SemEval-2016 Task 4: Sentiment Analysis in Twitter

Nakov, Preslav; Ritter, Alan; Rosenthal, Sara; Sebastiani, Fabrizio; Stoyanov, Veselin

doi:10.18653/v1/s16-1001

Cited by 567 publications

(569 citation statements)

References 46 publications

Supporting

Mentioning

549

Contrasting

Unclassified

Order By: Relevance

“…As a result, the presented system was ranked 1 st out of 34 participants, with an F1-score of 63.30 on the Twitter-2016 test set. See (Nakov et al, 2016) for further details. Table 4 summarizes the results of individual subsystems, as well as the final system on each test set.…”

Section: Resultsmentioning

confidence: 99%

SwissCheese at SemEval-2016 Task 4: Sentiment Classification Using an Ensemble of Convolutional Neural Networks with Distant Supervision

Deriu

Gonzenbach

Uzdilli

et al. 2016

Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016)

106

View full text Add to dashboard Cite

In this paper, we propose a classifier for predicting message-level sentiments of English micro-blog messages from Twitter. Our method builds upon the convolutional sentence embedding approach proposed by (Severyn and Moschitti, 2015a; Severyn and Moschitti, 2015b). We leverage large amounts of data with distant supervision to train an ensemble of 2-layer convolutional neural networks whose predictions are combined using a random forest classifier. Our approach was evaluated on the datasets of the SemEval-2016 competition (Task 4) outperforming all other approaches for the Message Polarity Classification task.

show abstract

Section: Resultsmentioning

confidence: 99%

SwissCheese at SemEval-2016 Task 4: Sentiment Classification Using an Ensemble of Convolutional Neural Networks with Distant Supervision

Deriu

Gonzenbach

Uzdilli

et al. 2016

Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016)

106

View full text Add to dashboard Cite

show abstract

“…The minimum count was set to 5 since words used fewer than 5 times do not add any information to the analysis and the default setting in Gensim is set to 5 as the minimum count [35]. Adding more training data could improve these results, however, a study by Nakov et al annotated 6,000 tweets and had similar F1-scores to our study [36]. Since these words occured fewer than 5 times, the algorithm was not able to identify these tweets as negative as it was not able to determine the words closer to these words.…”

Section: Classification Analysismentioning

confidence: 77%

Identifying Key Topics Bearing Negative Sentiment on Twitter: Insights Concerning the 2015-2016 Zika Epidemic (Preprint)

Mamidi¹,

Miller²,

Banerjee³

et al. 2018

Preprint

View full text Add to dashboard Cite

Background In order to understand public sentiment regarding the Zika virus, social media can be leveraged to understand how positive, negative, and neutral sentiments are expressed in society. Specifically, understanding the characteristics of negative sentiment could help inform federal disease control agencies' efforts to disseminate relevant information to the public about Zika related issues. Objective The purpose of this study was to analyze public sentiment concerning Zika using posts on Twitter and determine the qualitative characteristics of positive, negative and neutral sentiments expressed. Methods Machine learning techniques and algorithms were used to analyze the sentiment of tweets concerning Zika. A supervised machine learning classifier was built to classify tweets into three sentiment categories: positive, neutral, and negative. Tweets in each category were then examined using a topic modeling approach in order to determine the main topics for each category, with focus on the negative category. Results A total of 5,303 tweets were manually annotated and used to train multiple classifiers. These performed moderately well (F1 score = 0.69, 0.68) with text-based feature extraction. All 48,734 tweets were then categorized into the sentiment categories. Ten topics for each sentiment category were identified using topic modeling with a focus on the negative sentiment category. Conclusions Our study demonstrates how sentiment expressed within discussions of epidemics on Twitter can be discovered. This allows public health officials to understand public sentiment regarding an epidemic and enables them to address specific elements of negative sentiment in real-time. Our negative sentiment classifier was able to identify tweets concerning Zika with three broad themes: neural defects, Zika abnormalities, and reports and findings. These broad themes were based on domain expertise and from topics discussed in journals such as MMWR and Vaccine. Since the majority of topics in the negative sentiment category concerned symptoms, officials should focus on spreading information about prevention and treatment research.

show abstract

“…The goal of sentiment analysis is to specify the behavior of the person based on the general contextual polarity of the text (negative, positive, or neutral). It has many useful applications in education, commerce, heath, security and numerous others (Nakov et al, 2016;Imran et al, 2015). Since the sentiment is a language dependent, it is significantly important to consider the language and culture characteristics in analysing the sentiment.…”

Section: Introductionmentioning

confidence: 99%

“…Since the sentiment is a language dependent, it is significantly important to consider the language and culture characteristics in analysing the sentiment. Several sentiment analysis studies have been carried out based on the English language (Nakov et al, 2016;Poria et al, 2016;Agarwal and Mittal, 2016;Cambria, 2016). The Arabic language is one of the top Semitic languages; approximately 422 million persons use the Arabic language in 27 countries (UNESCO, 2012).…”

Section: Introductionmentioning

confidence: 99%

Hybrid approach for sentiment analysis of Arabic tweets based on deep learning model and features weighting

Altaher

2017

Int. j. adv. appl. sci.

View full text Add to dashboard Cite

The increasing adoption of social media networks as a platform for sharing opinions on different aspects emerged the sentiment analysis and opinion mining as an active research area. Recently, the sentiment analysis on Twitter has attracted considerable attention due to its many applications in various aspects of our lives. Many approaches have been presented for sentiment analysis based on English language, thus there is a need for efficient sentiment analysis approaches for Arabic language, since it has different structure when compared to other languages. This paper proposes a hybrid approach for sentiment analysis of Arabic tweets based on two stages. Firstly, the pre-processing methods like stop-word removal, tokenization and stemming are applied, and then two features weighting algorithms (information gain and chai square) are utilized to assign high weights to the most significant features of the Arabic tweets. Secondly, the deep learning technique is employed to effectively and accurately classify the Arabic tweets either as positive or negative tweets. The performance of the proposed approach has been compared with some of the classification methods such as Decision Tree (DT), Neural Networks (NN) and Support Vector Machine (SVM) using the dataset collected from Arabic tweets. The proposed approach outperforms the other approaches and achieved highest accuracy and precision of 90% and 93.7%, respectively.

show abstract

SemEval-2016 Task 4: Sentiment Analysis in Twitter

Cited by 567 publications

References 46 publications

SwissCheese at SemEval-2016 Task 4: Sentiment Classification Using an Ensemble of Convolutional Neural Networks with Distant Supervision

SwissCheese at SemEval-2016 Task 4: Sentiment Classification Using an Ensemble of Convolutional Neural Networks with Distant Supervision

Identifying Key Topics Bearing Negative Sentiment on Twitter: Insights Concerning the 2015-2016 Zika Epidemic (Preprint)

Hybrid approach for sentiment analysis of Arabic tweets based on deep learning model and features weighting

Contact Info

Product

Resources

About