CUFE at SemEval-2016 Task 4: A Gated Recurrent Model for Sentiment Classification

Abstract

In this paper we describe a deep learning system that has been built for SemEval 2016 Task4 (Subtask A and B). In this work we trained a Gated Recurrent Unit (GRU) neural network model on top of two sets of word embeddings: (a) general word embeddings generated from unsupervised neural language model; and (b) task specific word embeddings generated from supervised neural language model that was trained to classify tweets into positive and negative categories. We also added a method for analyzing and splitting multi-words hashtags and appending them to the tweet body before feeding it to our model. Our models achieved 0.58 F1-measure for Subtask A (ranked 12/34) and 0.679 Recall for Subtask B (ranked 12/19).

Towards a New Deep Learning Algorithm Based on GRU and CNN: NGRU

“…Both systems use word embeddings but with difference, in first CNN system's [4], so the word embeddings with 50M unsupervised to training word2vec [5,6], and 10M supervised to train CNN, but in the second, the GRU with the first word embeddings with 20.5M to train and the second set was obtained by training on supervised data using another GRU model, and add method for splitting hashtags and insert them in the body, before forwarding data in this step [7]. Therefore the amount of training data is very important without forgetting the quality with the treatment of tweets before the passes to the learning phase, CNN has taken a step forward with the amount of data on the other hand there is an equality with the pre-treatment of the text e.g.…”

“…Which it can capture long semantic patterns without tuning the model parameter, unlike CNN models where the model depends on the length of the convolutional feature maps for capturing long patterns, Which it achieved superior performance to CNNs [7]. And the network architecture is composed of a word embeddings layer, a merge layer, dropout layers, a GRU layer, a hyperbolic tangent tang layer and a soft-max classification layer.…”

“…Our architecture Fig. 1 is inspired by GRU [7], with changes in the input data, we use the method Adadelta to the learning rate, and the method Adam for optimization of the weights and the application of NGRU will be on the tweets from different languages.…”

deepSA at SemEval-2017 Task 4: Interpolated Deep Neural Networks for Sentiment Analysis in Twitter

“…The model performs well on three-point scale sentiment classification, while performing poorly on five-point scale sentiment classification. A GRU-based model with two kinds of embedding used for general and taskspecific purpose can be more efficient than CNN models (Nabil et al, 2016 3 Experiment…”

SSN MLRG1 at SemEval-2018 Task 1: Emotion and Sentiment Intensity Detection Using Rule Based Feature Selection

“…Twitter is a huge microblogging service with more than 500 million tweets per day from different locations of the world and in different languages (Saif and Felipe, 2017). Tweets are often used to convey ones emotions, opinions towards products, and stance over issues (Nabil et al, 2016). Automatically detecting emotion intensities in tweets has several applications, including commerce (Jansen et al, 2009), crisis management (Verma et al, 2011), tracking brand and product perception, tracking support for issues and policies, and tracking public health and well-being (Chew and Eysenbach, 2010).…”