Kevin Gimpel
YOU?
Author Swipe
View article: Discriminative Feature-Rich Modeling for Syntax-Based Machine Translation
Discriminative Feature-Rich Modeling for Syntax-Based Machine Translation Open
Fully-automated, high-quality machine translation promises to revolutionize human communication. But as anyone who has used a machine translation system knows, we are not there yet. In this thesis, we address four areas in which we believe…
View article: A baseline for detecting misclassified and out-of-distribution examples in neural networks
A baseline for detecting misclassified and out-of-distribution examples in neural networks Open
We consider the two related problems of detecting if an example is misclassified or out-of-distribution. We present a simple baseline that utilizes probabilities from softmax distributions. Correctly classified examples tend to have greate…
View article: Structured Tree Alignment for Evaluation of (Speech) Constituency Parsing
Structured Tree Alignment for Evaluation of (Speech) Constituency Parsing Open
We present the structured average intersection-over-union ratio (STRUCT-IOU), a similarity metric between constituency parse trees motivated by the problem of evaluating speech parsers. STRUCT-IOU enables comparison between a constituency …
View article: GEE! Grammar Error Explanation with Large Language Models
GEE! Grammar Error Explanation with Large Language Models Open
Grammatical error correction tools are effective at correcting grammatical errors in users' input sentences but do not provide users with \textit{natural language} explanations about their errors. Such explanations are essential for helpin…
View article: MAP's not dead yet: Uncovering true language model modes by conditioning away degeneracy
MAP's not dead yet: Uncovering true language model modes by conditioning away degeneracy Open
It has been widely observed that exact or approximate MAP (mode-seeking) decoding from natural language generation (NLG) models consistently leads to degenerate outputs (Holtzman et al., 2019; Stahlberg and Byrne, 2019). Prior work has att…
View article: Audio-Visual Neural Syntax Acquisition
Audio-Visual Neural Syntax Acquisition Open
We study phrase structure induction from visually-grounded speech. The core idea is to first segment the speech waveform into sequences of word segments, and subsequently induce phrase structure using the inferred segment-level continuous …
View article: The Benefits of Label-Description Training for Zero-Shot Text Classification
The Benefits of Label-Description Training for Zero-Shot Text Classification Open
Pretrained language models have improved zero-shot text classification by allowing the transfer of semantic knowledge from the training data in order to classify among specific label sets in downstream tasks. We propose a simple way to fur…
View article: The Benefits of Label-Description Training for Zero-Shot Text Classification
The Benefits of Label-Description Training for Zero-Shot Text Classification Open
Pretrained language models have improved zero-shot text classification by allowing the transfer of semantic knowledge from the training data in order to classify among specific label sets in downstream tasks. We propose a simple way to fur…
View article: Deep Clustering of Text Representations for Supervision-Free Probing of Syntax
Deep Clustering of Text Representations for Supervision-Free Probing of Syntax Open
We explore deep clustering of multilingual text representations for unsupervised model interpretation and induction of syntax. As these representations are high-dimensional, out-of-the-box methods like K-means do not work well. Thus, our a…
View article: Chess as a Testbed for Language Model State Tracking
Chess as a Testbed for Language Model State Tracking Open
Transformer language models have made tremendous strides in natural language understanding tasks. However, the complexity of natural language makes it challenging to ascertain how accurately these models are tracking the world state underl…
View article: "What makes a question inquisitive?" A Study on Type-Controlled Inquisitive Question Generation
"What makes a question inquisitive?" A Study on Type-Controlled Inquisitive Question Generation Open
We propose a type-controlled framework for inquisitive question generation. We annotate an inquisitive question dataset with question types, train question type classifiers, and finetune models for type-controlled question generation. Empi…
View article: “What makes a question inquisitive?” A Study on Type-Controlled Inquisitive Question Generation
“What makes a question inquisitive?” A Study on Type-Controlled Inquisitive Question Generation Open
We propose a type-controlled framework for inquisitive question generation. We annotate an inquisitive question dataset with question types, train question type classifiers, and finetune models for type-controlled question generation. Empi…
View article: Baked-in State Probing
Baked-in State Probing Open
Neural language models have been analyzed for their linguistic and extra-linguistic knowledge via probing. Of particular interest has been the following question: how much can a language model trained only on form learn about meaning? Rece…
View article: Paraphrastic Representations at Scale
Paraphrastic Representations at Scale Open
We present a system that allows users to train their own state-of-the-art paraphrastic sentence representations in a variety of languages. We release trained models for English, Arabic, German, Spanish, French, Russian, Turkish, and Chines…
View article: Substructure Distribution Projection for Zero-Shot Cross-Lingual Dependency Parsing
Substructure Distribution Projection for Zero-Shot Cross-Lingual Dependency Parsing Open
We present substructure distribution projection (SubDP), a technique that projects a distribution over structures in one domain to another, by projecting substructure distributions separately. Models for the target domain can then be train…
View article: SummScreen: A Dataset for Abstractive Screenplay Summarization
SummScreen: A Dataset for Abstractive Screenplay Summarization Open
We introduce SummScreen, a summarization dataset comprised of pairs of TV series transcripts and human written recaps. The dataset provides a challenging testbed for abstractive summarization for several reasons. Plot details are often exp…
View article: Reconsidering the Past: Optimizing Hidden States in Language Models
Reconsidering the Past: Optimizing Hidden States in Language Models Open
We present Hidden-State Optimization (HSO), a gradient-based method for improving the performance of transformer language models at inference time. Similar to dynamic evaluation (Krause et al., 2018), HSO computes the gradient of the log-p…
View article: Substructure Distribution Projection for Zero-Shot Cross-Lingual Dependency Parsing
Substructure Distribution Projection for Zero-Shot Cross-Lingual Dependency Parsing Open
We present substructure distribution projection (SubDP), a technique that projects a distribution over structures in one domain to another, by projecting substructure distributions separately. Models for the target domains can be then trai…
View article: TVStoryGen: A Dataset for Generating Stories with Character Descriptions
TVStoryGen: A Dataset for Generating Stories with Character Descriptions Open
We introduce TVStoryGen, a story generation dataset that requires generating detailed TV show episode recaps from a brief summary and a set of documents describing the characters involved. Unlike other story generation datasets, TVStoryGen…
View article: TVRecap: A Dataset for Generating Stories with Character Descriptions.
