Ekin Ekinci

An Ensemble Model using a BabelNet Enriched Document Space for Twitter Sentiment Classification

With the widespread usage of social media in our daily lives, user reviews emerged as an impactful factor for numerous fields including understanding consumer attitudes, determining political tendency, revealing strengths or weaknesses of many different organizations. Today, people are chatting with their friends, carrying out social relations, shopping and following many current events through the social media. However social media limits the size of user messages. The users generally express their opinions by using emoticons, abbreviations, slangs, and symbols instead of words. This situation makes the sentiment classification of social media texts more complex. In this paper a sentiment classification model for Twitter messages is proposed to overcome this difficulty. In the proposed model first the short messages are expanded with BabelNet which is a concept network. Then the expanded and the original form of the messages are included in an ensemble learning model. Consequently we compared our ensemble model with traditional classification algorithms and observed that the F-measure value is increased.

An annotated corpus for Turkish sentiment analysis at sentence level

With the rapid growth of unstructured data accessible via web, managing these data and finding undiscovered information in huge dataset become a necessary task. Consequently text mining, which can be defined as gleaning important information from natural language text, has emerged. In this study, in order to facilitate information management for aspect based sentiment analysis studies, a Turkish sentiment corpus, which is comprised of user reviews and is annotated semi-automatically, is constructed. In the constructed corpus, the root form of the words, the usage (aspect/multi-aspect/seedsentiment/absent) of these words, Part of Speech (POS) tags and their polarities are defined. Turkish hotel review dataset which contains 1000 reviews and 5364 sentences for this study was crawled from a web source. The system takes reviews, aspect and seedsentiment lists and returns JSON data structures of the annotated corpus. In this paper, both we provide a ready to use dataset for developing aspect based sentiment analysis applications and we make this dataset easy to use for Java applications by creating JSON data.

A Novel Method for Extracting Feature Opinion Pairs for Turkish

Reviews made by online users, are one of the most important sources for consumers who give importance them during decision process and for companies which benefit from them during development process. Since internet has become a part of our daily lives, the number of reviews expands; it is getting difficult day by day to obtain a comprehensive view of user opinions from these reviews manually. Thus, sentiment analysis becomes an indispensable task for analyzing user reviews automatically. Recently, feature-based opinion mining methods are gaining importance in terms of fine-grained sentiment analysis. In this paper, we propose a Push Down Automata (PDA) based Feature-Opinion Pair (FOP) extraction for Turkish hotel reviews. At first, context free grammars are proposed by using Turkish linguistic relations then PDA is applied for extracting FOPs. Experimental results are showed that the proposed approach provides an efficient solution for discovering accurate FOPs.

Using authorship analysis techniques in forensic analysis of electronic mails

As a result of rapid advances in information technology electronic mail has become one of todays most important communication tool. Electronic mail which provides conveniences to its user in many cases, is also an attractive environment for criminals. Malicious electronic mail whose actual owner is uncertain, is taking place in cyber crimes and authorship analysis has become necessary for determining the actual owner of this electronic mail. In this study 43 textual features were extracted from dataset of electronic mails which is obtained for 5 writers. These extracted textual features were processed with Artifical Neural Network (ANN), Support Vector Machines (SVM) and Decission Trees Method that are method of data mining classfication techniques in WEKA. As a result of the application. Decision Trees Method has been observed to be most succesful with F-measure rate of 83% in average for available dataset.