Ravichandran M, Kulanthaivel G, Chellatamilan T
Department of Computer Science and Engineering, Sathyabama University, Tamil Nadu 600119, India.
Educational Media Centre, NITTTR, Chennai 600113, India.
ScientificWorldJournal. 2015;2015:617358. doi: 10.1155/2015/617358. Epub 2015 Mar 18.
Every day, huge numbers of instant tweets (messages) are published on Twitter as it is one of the massive social media for e-learners interactions. The options regarding various interesting topics to be studied are discussed among the learners and teachers through the capture of ideal sources in Twitter. The common sentiment behavior towards these topics is received through the massive number of instant messages about them. In this paper, rather than using the opinion polarity of each message relevant to the topic, authors focus on sentence level opinion classification upon using the unsupervised algorithm named bigram item response theory (BIRT). It differs from the traditional classification and document level classification algorithm. The investigation illustrated in this paper is of threefold which are listed as follows: (1) lexicon based sentiment polarity of tweet messages; (2) the bigram cooccurrence relationship using naïve Bayesian; (3) the bigram item response theory (BIRT) on various topics. It has been proposed that a model using item response theory is constructed for topical classification inference. The performance has been improved remarkably using this bigram item response theory when compared with other supervised algorithms. The experiment has been conducted on a real life dataset containing different set of tweets and topics.
每天，大量即时推文（消息）在推特上发布，因为它是电子学习者互动的大型社交媒体之一。学习者和教师通过在推特上获取理想来源，讨论关于各种有趣研究主题的选项。通过大量关于这些主题的即时消息，可以了解对这些主题的普遍情感行为。在本文中，作者没有使用与主题相关的每条消息的观点极性，而是在使用名为二元组项目反应理论（BIRT）的无监督算法时，专注于句子级别的观点分类。它不同于传统的分类和文档级分类算法。本文阐述的研究有三个方面，如下所示：（1）基于词汇的推文消息情感极性；（2）使用朴素贝叶斯的二元组共现关系；（3）关于各种主题的二元组项目反应理论（BIRT）。有人提出构建一个使用项目反应理论的模型用于主题分类推理。与其他监督算法相比，使用这种二元组项目反应理论时性能有了显著提高。该实验是在一个包含不同推文和主题集的真实生活数据集上进行的。
