Ellie Pavlick

Ideas Mova took (1)

From the ideas ledger: an idea tried, kept or queued, or measured with no gain, and what it gave.

• kept — Use the paper's verb–signature table as the seed of the implication signatures proposed for Mova's frame factuality (see rudinger-2018-neural-models-factuality I1): our own list built from these published class memberships, stored in prag/data with the… (How well do NLI models capture verb veridicality? · the signature table seeded prag/data/signatures.tsv (six of the eight classes, «to» complements only) (35db4e3f7))

Papers

1. Which Linguist Invented the Lightbulb? Presupposition Verification for Question-Answering (2021)
   The paper studies unanswerable questions in Natural Questions. In a hand analysis of 100 of them, 30% are unanswerable because a presupposition of the question cannot be verified in the linked document («What is the stock symbol for Mars candy?» presupposes that there is one). Most such presuppositions come from the…
2. How well do NLI models capture verb veridicality? (2019) 1 idea kept
   From MultiNLI sentences with a verb + to/that complement the authors build 1,498 sentences over 137 verbs, each paired twice with its complement as hypothesis: once as is and once with the main verb negated («He knows / does not know that the answer is 5» → «The answer is 5»).
3. Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference (2019)
   Statistical NLI models can score high on MNLI while relying on three syntactic shortcuts: lexical overlap (a premise entails any hypothesis built from its words), the subsequence heuristic (it entails any contiguous word span), and the constituent heuristic (it entails any complete subtree).
4. Inherent Disagreements in Human Textual Inferences (2019)
   The paper asks whether disagreement among annotators of NLI pairs is noise that aggregation removes, or a stable signal.
5. Learning Scalar Adjective Intensity from Paraphrases (2018)
   The paper orders scalar adjectives (warm, hot, scalding) by intensity from paraphrase pairs such as "really hot" ↔ "scalding".
6. Optimizing Statistical Machine Translation for Text Simplification (2016)
   The paper adapts a syntax-based statistical translation system to simplify sentences. Its paraphrase rules come from large bilingual paraphrase tables (PPDB), which are easier to get than manually simplified text, and they are ranked with simplicity-specific features. The main contribution is the metric SARI. It…
7. An Empirical Analysis of Formality in Online Communication (2016)
   The paper studies sentence-level formality. Crowd workers rated sentences from four genres (news, blogs, emails and Yahoo Answers) on a scale from -3 (very informal) to +3 (very formal), and the agreement among raters is measured. A ridge regression model is then trained on eleven groups of features: n-grams, word…
8. PPDB 2.0: Better paraphrase ranking, fine-grained entailment relations, word embeddings, and style classification (2015)
   PPDB 1.0 ranked its pivot-extracted paraphrases with a fixed heuristic, and this paper replaces the ranking with a discriminative model.

Co-authors (18)

• Chris Callison-Burch ×3
• Alexis Ross
• Anne Cocos
• Benjamin Van Durme
• Burcu Karagol Ayan
• Courtney Napoles
• Deepak Ramachandran
• Joel Tetreault
• Juri Ganitkevitch
• Marianna Apidianaki
• Najoung Kim
• Pushpendre Rastogi
• Quanze Chen
• R. Thomas McCoy
• Skyler Wharton
• Tal Linzen
• Tom Kwiatkowski
• Wei Xu

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/authors/ellie-pavlick/