The problem is to improve the retrieval performance of an information retrieval system by implementing and evaluating query expansion techniques. The primary challenge is to enhance query coverage without introducing noise, and to evaluate the effectiveness of these expansions using a graded relevance system. The goal is to determine whether query expansion improves the ranking of relevant documents in terms of user satisfaction.
Evaluation Strategy
This project will evaluate the impact of query expansion techniques by comparing the performance of the baseline system (no expansion) with expanded queries. We will implement synonym expansion and contextual expansion to enhance query coverage and use ranking algorithms like BM25 and TF-IDF. The evaluation will focus on metrics like Normalized Discounted Cumulative Gain (NDCG), which is well-suited for graded relevance judgments (2, 1, 0, -1), and Mean Average Precision (MAP) to measure the overall retrieval performance. Traditional metrics like precision, recall, and F1 score will also be used.
Title
Query Expansion and Ranking Evaluation
Team Name
Retrievers
Email
202318002@daiict.ac.in
Team Member 1 Name
Bhavik Manwani
Team Member 1 Id
202318002
Team Member 2 Name
Manthan Solanki
Team Member 2 Id
202318002
Team Member 3 Name
Anmol Poonia
Team Member 3 Id
202318009
Team Member 4 Name
Kanishk Pareek
Team Member 4 Id
202101134
Category
Evaluation Track Problem
Problem Statement
The problem is to improve the retrieval performance of an information retrieval system by implementing and evaluating query expansion techniques. The primary challenge is to enhance query coverage without introducing noise, and to evaluate the effectiveness of these expansions using a graded relevance system. The goal is to determine whether query expansion improves the ranking of relevant documents in terms of user satisfaction.
Evaluation Strategy
This project will evaluate the impact of query expansion techniques by comparing the performance of the baseline system (no expansion) with expanded queries. We will implement synonym expansion and contextual expansion to enhance query coverage and use ranking algorithms like BM25 and TF-IDF. The evaluation will focus on metrics like Normalized Discounted Cumulative Gain (NDCG), which is well-suited for graded relevance judgments (2, 1, 0, -1), and Mean Average Precision (MAP) to measure the overall retrieval performance. Traditional metrics like precision, recall, and F1 score will also be used.
Dataset
https://huggingface.co/datasets/nreimers/trec-covid
Resources
Title: Learning to rank query expansion terms for COVID-19 scholarly search Link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10174726/