Onur Kucuktunc - Research

Citation Recommendation We developed a method, direction-aware random walk with restart, to suggest relevant papers using the relations in citation graph. Using the metadata harvested mainly from open-access databases, we implemented a scalable web interface of the service, accessible via theadvisor.
Diversity Search We investigate the concepts of diversity and diverse nearest neighbor search in various fields and application areas with different types of data, such spatial, high-dimensional, and time-series data. Although the definition of diversity depends on the field, we propose techniques without neglecting the general objective of diversification, which is to maximize the similarity of search results to the query by minimizing pairwise similarities between the results.
Biclustering We address this problem of evaluating the strengths and weaknesses of existing biclustering methods for gene expression data that were recently published or have not been extensively studied.
Sentiment Analysis Our analysis starts with a large-scale study on the correlation of various features with the observed attitude and sentimentality. We investigate the correlation of sentiments with textual, topical, demographical, spatial, and temporal features.

Video Copy Detection Content-based copy detection task for TRECVID 2008, proposed a video copy detection framework that detects copy segments by fusing the results of three different techniques: facial shot matching, activity subsequence matching, and non-facial shot matching using low-level features.
Tag Suggestion Computer vision-based tag recommendation for photo sharing websites, such as Flickr.
NLP Quering We introduced a natural language-based query language for a multimedia database, and integrated the NLP part with the web-based visual query interface.
Animation We presented an efficient method for the simulation of boiling water by modeling bubbles and waves as particles, and adding a grid-based fluid surface.