University of Arkansas at Little Rock (UALR) participated for the first time this year in Text REtrieval Conference (TREC) organized by National Institute of Standards and Technology (NIST). Information Retrieval (IR) group at UALR applied Science department participated in the Blog Track. TREC consists of different tracks and each track has unique challenge in Information Retrieval. Blog track particularly focuses on the peculiar type of challenges associated with Blogs. Blogs are online web diaries written by millions of people around the world. What is interesting about blogs is its representation of people├óÔéČÔäós opinions about some event, something or someone in particular. Although not all blogs are opinionated, most blogs are personal and hence likely to be opinionated.
├é ├é ├é ├é ├é ├é ├é ├é ├é ├é ├é The challenge this year in Blog track was to determine opinionated blogs given a particular topic. Target topic could be name of person, event, product or even organization/company. Neilson Buzz Metric has been a popular business where they generate popularity ratings of various TV shows, political polls and various issues of importance. Over 3 million blogs were provided the challenge was to identify and rank those which are most opinionated about say ├óÔéČ┼ôipod├óÔéČ┬Ł.
├é ├é ├é ├é ├é ├é ├é ├é ├é ├é ├é Twenty eight teams from all over the world participated this year in Blog Track and only 14 teams could submit final results. One of the reasons why it is so difficult is because, each team uses different approach to solve the given problem and in a given time of around 2 months during summer. It was long sweaty nights and days of summer but in the end it was all worth it when we could qualify and submit 5 different runs. We used 5 different approaches to rank the most relevant and opinionated blogs about the topic. Each team is allowed maximum of five runs.├é To give you an idea, some of the participants along with UALR are John Hopkins University, Carnegie Mellon University, CSIRO ICT Centre in Australia, Univ. of Maryland (college park), Indiana University, Univ. of Pisa in Italia, University of Amsterdam, NEC labs, Chinese academy of Sciences, University of Chicago and so on.
├é ├é ├é ├é ├é ├é ├é ├é ├é ├é ├é For the opinion retrieval task (rank blog-posts by the most opinionated first), the median Mean Average Precision (MAP) is 0.1059 and UALR-IR group achieved performance of 0.0715 beating MAP scores by Chinese Academy of Sciences, National Institute of Informatics and Robert Gordon University in UK. Considering this was the first year for UALR participation, we feel satisfied with our performance.
├é ├é ├é ├é ├é ├é ├é ├é ├é ├é ├é Among all participants, only 5 teams get a chance to present their work at the actual conference. We submitted speaker proposal and were selected to present our work and results with the Information Retrieval community on November 16th, 2006. In fact, we were the first team to present research work for the blog track. The travel and data acquisition costs were covered by Graduate Dean├óÔéČÔäós Office, Department of Applied Science, and Dr. Bayrak├óÔéČÔäós grants.
A Special Note from the Student
For me as a student, it was a wonderful experience to be able to participate and represent our school at international level. The event was very stimulation and motivational. I have learned a lot and made many new friends and met with lot of colleagues at the conference venue. I have been able to exchange ideas with the research scholars from around the world and I feel proud to say ├óÔéČ┼ôWe did it├óÔéČ┬Ł. ├óÔéČÔÇť Hemant Joshi, PhD candidate, Applied Science