Exploiting Context for Rumour Detection in Social Media

Zubiaga, Arkaitz; Liakata, Maria; Procter, Rob

doi:10.1007/978-3-319-67217-5_8

Cited by 183 publications

(121 citation statements)

References 28 publications

Supporting

Mentioning

116

Contrasting

Unclassified

Order By: Relevance

“…Then, the augmented tweets for the PHEME5 events are merged with the original PHEME5. Table VI 12 shows the number of source tweets and replies obtained via our data augmentation method and those after balancing augmented data and merging the balanced data with the original PHEME5. The number of conversational threads in original PHEME5 is provided in parentheses for comparison.…”

Section: A Data Augmentationmentioning

confidence: 99%

“…The artificial augmentation of training data helps to alleviate data sparseness and class imbalance, reduce over-fitting, and reduce generalization error, thereby sustaining deeper networks and improving their performance. We argue that enriching existing labeled rumor data with duplicated (but unique) tweets or corresponding variants is a promising attempt for early rumor detection methods [12] that rely on the structure of rumor propagation. Recent findings [13,6] show that rumors spread via the distribution of original sources.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Neural language model based training data augmentation for weakly supervised early rumor detection

Han

Gao

Ciravegna

2019

Proceedings of the 2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining

View full text Add to dashboard Cite

The scarcity and class imbalance of training data are known issues in current rumor detection tasks. We propose a straight-forward and general-purpose data augmentation technique which is beneficial to early rumor detection relying on event propagation patterns. The key idea is to exploit massive unlabeled event data sets on social media to augment limited labeled rumor source tweets. This work is based on rumor spreading patterns revealed by recent rumor studies and semantic relatedness between labeled and unlabeled data. A state-ofthe-art neural language model (NLM) and large credibilityfocused Twitter corpora are employed to learn context-sensitive representations of rumor tweets. Six different real-world events based on three publicly available rumor datasets are employed in our experiments to provide a comparative evaluation of the effectiveness of the method. The results show that our method can expand the size of an existing rumor data set nearly by 200% and corresponding social context (i.e., conversational threads) by 100% with reasonable quality. Preliminary experiments with a state-of-the-art deep learning-based rumor detection model show that augmented data can alleviate over-fitting and class imbalance caused by limited train data and can help to train complex neural networks (NNs). With augmented data, the performance of rumor detection can be improved by 12.1% in terms of F-score. Our experiments also indicate that augmented training data can help to generalize rumor detection models on unseen rumors.

show abstract

Section: A Data Augmentationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Neural language model based training data augmentation for weakly supervised early rumor detection

Han

Gao

Ciravegna

2019

Proceedings of the 2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining

View full text Add to dashboard Cite

show abstract

“…Our research also sheds light on open research questions that we suggest should be addressed in future work. Our work here complements other components of a rumour classification system that we implemented in the PHEME project, including a rumour detection component [20,21], as well as a study into the diffusion of and reactions to rumour [22].…”

Section: Introductionmentioning

confidence: 99%

Discourse-aware rumour stance classification in social media using sequential classifiers

Zubiaga

Kochkina

Liakata

et al. 2018

Information Processing & Management

Self Cite

117

View full text Add to dashboard Cite

Rumour stance classification, defined as classifying the stance of specific social media posts into one of supporting, denying, querying or commenting on an earlier post, is becoming of increasing interest to researchers. While most previous work has focused on using individual tweets as classifier inputs, here we report on the performance of sequential classifiers that exploit the discourse features inherent in social media interactions or 'conversational threads'. Testing the effectiveness of four sequential classifiers -Hawkes Processes, Linear-Chain Conditional Random Fields (Linear CRF), Tree-Structured Conditional Random Fields (Tree CRF) and Long Short Term Memory networks (LSTM) -on eight datasets associated with breaking news stories, and looking at different types of local and contextual features, our work sheds new light on the development of accurate stance classifiers. We show that sequential classifiers that exploit the use of discourse properties in social media conversations while using only local features, outperform non-sequential classifiers. Furthermore, we show that LSTM using a reduced set of features can outperform the other sequential classifiers; this performance is consistent across datasets and across types of stances. To conclude, our work also analyses the different features under study, identifying those that best help characterise and distinguish between stances, such as supporting tweets being more likely to be accompanied by evidence than denying tweets. We also set forth a number of directions for future research.

show abstract

“…Classification based approaches classify tweets into credible and not credible based on features extracted from them using machine learning techniques especially supervised techniques [8,[12][13][14][15][17][18][19]. Supervised machine learning techniques require a ground truth that contains a dataset of annotated tweets with the features related to them.…”

Section: Introductionmentioning

confidence: 99%

“…Sourcebased features consider characteristics of the user such as the number of followers and if the user is verified. Some studies used a combination of content-based and source-based features [8,17,18]. After the feature dataset is built, the next step is to determine the optimal classification algorithm to train them.…”

Section: Introductionmentioning

confidence: 99%

Credibility Detection in Twitter Using Word N-gram Analysis and Supervised Machine Learning Techniques

Hassan¹,

Gomaa²,

Khoriba³

et al. 2020

IJIES

View full text Add to dashboard Cite

With the evolution of social media platforms, the Internet is used as a source for obtaining news about current events. Recently, Twitter has become one of the most popular social media platforms that allows public users to share the news. The platform is growing rapidly especially among young people who may be influenced by the information from anonymous sources. Therefore, predicting the credibility of news in Twitter becomes a necessity especially in the case of emergencies. This paper introduces a classification model based on supervised machine learning techniques and word-based N-gram analysis to classify Twitter messages automatically into credible and not credible. Five different supervised classification techniques are applied and compared namely: Linear Support Vector Machines (LSVM), Logistic Regression (LR), Random Forests (RF), Naïve Bayes (NB) and K-Nearest Neighbors (KNN). The research investigates two feature representations (TF and TF-IDF) and different word N-gram ranges. For model training and testing, 10-fold cross validation is performed on two datasets in different languages (English and Arabic). The best performance is achieved using a combination of both unigrams and bigrams, LSVM as a classifier and TF-IDF as a feature extraction technique. The proposed model achieves 84.9% Accuracy, 86.6% Precision, 91.9% Recall, and 89% F-Measure on the English dataset. Regarding the Arabic dataset, the model achieves 73.2% Accuracy, 76.4% Precision, 80.7% Recall, and 78.5% F-Measure. The obtained results indicate that word N-gram features are more relevant for the credibility prediction compared with content and source-based features, also compared with character N-gram features. Experiments also show that the proposed model achieved an improvement when compared to two models existing in the literature.

show abstract

Exploiting Context for Rumour Detection in Social Media

Cited by 183 publications

References 28 publications

Neural language model based training data augmentation for weakly supervised early rumor detection

Neural language model based training data augmentation for weakly supervised early rumor detection

Discourse-aware rumour stance classification in social media using sequential classifiers

Credibility Detection in Twitter Using Word N-gram Analysis and Supervised Machine Learning Techniques

Contact Info

Product

Resources

About