Keith Hall
Also known as Keith B. Hall
Keith Hall is a natural language processing researcher whose work runs from parsing speech to dense text retrieval. He received a PhD from Brown University in 2005, served as a Research Professor at Johns Hopkins University, and joined Google as a research scientist. In 2013, he was one of thirteen authors of Universal Dependency Annotation for Multilingual Parsing, which won the ACL Test-of-Time Paper Award in 2023. His later papers include Sentence-T5, Large Dual Encoders Are Generalizable Retrievers and Promptagator in 2021 and 2022, followed by OpenMSD, a multilingual dataset of scientific papers, in 2023.
- Badges
Timeline
-
2004–2005
PhD at Brown University
He coauthors Attention Shifting for Parsing Speech with Mark Johnson in 2004, presented at the ACL meeting in Barcelona. He receives his PhD from Brown University in 2005 and serves as a Research Professor at Johns Hopkins University before moving to Google.
-
2007
K-best spanning tree parsing
His single-author paper K-best Spanning Tree Parsing appears at the 45th Annual Meeting of the Association of Computational Linguistics in Prague in 2007.
-
2010–2011
Structured perceptron and parser transfer
As a Google research scientist, he coauthors Distributed Training Strategies for the Structured Perceptron with Ryan McDonald and Gideon Mann in 2010. A 2011 paper with McDonald and Slav Petrov, Multi-Source Transfer of Delexicalized Dependency Parsers, appears at the Conference on Empirical Methods in Natural Language Processing in Edinburgh.
-
2013
Universal dependency annotation
In 2013, he is one of thirteen authors of Universal Dependency Annotation for Multilingual Parsing, presented at ACL. He also co-chairs the ACL 2013 tutorial program with Johan Bos.
-
2020
The Dakshina dataset
The Dakshina dataset, which he coauthors in 2020, offers text in both the Latin and native scripts for 12 South Asian languages.
-
2021–2022
Dense retrieval and sentence embeddings
In 2021, he coauthors Sentence-T5, which extracts sentence embeddings from T5 models, and Large Dual Encoders Are Generalizable Retrievers, which finds that scaling up dual encoders improves out-of-domain retrieval. In 2022, he coauthors Promptagator, which builds task-specific dense retrievers from no more than 8 examples.
-
2023
An ACL Test-of-Time award
In 2023, the ACL Test-of-Time Paper Award goes to Universal Dependency Annotation for Multilingual Parsing, one of two papers honored from 2013.
-
2023–2024
OpenMSD for scientific papers
He coauthors OpenMSD in 2023, a dataset of 74 million papers in 103 languages and 778 million citation pairs for multilingual scientific document similarity. The paper appears at LREC-COLING 2024 in Torino.
Career
- 2010–2011GoogleResearch Scientist
- Johns Hopkins UniversityResearch Professor
Education
- 2005Brown UniversityPhD
Sources and credits
Photo: We have no published photo of Keith Hall yet. Send us a free-licensed photo.
- Keith B. Hall - Research at Google
- Attention Shifting for Parsing Speech - ACL Anthology
- K-best Spanning Tree Parsing - ACL Anthology
- Distributed Training Strategies for the Structured Perceptron - ACL Anthology
- Multi-Source Transfer of Delexicalized Dependency Parsers - ACL Anthology
- Universal Dependency Annotation for Multilingual Parsing - ACL Anthology
- [Corpora-List] ACL-2013: call for tutorials
- Processing South Asian Languages Written in the Latin Script: the Dakshina Dataset
- Sentence-T5: Scalable Sentence Encoders from Pre-trained Text-to-Text Models
- Large Dual Encoders Are Generalizable Retrievers
- Promptagator: Few-shot Dense Retrieval From 8 Examples
- Announcement of the 2023 ACL Test-of-Time Paper Award
- OpenMSD: Towards Multilingual Scientific Documents Similarity Measurement
- OpenMSD: Towards Multilingual Scientific Documents Similarity Measurement - ACL Anthology



