%0 Generic %1 liang2020embeddingbased %A Liang, Davis %A Xu, Peng %A Shakeri, Siamak %A Santos, Cicero Nogueira dos %A Nallapati, Ramesh %A Huang, Zhiheng %A Xiang, Bing %D 2020 %K masterthesis passage-retrieval word-embeddings %T Embedding-based Zero-shot Retrieval through Query Generation %U http://arxiv.org/abs/2009.10270 %X Passage retrieval addresses the problem of locating relevant passages, usually from a large corpus, given a query. In practice, lexical term-matching algorithms like BM25 are popular choices for retrieval owing to their efficiency. However, term-based matching algorithms often miss relevant passages that have no lexical overlap with the query and cannot be finetuned to downstream datasets. In this work, we consider the embedding-based two-tower architecture as our neural retrieval model. Since labeled data can be scarce and because neural retrieval models require vast amounts of data to train, we propose a novel method for generating synthetic training data for retrieval. Our system produces remarkable results, significantly outperforming BM25 on 5 out of 6 datasets tested, by an average of 2.45 points for Recall@1. In some cases, our model trained on synthetic data can even outperform the same model trained on real data

@misc{liang2020embeddingbased, abstract = {Passage retrieval addresses the problem of locating relevant passages, usually from a large corpus, given a query. In practice, lexical term-matching algorithms like BM25 are popular choices for retrieval owing to their efficiency. However, term-based matching algorithms often miss relevant passages that have no lexical overlap with the query and cannot be finetuned to downstream datasets. In this work, we consider the embedding-based two-tower architecture as our neural retrieval model. Since labeled data can be scarce and because neural retrieval models require vast amounts of data to train, we propose a novel method for generating synthetic training data for retrieval. Our system produces remarkable results, significantly outperforming BM25 on 5 out of 6 datasets tested, by an average of 2.45 points for Recall@1. In some cases, our model trained on synthetic data can even outperform the same model trained on real data}, added-at = {2021-01-07T16:06:45.000+0100}, author = {Liang, Davis and Xu, Peng and Shakeri, Siamak and Santos, Cicero Nogueira dos and Nallapati, Ramesh and Huang, Zhiheng and Xiang, Bing}, biburl = {https://www.bibsonomy.org/bibtex/25c53370740830fb0d911d9478b587200/festplatte}, description = {[2009.10270] Embedding-based Zero-shot Retrieval through Query Generation}, interhash = {9fe70666b093f9e227e5240c15181af4}, intrahash = {5c53370740830fb0d911d9478b587200}, keywords = {masterthesis passage-retrieval word-embeddings}, note = {cite arxiv:2009.10270}, timestamp = {2021-01-07T16:08:48.000+0100}, title = {Embedding-based Zero-shot Retrieval through Query Generation}, url = {http://arxiv.org/abs/2009.10270}, year = 2020 }