Twitter Stance Detection — A Subjectivity and Sentiment Polarity Inspired Two-Phase Approach
Citations
131 citations
Cites background from "Twitter Stance Detection — A Subjec..."
...These studies include [Addawood et al. 2017; Bar-Haim et al. 2017; Dey et al. 2017; Gadek et al. 2017; Grčar et al. 2017; HaCohen-kerner et al. 2017; Hercig et al. 2017; Küçük 2017a,b; Küçük and Can 2018; Kucher et al. 2017; Lai et al. 2018; Mohammad et al. 2017; Rohit and Singh 2018; Sen et al.…...
[...]
101 citations
58 citations
Cites methods from "Twitter Stance Detection — A Subjec..."
...ysis: Table 5 and Table 6 describe the performances of all models on SemEval dataset and MPCHI dataset respectively. Since we could not reproduce the Frame Semantics feature of the two-step SVM model [5], we have reported both the performances of our implementation and that reported in the original paper [5] for the SemEval datasets (the original paper worked only on the SemEval datasets, not the MPC...
[...]
51 citations
Cites background or methods from "Twitter Stance Detection — A Subjec..."
...Dey [49] employs so many elaborate features (such as sentiment lexicons, n-grams, POS tags) to capture the syntactic and semantic meaning of the tweets; ii) the small number of training data in this field limit the representation ability and the efficiency of the deep learning basedmodel....
[...]
...Although many deep learning based approaches [33], [41], [46], [48] obtain better results than the traditional feature based models [34]–[36], the traditional feature engineering based method [49] performs the best....
[...]
...Likewise, Dey [49] devises a two stage approach by using various features (e.g. MPQA subjectivity lexicon, WordNet Adjective) and obtains the best overall performance (Favg : 74.44%)....
[...]
...[49] K. Dey, R. Shrivastava, and S. Kaushik, ‘‘Twitter stance detection—...
[...]
...Likewise, Dey [49] devises a two stage approach by using various features (e....
[...]
40 citations
Cites methods from "Twitter Stance Detection — A Subjec..."
...Another SVM-based model consisted of two-step SVMs (Dey et al., 2017)....
[...]
References
73,978 citations
72,897 citations
"Twitter Stance Detection — A Subjec..." refers background in this paper
...Different models, including deep learning approaches such as convolutional neural networks (CNN) [16] [21], recurrent neural networks (RNN) [7] and long short-term memory (LSTM) [14] [15], traditional machine learning and genetic algorithms, were tried....
[...]
...as convolutional neural networks (CNN) [16] [21], recurrent neural networks (RNN) [7] and long short-term memory (LSTM) [14] [15], traditional machine learning and genetic algorithms, were tried....
[...]
...The second layer is composed of 128 Long Short-Term Memory (LSTM) units [15]....
[...]
20,077 citations
19,603 citations
"Twitter Stance Detection — A Subjec..." refers methods in this paper
...We perform preprocessing, followed by a two-phase SVM based machine learning using Weka [11] with a linear kernel and default parameters (cost function C = 1 and L2 regularizer), with 10% held-out data for the purpose of model development....
[...]
...Further note that, we use the Weka [11] tool to perform our machine learning activities....
[...]
15,068 citations
"Twitter Stance Detection — A Subjec..." refers methods in this paper
...We extensively use Wordnet [26], to enable the construction of this syntactic feature, wherein, we aim to detect whether a token given in the tweet content, exists in Wordnet and is marked as an adjective there, and construct a boolean feature accordingly....
[...]
...Note that, words that at all can be used as an adjective (as per Wordnet), are used to form this feature, and not the actual usage of the word in text (which would be tagged by the PoS tagger)....
[...]
...• Wordnet Based Potential Adjective Recognition: Hatzivassiloglou and Wiebe [13] show that different kinds of adjectives, such as dynamic adjectives, semantically oriented adjectives, and gradable adjectives are strong predictors of presence of subjectivity....
[...]
...Feature Novelty: We used two features that are novel in the context of the task at hand: the Wordnet-based potential adjective recognition feature in the first phase and the frame semantics feature in the second phase....
[...]
...Note that, the two features that are novel in the context of the task at hand, namely Wordnet-based potential adjective recognition and the frame semantics feature, deliver strong impacts, as is clear from the improvement these deliver in the average F-score values....
[...]