DataFinder: Scientific Dataset Recommendation from Natural Language Descriptions

TL;DR

DataFinder introduces a natural language-based dataset recommendation system that uses a new dataset and a bi-encoder model to improve search relevance for researchers seeking suitable datasets.

Contribution

The paper presents the DataFinder Dataset and a novel bi-encoder retrieval method for recommending datasets from natural language descriptions, outperforming existing search engines.

Findings

Bi-encoder retriever outperforms baselines in relevance.

DataFinder Dataset includes 17.5K queries for training.

System effectively matches datasets to research needs.

Abstract

Modern machine learning relies on datasets to develop and validate research ideas. Given the growth of publicly available data, finding the right dataset to use is increasingly difficult. Any research question imposes explicit and implicit constraints on how well a given dataset will enable researchers to answer this question, such as dataset size, modality, and domain. We operationalize the task of recommending datasets given a short natural language description of a research idea, to help people find relevant datasets for their needs. Dataset recommendation poses unique challenges as an information retrieval problem; datasets are hard to directly index for search and there are no corpora readily available for this task. To facilitate this task, we build the DataFinder Dataset which consists of a larger automatically-constructed training set (17.5K queries) and a smaller expert-annotated evaluation set (392 queries). Using this data, we compare various information retrieval algorithms on our test set and present a superior bi-encoder retriever for text-based dataset recommendation. This system, trained on the DataFinder Dataset, finds more relevant search results than existing third-party dataset search engines. To encourage progress on dataset recommendation, we release our dataset and models to the public.