Emoticon Smoothed Language Models for Twitter Sentiment Analysis

By Kun Liu, Wu-Jun Li, Minyi Guo. 2012

285 citations recorded by the source corpus.

Twitter sentiment analysis (TSA) has become a hot research topic in recent years. The goal of this task is to discover the attitude or opinion of the tweets, which is typically formulated as a machine learning based text classification problem. Some methods use manually labeled data to train fully supervised models, while others use some noisy labels, such as emoticons and hashtags, for model training. In general, we can only get a limited number of training data for the fully supervised models because it is very labor-intensive and time-consuming to manually label the tweets. As for the models with noisy labels, it is hard for them to achieve satisfactory performance due to the noise in the labels although it is easy to get a large amount of data for training. Hence, the best strategy is to utilize both manually labeled data and noisy labeled data for training. However, how to seamlessly integrate these two different kinds of data into the same learning framework is still a challenge. In this paper, we present a novel model, called emoticon smoothed language model (ESLAM), to handle this challenge. The basic idea is to train a language model based on the manually labeled data, and then use the noisy emoticon data for smoothing. Experiments on real data sets demonstrate that ESLAM can effectively integrate both kinds of data to outperform those methods using only one of them.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Ting-Wei Liu
2. Bing Qin
3. Duyu Tang
4. Furu Wei
5. M. Zhou
6. Erik Cambria
7. Jiliang Tang
8. Fabio A. Crestani

Citations in the corpus

• "Blissfully Happy" or "Ready toFight": Varying Interpretations of Emoji (2016, background): “Researchers such as Liu, Li, and Guo (2012) and Novak et al. (2015) have developed classifiers of emoji sentiment by labeling emoji with the sentiment of the surrounding text.”
• Distantly Supervised Lifelong Learning for Large-Scale Social Media Sentiment Analysis (2017, methodology): “[18] presented a model called emoticon smoothed language model (ESLAM) to integrate manually labeled data and noisy labeled data into one framework.”
• Unsupervised sentiment analysis with emotional signals (2013, background): “use emoticons, they are effectively marking up the text with an emotional state [22].”
• Twitter Sentiment Analysis (2015, background): “In recent years the role of emoticons has been investigated (Pozzi et al., 2013a; Hogenboom et al., 2013; Liu et al., 2012; Zhao et al., 2012).”
• Twitter brand sentiment analysis: A hybrid system using n-gram analysis and dynamic artificial neural network (2013, background): “…have also evaluated features that are more specific to the domain, including acronyms (Agarwal et al., 2011; Barbosa & Feng, 2010; Kouloumpis et al., 2011) and emoticons (Agarwal et al., 2011; Jiang et al., 2011; Kouloumpis et al., 2011; Liu, Li, & Guo, 2012; Zhang, Ghosh, Dekhil, Hsu, & Liu, 2011).”
• Learning Sentiment-Specific Word Embedding for Twitter Sentiment Classification (2014, methodology): “Instead of directly using the distantsupervised data as training set, Liu et al. (2012) adopt the tweets with emoticons to smooth the language model and Hu et al. (2013) incorporate the emotional signals into an unsupervised learning framework for Twitter sentiment classification.”
• Sentiment Embeddings with Applications to Sentiment Analysis (2016, background): “How to automatically collecting large-scale and robust tweets with neutral sentiment remains as an open problem, even though there are some preliminary attempts [55], [56].”
• Evaluation Datasets for Twitter Sentiment Analysis: A survey and a new dataset, the STS-Gold (2013, methodology): “The dataset has been used in [3,12,5] for polarity and subjectivity classification of tweets.”
• Predicting Polarities of Tweets by Composing Word Embeddings with Long Short-Term Memory (2015, background): “While some other works utilize the microblogging features uncommon in the formal literature, such as hashtags, emoticons (Hu et al., 2013a; Liu et al., 2012).”
• Contextual Bidirectional Long Short-Term Memory Recurrent Neural Network Language Models: A Generative Approach to Sentiment Analysis (2017, methodology): “Liu et al. (2012) have employed an emoticon smoothed unigram LM to perform sentiment classification.”
• Predictive Analysis on Twitter: Techniques and Applications (2018, background): “, Txt, [18] F1=0.67 Emotion Detection RT, REPTree, TwM [22] F1=˘0.60 BNet, LogR and UsM [29] F0.5=0.62 [21] Acc=0.82 [6] Acc=0.61 [139] Mean error=0.071 [140] Acc=˘0.71 [141] F0.5= 0.76 [64] F1=˘0.73 [26] F1=0.79 [101] Acc=˘0.83 Location Estimation NB, SVM, DT, Ling, TwM, UsM [142] Mean error=0.39 L”
• A Systematic Review of Emoji: Current Research and Future Perspectives (2019, methodology): “An example of supervised learning is the emoticon smoothed language model (ESLAM) proposed by Liu et al. (2012), which classifies twitter based on a model trained by a human annotated corpus.”