Alfan Farizki Wicaksono

Buzzer Detection and Sentiment Analysis for Predicting Presidential Election Results in a Twitter Nation

In this paper, we present our approach for predicting the results of Indonesian Presidential Election using Twitter as our main resource. We explore the possibility of easy-togather Twitter data to be utilized as a survey supporting tool to understand public opinion. First, we collected Twitter data during the campaign period. Second, we performed automatic buzzer detection on our Twitter data to remove those tweets generated by computer bots, paid users, and fanatic users that usually become noise in our data. Third, we performed a fine-grained political sentiment analysis to partition each tweet into several sub-tweets and subsequently assigned each sub-tweet with one of the candidates and its sentiment polarity. Finally, to predict the election results, we leveraged the number of positive sub-tweets for each candidate. Our experiment shows that the mean absolute error (MAE) of our Twitter-based prediction is 0.61%, which is surprisingly better than the prediction results published by several independent survey institutions (offline polls). Our study suggests that Twitter can serve as an important resource for any political activity, specifically for predicting the final outcomes of the election itself.

Named entity recognition on Indonesian microblog messages

This paper describes a model to address the task of named-entity recognition on Indonesian microblog messages due to its usefulness for higher-level tasks or text mining applications on Indonesian microblogs. We view our task as a sequence labeling problem using machine learning approach. We also propose various word-level and orthographic features, including the ones that are specific to the Indonesian language. Finally, in our experiment, we compared our model with a baseline model previously proposed for Indonesian formal documents, instead of microblog messages. Our contribution is two-fold: (1) we developed NER tool for Indonesian microblog messages, which was never addressed before, (2) we developed NER corpus containing around 600 Indonesian microblog messages available for future development.

Single-output recurrent neural networks for sentence binary classification

We report several experiments on using Recurrent Neural Networks (RNNs) for sentence binary classification task. In terms of sentence classification, RNNs have an important advantage compared to well-known traditional machine learning models (e.g. SVM and Maximum Entropy), in which it can naturally take into account neighboring information between contiguous words. In addition, to perform binary classification task, we employed Single-Output RNNs (SORNNs) which only consists of a single output layer located in the last time step. The output layer itself is a vector consisting of two units (since we perform binary classification), in which each unit corresponds to a single label. Our results showed that SORNN achieved better performance than other traditional machine learning models, such as SVM, Maximum Entropy, and Naive Bayes, which have been widely used for sentence classification.

A two-stage emotion detection on Indonesian tweets

Emotion is a vital component in various Affective Computing areas such as opinion mining, sentiment analysis, e-learning applications, human-computer interaction and humor recognition. In this paper, we propose a two-stage approach for detecting emotions on Indonesian tweets. In the first stage, we extract emotion-bearing tweets from a huge number of raw tweets. In the second stage, all the extracted tweets are then classified into five well-known pre-defined emotion classes, namely love, joy, sad, fear, and anger. To do that, we devise various features (i.e., linguistic, semantic, and orthographic features) and subsequently use those proposed features to build a computational model based on machine learning approach. Our experimental results show that the proposed method is very effective. It is also worth noting that the work described in this paper is the first work on emotion analysis on Indonesian data.

Detecting Controversial Articles on Citizen Journalism

Someones understanding and stance on a particular controversial topic can be influenced by daily news or articles he consume everyday. Unfortunately, readers usually do not realize that they are reading controversial articles. In this paper, we address the problem of automatically detecting controversial article from citizen journalism media. To solve the problem, we employ a supervised machine learning approach with several hand-crafted features that exploits linguistic information, meta-data of an article, structural information in the commentary section, and sentiment expressed inside the body of an article. The experimental results shows that our proposed method manages to perform the addressed task effectively. The best performance so far is achieved when we use all proposed feature with Logistic Regression as our model (82.89\% in terms of accuracy). Moreover, we found that information from commentary section (structural features) contributes most to the classification task.