Katia Lida Kermanidis

A comparison between semi-supervised and supervised text mining techniques on detecting irony in greek political tweets

The present work describes a classification schema for irony detection in Greek political tweets. Our hypothesis states that humorous political tweets could predict actual election results. The irony detection concept is based on subjective perceptions, so only relying on human-annotator driven labor might not be the best route. The proposed approach relies on limited labeled training data, thus a semi-supervised approach is followed, where collective-learning algorithms take both labeled and unlabeled data into consideration. We compare the semi-supervised results with the supervised ones from a previous research of ours. The hypothesis is evaluated via a correlation study between the irony that a party receives on Twitter, its respective actual election results during the Greek parliamentary elections of May 2012, and the difference between these results and the ones of the preceding elections of 2009.

Mood Classification Using Lyrics and Audio: A Case-Study in Greek Music

This paper presents a case-study of the effectiveness of a trained system into classifying Greek songs according to their audio characteristics or/and their lyrics into moods. We examine how the usage of different algorithms, featureset combinations and pre-processing parameters affect the precision and recall percentages of the classification process for each mood model characteristic. Experimental results indicate that the current selection of features offers accuracy results, the superiority of lyrics content over generic audio features as well as potential caveats with current research in Greek language stemming pre-processing methods.

Learning automatic acquisition of subcategorization frames using Bayesian inference and support vector machines

Learning Bayesian belief networks (BBN) from corpora and support vector machines (SVM) have been applied to the automatic acquisition of verb subcategorization frames for Modern Greek. We are incorporating minimal linguistic resources, i.e. basic morphological tagging and phrase chunking, to demonstrate that verb subcategorization, which is of great significance for developing robust natural language human computer interaction systems, could be achieved using large corpora, without having any general-purpose, syntactic parser at all. In addition, apart from BBN and SVM, which have not previously used for this task, we have experimented with three well-known machine learning methods (feedforward backpropagation neural networks, learning vector quantization and decision tables), which are also being applied to the task of verb subcategorization frame defection for the first time. We argue that both BBN and SVM are well suited for learning to identify verb subcategorization frames. Empirical results will support this claim. Performance has been methodically evaluated using two different corpora types, one balanced and one domain-specific in order to determine the unbiased behaviour of the trained models. Limited training data are proved to endow with satisfactory results. We have been able to achieve precision exceeding 80% on the identification of subcategorization frames which were not known beforehand.

Success Is Hidden in the Students' Data

The contribution of data mining to education as well as research in this area is done on a variety of levels and can affect the instructors’ approach to learning. This particular study focuses on problems associated with classification and attribute selection. An effort to forecast the results takes place before the educational process ends in order to prevent a potential learning failure.

Automatic Speech Recognition System for Home Appliances Control

In the present work we study the performance of a speech recognizer for the Greek language, in a smart-home environment. This recognizer operates in spoken interaction scenarios, where the users are able to control various home appliances. In contrast to command and control systems, in our application the users speak spontaneously, beyond the use of a standardized set of isolated commands. The operational performance was tested over various environmental conditions, for two different types of microphones. In all experiments, regardless of the difference in the word error rates obtained for different scenarios, a task completion rate of 100% was observed.

Political sentiment analysis of tweets before and after the Greek elections of May 2012

Political sentiment analysis using social media content has attracted significant interest in the last years. The present work utilises limited linguistic resources in a knowledge-poor approach to political sentiment identification from Greek tweets, posted a few days before and a few days after the Greek parliamentary elections of May 6th, 2012. Rather than the accurate prediction of the election results, the main focus of the proposed work is the investigation of the alignment between web and actual political sentiment in a bi-directional manner: the impact/reflection of the tweets’ sentiment on the elections as well as the impact of the elections’ results on web sentiment, and its shift around such a major political event.

Combining LSTM and Feed Forward Neural Networks for Conditional Rhythm Composition

Algorithmic music composition has long been in the spotlight of music information research and Long Short-Term Memory (LSTM) neural networks have been extensively used for this task. However, despite LSTM networks having proven useful in learning sequences, no methodology has been proposed for learning sequences conditional to constraints, such as given metrical structure or a given bass line. In this paper we examine the task of conditional rhythm generation of drum sequences with Neural Networks. The proposed network architecture is a combination of LSTM and feed forward (conditional) layers capable of learning long drum sequences, under constraints imposed by metrical rhythm information and a given bass sequence. The results indicate that the role of the conditional layer in the proposed architecture is crucial for creating diverse drum sequences under conditions concerning given metrical information and bass lines.

Detecting Irony on Greek Political Tweets: A Text Mining Approach

The present work describes the classification schema for irony detection in Greek political tweets. The proposed approach relies on limited labeled training data, and its performance on a larger unlabeled dataset is evaluated qualitatively (implicitly) via a correlation study between the irony that a party receives on Twitter, its respective actual election results during the Greek parliamentary elections of May 2012, and the difference between these results and the ones of the preceding elections of 2009. The machine learning results on the labeled dataset were highly encouraging and uncovered a trend whereby the volume of ironic tweets can predict the fluctuation from previous elections.

The Greek Audio Dataset

The Greek Audio Dataset (GAD), is a freely available collection of audio features and metadata for a thousand popular Greek tracks. In this work, the creation process of the dataset is described together with its contents. Following the methodology of existing datasets, the GAD dataset does not include the audio content of the respective data due to intellectual property rights but it includes MIR important features extracted directly from the content in addition to lyrics and manually annotated genre and mood for each audio track. Moreover, for each track a link to available audio content in YouTube is provided in order to support researchers that require the extraction of new feature-sets, not included in the GAD. The selection of the features extracted has been based on the Million Song Dataset in order to ensure that researchers do not require new programming interfaces in order to take advantage of the GAD.

Studying massive open online courses: recommendation in social media

This paper is a research approach linking Massively Open Online Courses (MOOCs) data from Coursera with social media. More specifically it examines information through web data that have been retrieved from the MOOCs Information Pages of the company. These pages can be recommended by the site visitors of the social networks and be shared by them. Our studys purpose is to trace those attributes of these pages that are the most valuable for the visitors recommendation procedure. Moreover, we study the correlation among the three largest social networks, Facebook, Google+ and Twitter, based on the data of the Information Pages. Finally, we present interesting statistics concerning the course categories and the Universities that participate.