{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/learning-dense-representations-of-phrases-at","title":"Learning Dense Representations of Phrases at Scale","arxiv_id":"2012.12624","date":"2020-12-23","proceeding":"ACL 2021 5","authors":["Jinhyuk Lee","Mujeen Sung","Jaewoo Kang","Danqi Chen"],"abstract":"Open-domain question answering can be reformulated as a phrase retrieval problem, without the need for processing documents on-demand during inference (Seo et al., 2019). However, current phrase retrieval models heavily depend on sparse representations and still underperform retriever-reader approaches. In this work, we show for the first time that we can learn dense representations of phrases alone that achieve much stronger performance in open-domain QA. We present an effective method to learn phrase representations from the supervision of reading comprehension tasks, coupled with novel negative sampling methods. We also propose a query-side fine-tuning strategy, which can support transfer learning and reduce the discrepancy between training and inference. On five popular open-domain QA datasets, our model DensePhrases improves over previous phrase retrieval models by 15%-25% absolute accuracy and matches the performance of state-of-the-art retriever-reader models. Our model is easy to parallelize due to pure dense representations and processes more than 10 questions per second on CPUs. Finally, we directly use our pre-indexed dense phrase representations for two slot filling tasks, showing the promise of utilizing DensePhrases as a dense knowledge base for downstream tasks.","url_abs":"https://arxiv.org/abs/2012.12624v3","url_pdf":"https://arxiv.org/pdf/2012.12624v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"learning-dense-representations-of-phrases-at","repo_url":"https://github.com/jhyuklee/DensePhrases","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"learning-dense-representations-of-phrases-at","repo_url":"https://github.com/princeton-nlp/DensePhrases","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"learning-dense-representations-of-phrases-at","repo_url":"https://github.com/dmis-lab/gener","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"learning-dense-representations-of-phrases-at","repo_url":"https://github.com/princeton-nlp/SimCSE","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"open-domain-question-answering","task_name":"Open-Domain Question Answering"},{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"question-generation","task_name":"Question Generation"},{"task_slug":"reading-comprehension","task_name":"Reading Comprehension"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"slot-filling","task_name":"Slot Filling"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/question-answering-on-natural-questions-long","task":"Question Answering","dataset":"Natural Questions (long)","model":"DensePhrases","rank_in_archive_order":1,"of":13,"metrics":{"EM":"71.9","F1":"79.6"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-squad11-dev","task":"Question Answering","dataset":"SQuAD1.1 dev","model":"DensePhrases","rank_in_archive_order":18,"of":55,"metrics":{"EM":"78.3","F1":"86.3"},"uses_additional_data":false},{"leaderboard":"/sota/slot-filling-on-kilt-t-rex","task":"Slot Filling","dataset":"KILT: T-REx","model":"DensePhrases","rank_in_archive_order":8,"of":20,"metrics":{"Accuracy":"53.9","F1":"61.74","KILT-AC":"27.84","KILT-F1":"32.34","R-Prec":"37.62","Recall@5":"40.07"},"uses_additional_data":false},{"leaderboard":"/sota/slot-filling-on-kilt-zero-shot-re","task":"Slot Filling","dataset":"KILT: Zero Shot RE","model":"DensePhrases","rank_in_archive_order":7,"of":21,"metrics":{"Accuracy":"47.42","F1":"54.75","KILT-AC":"41.34","KILT-F1":"46.79","R-Prec":"57.43","Recall@5":"60.47"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2012.12624","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}