2 (en) * | 2012-10-30 | 2015-10-13 | International Business Machines Corporation | Category-based lemmatizing of a phrase in a document |
| CN103176963B (en) * | 2013-03-08 | 2015-06-03 | 北京理工大学 | Chinese sentence meaning structure model automatic labeling method based on CRF ++ |
-
2016
- 2016-03-04 CN CN201610124157.4A patent/CN105740238B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177089A (en) * | 2013-03-08 | 2013-06-26 | 北京理工大学 | Sentence meaning composition relationship lamination identification method based on central blocks |
| CN103544255A (en) * | 2013-10-15 | 2014-01-29 | 常州大学 | Text semantic relativity based network public opinion information analysis method |
Non-Patent Citations (2)
| Title |
|---|
| Automatic Event Trigger Word Extraction in Chinese Event;Long Tian 等;《Journal of Software Engineering and Applications》;20120531;第208-212页 |
| 融合句义结构模型的微博话题摘要算法;林萌 等;《浙江大学学报》;20151231;第49卷(第12期);第2316-2325页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105740238A (en) | 2016-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110516067B (en) | Public opinion monitoring method, system and storage medium based on topic detection | |
| CN107609132B (en) | Semantic ontology base based Chinese text sentiment analysis method | |
| Salton et al. | Idiom token classification using sentential distributed semantics | |
| CN103049435B (en) | Text fine granularity sentiment analysis method and device | |
| CN106407236B (en) | A kind of emotion tendency detection method towards comment data | |
| CN105740238B (en) | A kind of event relation intensity map construction method merging sentence justice information | |
| CN103853834B (en) | Text structure analysis-based Web document abstract generation method | |
| Pilehvar et al. | Card-660: Cambridge rare word dataset-a reliable benchmark for infrequent word representation models | |
| CN103605658A (en) | Search engine system based on text emotion analysis | |
| CN105183717A (en) | OSN user emotion analysis method based on random forest and user relationship | |
| CN106547875B (en) | Microblog online emergency detection method based on emotion analysis and label | |
| WO2017198031A1 (en) | Semantic parsing method and apparatus | |
| Khasawneh et al. | Arabic sentiment polarity identification using a hybrid approach | |
| Celli et al. | Relationships between personality and interactions in facebook | |
| CN108009297B (en) | Text emotion analysis method and system based on natural language processing | |
| Stavrianou et al. | NLP-based feature extraction for automated tweet classification | |
| CN115017303A (en) | Method, computing device and medium for enterprise risk assessment based on news text | |
| CN105069647A (en) | Improved method for extracting evaluation object in Chinese commodity review | |
| Yan et al. | Tibetan sentence sentiment analysis based on the maximum entropy model | |
| US20230073602A1 (en) | System of and method for automatically detecting sarcasm of a batch of text | |
| JP7186080B2 (en) | Text information judgment device and its program | |
| Caicedo et al. | Bootstrapping semi-supervised annotation method for potential suicidal messages | |
| CN104978308B (en) | A kind of microblogging theme emotion evolution analysis method | |
| Celikyilmaz et al. | A joint model for discovery of aspects in utterances | |
| Chen et al. | Double Lda: a sentiment analysis model based on topic model |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190201 |