Papers by Emily Pitler
New Protocols and Negative Results for Textual Entailment Data Collection (2020.emnlp-main)
Copied to clipboard
| Challenge: | Natural language inference data has proven useful in benchmarking and as pretraining data for tasks requiring language understanding. |
| Approach: | They propose four alternative protocols to improve annotation quality and diversity . they use 8.5k-example training sets to compare different protocols . |
| Outcome: | The proposed protocols improve the ease of training and quality of the examples. |
A Challenge Set and Methods for Noun-Verb Ambiguity (D18-1)
Copied to clipboard
| Challenge: | English part-of-speech taggers make egregious errors related to noun-verb ambiguity, despite having achieved 97%+ accuracy on the WSJ Penn Treebank since 2002. |
| Approach: | They propose to use a WSJ dataset to identify 30,000 examples of noun-verb ambiguity . they find that english part-of-speech taggers make egregious errors related to nouns and verbs . |
| Outcome: | The proposed model improves on the WSJ Penn Treebank by 14% and 52% relative to the previous model. |
Synthetic QA Corpora Generation with Roundtrip Consistency (P19-1)
Copied to clipboard
| Challenge: | Existing methods for generating synthetic question answering corpora are not suitable for QA, but can be constructed from widely available natural text. |
| Approach: | They propose a method for generating synthetic question answering corpora by combining question generation and answer extraction models and filtering the results to ensure roundtrip consistency. |
| Outcome: | The proposed model achieves exact match and F1 at less than 0.1% and 0.4% from human performance on SQuAD2 and NQ. |
Evaluating the Impact of Model Scale for Compositional Generalization in Semantic Parsing (2022.emnlp-main)
Copied to clipboard
Linlu Qiu, Peter Shaw, Panupong Pasupat, Tianze Shi, Jonathan Herzig, Emily Pitler, Fei Sha, Kristina Toutanova
| Challenge: | Pre-trained language models struggle on out-of-distribution compositional generalization . recent work shows considerable improvements on many NLP tasks from model scaling . |
| Approach: | They evaluate encoder-decoder models up to 11B parameters and decoder-only models up 540B parameters . they compare scaling curves for fine-tuning, prompt tuning, and in-context learning methods . |
| Outcome: | The proposed scaling methods improve compositional generalization on many tasks . fine-tuning generally has flat or negative scaling curves on out-of-distribution compositional . larger models are better at modeling the syntax of the output space, the study finds . |
Syntactic Data Augmentation Increases Robustness to Inference Heuristics (2020.acl-main)
Copied to clipboard
| Challenge: | Pretrained neural models lack sensitivity to word order on controlled challenge sets . augmentation methods that improve accuracy on standard training sets may be a problem . |
| Approach: | They propose to augment standard training sets with syntactically informative examples by applying syntastic transformations to sentences from the MNLI corpus. |
| Outcome: | The proposed method improved BERT’s accuracy on controlled examples that diagnose sensitivity to word order from 0.28 to 0.73 without affecting performance on the MNLI test set. |
Giving BERT a Calculator: Finding Operations and Arguments with Reading Comprehension (D19-1)
Copied to clipboard
| Challenge: | End-to-end reading comprehension models have been successful at extracting text answers, but there are still problems with generalizing them to abstractive numerical reasoning. |
| Approach: | They propose to augment a BERT-based reading comprehension model with a set of executable ‘programs’ which encompass simple arithmetic as well as extraction. |
| Outcome: | The proposed model can perform 33% absolute improvement on the DROP dataset, with very few training examples. |
Morphosyntactic Tagging with a Meta-BiLSTM Model over Context Sensitive Token Encodings (P18-1)
Copied to clipboard
| Challenge: | recurrent neural networks have produced significant advances in part-of-speech tagging accuracy . a common feature of these models is the presence of rich initial word encodings . however, word or sub-word information interacts only through subsequent recursive layers . |
| Approach: | They propose to use recurrent neural networks with sentence-level context for initial character and word-based representations. |
| Outcome: | The proposed model has the highest accuracy of all participating systems in the CoNLL 2017 task. |