Dense Passage Retrieval
Year: 2,020
Journal: Conference on Empirical Methods in Natural Language Processing
Languages: English
Programming languages: Python
Input data:
question
Output data:
answer
Project website: https://github.com/facebookresearch/DPR
Open-domain question answering relies on efficient passage retrieval to select candidate contexts, where traditional sparse vector space models, such as TF-IDF or BM25, are the de facto method. In this work, we show that retrieval can be practically implemented using dense representations alone, where embeddings are learned from a small number of questions and passages by a simple dualencoder framework.