1. What are the contributions mentioned in the paper "Predicting query performance by query-drift estimation" ?
Their novel approach to addressing this challenge is based on estimating the potential amount of query drift in the result list, i. e., the presence ( and dominance ) of aspects or topics not related to the query in top-retrieved documents.. Empirical evaluation demonstrates the prediction effectiveness of their approach for several retrieval models.. The authors argue that query-drift can potentially be estimated by measuring the diversity ( e. g., standard deviation ) of the retrieval scores of these documents.
read more
2. What is the effective method for predicting the performance of a query?
The most effective prediction approaches employ postretrieval analysis of the result list — the documents most highly ranked in response to the query.
read more
3. What is the main argument for this paper?
using insights from work on pseudo-feedback-based query expansion [2] the authors argue that high standard deviation of retrieval scores in the result list correlates with reduced query-drift, and consequently, with improved effectiveness.
read more
4. How can NQC predict the performance of a query?
if NQC is employed for TREC123 over a much larger result-list, then prediction success can improve up to a Pearson correlation of 0.7; the same holds for WIG.
read more
![Fig. 1. Geometric interpretation of NQC. The two leftmost graphs present retrievalscores curves for “difficult” and “easy” queries chosen by average-precision (AP) performance (query-likelihood model [20], ROBUST benchmark). Right: the shift between these two scenarios amounts to clockwise rotation of the retrieval-scores line.](/figures/fig-1-geometric-interpretation-of-nqc-the-two-leftmost-3ri3pd2z.webp)
