Skip to content

Latest commit

 

History

History
13 lines (6 loc) · 870 Bytes

README.md

File metadata and controls

13 lines (6 loc) · 870 Bytes

Task Overview

This repository focuses on the task of Dataset Search for the NTCIR-16 Conference. Given a query and a dataset collection, the goal is to retrieve a ranked list of datasets. We perform experimental evaluations by using document retrieval approaches, sentence encoders and language models such as BERT. For that, we use the public data of the NTCIR-16 competition: a shared task for ad-hoc dataset retrieval. As follows, we present the main results in terms of NDCG and recall at several cutoffs for each search methodology.

Retrieval Results

Screen Shot 2022-02-09 at 19 39 03

Ranking Results

Screen Shot 2022-02-09 at 19 36 01