Tweester at SemEval-2016 Task 4: Sentiment Analysis in Twitter Using Semantic-Affective Model Adaptation

Abstract

We describe our submission to SemEval2016 Task 4: Sentiment Analysis in Twitter. The proposed system ranked first for the subtask B. Our system comprises of multiple independent models such as neural networks, semantic-affective models and topic modeling that are combined in a probabilistic way. The novelty of the system is the employment of a topic modeling approach in order to adapt the semantic-affective space for each tweet. In addition, significant enhancements were made in the main system dealing with the data preprocessing and feature extraction including the employment of word embeddings. Each model is used to predict a tweet's sentiment (positive, negative or neutral) and a late fusion scheme is adopted for the final decision.

NileTMRG at SemEval-2017 Task 4: Arabic Sentiment Analysis

“…They trained the SVM using hidden CNN activations with additional linguistic information. The "Tweester" team (Palogiannidi et al 2016) used multiple independent classifiers: neural networks, semantic-affective models and topic modeling. Each classifier predicts the result and late fusion is done to generate the final result.…”

Tweester at SemEval-2017 Task 4: Fusion of Semantic-Affective and pairwise classification models for sentiment analysis in Twitter

“…The submitted system is based on the fusion of several systems. Specifically the system consists of: 1) the semantic-affective system (submitted to the SemEval 2016 Task 4 (Palogiannidi et al, 2016)) that incorporates affective, semanticsimilarity, sarcasm/irony and topic modeling features, 2) a single and a two-step convolutional neural network, 3) a system based on word embeddings, 4) a "stacking" based system that transforms the 3-class polarity problem of Subtask A, into 2-class binary problems and finally 5) the open-source system submitted to the SemEval 2015 Task 10 ( Rosenthal et al, 2015a).…”

“…We hypothesize that hashtags are able to express user's sentiment with regard to some topic or events (e.g., "Jazz all day #lovemusic"). Following this assumption, hashtag expansion into word strings (Palogiannidi et al, 2016) was performed using the Viterbi algorithm (Forney, 1973). The absolute and relative frequencies of hashtags to be expanded are used as features, as well as the binary indicators that a tweet contains hashtags that require expansion.…”

“…The challenging nature of sentiment analysis in Twitter motivated the organization of numerous tasks within the Semantics Evaluation (SemEval) workshop. In this paper, we show how our sentiment analysis framework called "Tweester" (winner of Subtask B in SemEval-2016 (Palogiannidi et al, 2016)), can be applied to all subtasks, namely Subtask A (message polarity classification), Subtask B (tweet classification according to a two-point scale), Subtask C (sentiment conveyed by a tweet towards the topic on a five-point scale), Subtask D (estimate the distribution of the tweets across a twopoint scale), Subtask E (estimate the distribution of the tweets across a five-point scale ). The system achieved high performance ranking 5th, 3rd, 4th, 5th and 8th for Subtasks A, B, C, D, and E, respectively (S. Rosenthal et al, 2017).…”

DataStories at SemEval-2017 Task 4: Deep LSTM with Attention for Message-level and Topic-based Sentiment Analysis

“…In previous runs of the Task, sentiment analysis was usually tackled using hand-crafted features and/or sentiment lexicons (Mohammad et al, 2013;Kiritchenko et al, 2014;Palogiannidi et al, 2016), feeding them to classifiers such as Naive Bayes or Support Vector Machines (SVM). These approaches require a laborious 1 github.com/cbaziotis/ekphrasis feature-engineering process, which may also need domain-specific knowledge, usually resulting both in redundant and missing features.…”