Completed

Digital Humanities is an area where computing and humanistic diciplines intersect. In this project, we intersect literature with computing by visualizing Nazım Hikmet’s poetry. The project is intended to notice the distinctive parameters of style of Nazım Hikmet, examine the change of his style among his works by applying text-mining methodologies to analyze  and show the results of our analysis by using an interactive visualizations. Our goal is developing an interactive tool where users can see various visualization reated using Nazım’s works.

Time series analysis is used to estimate and predict behaviour of time dependent processes. In this project we have made use of two stochastic volatility models: Univariate Stochastic Volatility Model, and Multivariate Stochastic Volatility via Wishart Distribution. We estimated their parameters using combinations of MCMC methods, which are mainly Metropolis-Hastings Algorithm and Gibbs sampler. The work done gave us tools to estimate time dependent variance structure of the time series.

Sentiment Analysis which is part of Natural Language Processing, means making text analysis and computational linguistics to identify and extract subjective information in source materials. Sentiment analysis is the most common technique for natural language processing in order to understand human languages. In spite of the fact that sentiment analysis has many sections, we interest on “Sentiment Prediction” by classifying the words as “positive”, “negative” or “neutral”. This is the main aim of our project.

In this paper, we present a system to analyze consumer be- havior and cluster the customers accordingly. It is based on matrix and tensor factorization techniques and k-means clus- tering algorithm. Factorizing the matrix that contains cus- tomers and the items they purchased, we get a relatively low dimensional latent factors for users. Then this factors are used to cluster the customers. We used this system on the Obase data set that contains past sales data and customer demographics for a major retail store.

We address the problem of Part of Speech tagging for noisy texts in Turkish language. The popular social media network Twitter has gained so much attention in Turkey recent years and for this reason we considered Twitter as a perfect candidate for noisy text source for Turkish language and consequently have decided to supply our data set from Twitter. We used Hidden Markov Models (HMMs) for modeling our problem.We have created a data set in order to train and test our model which constitutes around 500 tweets.