jump to: about | research | grapevine | publications | other
I am currently a PhD Candidate at the Department of Computer Science at the University of Toronto, working with Prof. Nick Koudas.
My main research interests lie in the intersection of Database Systems and Information Retrieval. I am partial to problems dealing with ranking structured information, in the context of user generated content, such as:
Using services such as blogs, micro-blogs, and online social-networks, people around the globe generate millions of documents (posts, articles, status updates, etc.) every day.
This user-generated content is more than just a massive stream of text. On many dimensions, it embeds a significant amount of structural information, such as social links, mentions of real-world entities (people, locations, products, etc.), user profile information, etc., giving rise to large graph structures.
Exploiting this structure, as opposed to considering the textual aspect of user-generated content alone, enables valuable applications, as diverse as:
and so on. At the core of such computations, is the necessity of efficiently and effectively prioritizing/ranking information.
My research focuses on devising efficient algorithms to enable such computations, i.e. ranking in the context of user-generated content.
Existing algorithmic techniques, such as the Threshold Algorithm, can be adapted to the structured setting of user-generated content, by interleaving richer computations into the algorithm's core.
At the same time, traditional assumptions need to be re-examined, such as how to schedule data accesses for optimal performance.
In my work, I strive for balance between designing novel algorithms, analytically exploring their properties, and experimentally validating them.
Grapevine is a system that conducts large scale data analysis on the social media collective, extracting information in real time. The goal of Grapevine is to distill information and provide insights, by capturing popular trends as they emerge.
Grapevine facilitates the interactive exploration of content, allowing users to discover interesting or surprising stories, optionally narrowed down on a specific demographic of interest (e.g. "What are Torontonian teens talking about on blogs?"). Stories of interest can be explored in a variety of ways, such as modifying their scope (e.g. "How is Barack Obama related to this story?"), viewing related content (blog posts, news, videos, etc.), and examining their temporal evolution.
Grapevine, currently formerly live at www.onthegrapevine.ca, was developed in collaboration with fellow graduate student Nikos Sarkas. Supporting this functionality has led us to consider exciting research questions, such as
Chronological list of publications (see also DBLP)
Dense Subgraph Maintenance under Streaming Edge Weight Updates for Real-time Story Identification [Paper | Tech. Report], in VLDB 2012
, Nick Koudas, Nikos Sarkas, Divesh Srivastava
Efficient Diversity-Aware Search [Paper | Tech. Report], in SIGMOD 2011
, Nick Koudas
Efficient Identification of Coupled Entities in Document Collections, in ICDE 2010
Nikos Sarkas, , Nick Koudas, Divesh Srivastava
Ranking Objects Based on Relationships and Fixed Associations [Paper | Tech. Report ], in EDBT 2009
, Surajit Chaudhuri, Gautam Das, Nick Koudas
What's on the Grapevine ? [Demo Paper], in SIGMOD 2009
, Nick Koudas, Nikos Sarkas, Divesh Srivastava
Qualitative Geocoding of Persistent Web Pages [Paper], in ACM GIS 2008
, Alexandros Efentakis, Chara Lontou, Dieter Pfoser
More coming soon... (currently under submission)
MSc Research Paper (in lieu of MSc Thesis), University of Toronto
Supervisor: Prof. Nick Koudas
Available upon request (currently under submission)
Geographic Information Extraction from Text [abstract | full text (in Greek)]
Diploma Thesis, National Technical University of Athens
Supervisor: Prof. Timos Sellis. Co-supervisor: Dieter Pfoser
