Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016) 2016
DOI: 10.18653/v1/s16-1066
|View full text |Cite
|
Sign up to set email alerts
|

UWB at SemEval-2016 Task 6: Stance Detection

Abstract: This paper describes our system participating in the SemEval 2016 task: Detecting stance in Tweets. The goal was to identify whether the author of a tweet is in favor of the given target or against. Our approach is based on a maximum entropy classifier, which uses surface-level, sentiment and domain-specific features. We participated in both the supervised and weakly supervised subtasks and received promising results for most of the targets.

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
12
0
3

Year Published

2016
2016
2022
2022

Publication Types

Select...
4
2
2

Relationship

0
8

Authors

Journals

citations
Cited by 22 publications
(15 citation statements)
references
References 10 publications
0
12
0
3
Order By: Relevance
“…At the feature extraction phase, the following feature types were generated [25], [30]: word and character n-grams, dependency features, target features, stanceindicative features, linguistic features, stylistic features, sentiment features and word embeddings (Table I). These feature types are described in detail below.…”
Section: B Feature Extractionmentioning
confidence: 99%
See 1 more Smart Citation
“…At the feature extraction phase, the following feature types were generated [25], [30]: word and character n-grams, dependency features, target features, stanceindicative features, linguistic features, stylistic features, sentiment features and word embeddings (Table I). These feature types are described in detail below.…”
Section: B Feature Extractionmentioning
confidence: 99%
“…Stance-indicative features are formed on the basis of Favor and Against lists generated as proposed in [30]. For each word the ratio of its frequency in the texts of the favor class to its frequency in the whole training corpus is calculated.…”
Section: Stance-indicative Featuresmentioning
confidence: 99%
“…Table III shows the comparison of our proposed system with the existing state-of-the-art system of SemEval 2016 Task 6 for the sentiment dataset. [8] used feature-based SVM, [40] used keyword rules, LitisMind relied on hashtag rules on external data, [39] utilized a combination of sentiment classifiers and rules, whereas [38] used a maximum entropy classifier with domain-specific features. Our system comfortably surpasses the existing best system at SemEval.…”
Section: Modelsmentioning
confidence: 99%
“…Task A External resources: More than 1.5 million of tweets were added by using representative hashtag for target-stance pairs. UWB [20] Overall approach: Maximum entropy classifier. Tasks A and B External resources: N-grams, PoS labels, General Inquirer.…”
Section: Our Approachmentioning
confidence: 99%