TVRecap: A Dataset for Generating Stories with Character Descriptions. Open
We introduce TVRecap, a story generation dataset that requires generating detailed TV show episode recaps from a brief summary and a set of documents describing the characters involved. Unlike other story generation datasets, TVRecap conta…
View article: Paraphrastic Representations at Scale
Paraphrastic Representations at Scale Open
We present a system that allows users to train their own state-of-the-art paraphrastic sentence representations in a variety of languages. We also release trained models for English, Arabic, German, French, Spanish, Russian, Turkish, and C…
View article: SummScreen: A Dataset for Abstractive Screenplay Summarization
SummScreen: A Dataset for Abstractive Screenplay Summarization Open
We introduce SummScreen, a summarization dataset comprised of pairs of TV series transcripts and human written recaps. The dataset provides a challenging testbed for abstractive summarization for several reasons. Plot details are often exp…
View article: Learning Chess Blindfolded: Evaluating Language Models on State Tracking.
Learning Chess Blindfolded: Evaluating Language Models on State Tracking. Open
Transformer language models have made tremendous strides in natural language understanding tasks. However, the complexity of natural language makes it challenging to ascertain how accurately these models are tracking the world state underl…
View article: Chess as a Testbed for Language Model State Tracking
Chess as a Testbed for Language Model State Tracking Open
Transformer language models have made tremendous strides in natural language understanding tasks. However, the complexity of natural language makes it challenging to ascertain how accurately these models are tracking the world state underl…
View article: Substructure Substitution: Structured Data Augmentation for NLP
Substructure Substitution: Structured Data Augmentation for NLP Open
We study a family of data augmentation methods, substructure substitution (SUB2), for natural language processing (NLP) tasks. SUB2 generates new examples by substituting substructures (e.g., subtrees or subsequences) with ones with the sa…
View article: Unsupervised Label Refinement Improves Dataless Text Classification
Unsupervised Label Refinement Improves Dataless Text Classification Open
Dataless text classification is capable of classifying documents into previously unseen labels by assigning a score to any document paired with a label description.While promising, it crucially relies on accurate descriptions of the label …
View article: On Generalization in Coreference Resolution
On Generalization in Coreference Resolution Open
While coreference resolution is defined independently of dataset domain, most models for performing coreference resolution do not transfer well to unseen domains. We consolidate a set of 8 coreference resolution datasets targeting differen…
View article: Substructure Substitution: Structured Data Augmentation for NLP
Substructure Substitution: Structured Data Augmentation for NLP Open
We study a family of data augmentation methods, substructure substitution (SUB 2 ), that generalizes prior methods.SUB 2 generates new examples by substituting substructures (e.g., subtrees or subsequences) with others having the same labe…
View article: FlowPrior: Learning Expressive Priors for Latent Variable Sentence Models
FlowPrior: Learning Expressive Priors for Latent Variable Sentence Models Open
Variational autoencoders (VAEs) are widely used for latent variable modeling of text. We focus on variations that learn expressive prior distributions over the latent variable. We find that existing training strategies are not effective fo…
View article: WikiTableT: A Large-Scale Data-to-Text Dataset for Generating Wikipedia Article Sections
WikiTableT: A Large-Scale Data-to-Text Dataset for Generating Wikipedia Article Sections Open
Datasets for data-to-text generation typically focus either on multi-domain, single-sentence generation or on single-domain, long-form generation.In this work, we cast generating Wikipedia sections as a data-to-text generation task and cre…