BerkeleyNLP

@BerkeleyNLP

We work on natural language processing, machine learning, linguistics, and deep learning. PIs: Dan Klein, @alsuhr, @sewon__min

Berkeley, California

nlp.cs.berkeley.edu

Joined September 2019

Following

6,698

Followers

Replies Reposts Media

BerkeleyNLP
@BerkeleyNLP
Jun 7, 2021
This is today in Session 3D (Grounding), 11:40am-12:30pm Pacific (talks), and 12:30-1pm Pacific (poster). #NAACL2021
Daniel Fried
@dan_fried
Jun 6, 2021
On Monday at 11:40a PT at #NAACL2021 session 3D: see how modular neural networks for embodied instruction following improve compositional generalization. Work by @_rodolfocorona_, with @ColinearDevin, Dan Klein (@BerkeleyNLP), and @trevordarrell. Paper: arxiv.org/abs/2010.12764
BerkeleyNLP
@BerkeleyNLP
Nov 17, 2020
The dominant approach in unsupervised parsing is to optimize latent variable models. In contrast, taking inspiration from how linguists decide phrase structure, our #emnlp2020 paper performs unsupervised parsing via constituency tests. Paper: arxiv.org/abs/2010.03146 [1/3]
2
BerkeleyNLP
@BerkeleyNLP
Nov 16, 2020
This is happening in an hour! #EMNLP2020
Ruiqi Zhong
@ZhongRuiqi
Nov 16, 2020
Our Zoom Q&A session will start at 9 a.m. PT tomorrow (session 2D, UTC 17:00). Feel free to join, especially if you are interested in building semantic-based system evaluations! #EMNLP2020
BerkeleyNLP
@BerkeleyNLP
Oct 27, 2020
Our #emnlp2020 paper(arxiv.org/abs/2010.02840) approximates the semantic accuracy of semantic parsing models by comparing the predicted meanings for “multiple possible worlds” rather than the logical forms. It is now the official metric of SPIDER, SParC, and CoSQL.
arxiv.org
Semantic Evaluation for Text-to-SQL with Distilled Test Suites
We propose test suite accuracy to approximate semantic accuracy for Text-to-SQL models. Our method distills a small test suite of databases that achieves high code coverage for the gold query from...
1
BerkeleyNLP
@BerkeleyNLP
Oct 6, 2020
Super excited to announce our paper about an efficient streaming implementation of beam search has been accepted to EMNLP 2020! When using variable-length decoding on GPU, refill your batch after beams terminate to substantially improve speed. (1/2)
1