• Publications
  • Influence
Representing Numbers in NLP: a Survey and a Vision
TLDR
This work synthesizes best practices for representing numbers in text and articulate a vision for holistic numeracy in NLP, comprised of design trade-offs and a unified evaluation.
SWOW-8500: Word Association task for Intrinsic Evaluation of Word Embeddings
TLDR
A novel intrinsic evaluation task employing large word association datasets (particularly the Small World of Words dataset) is proposed, and correlations not just between performances on SWOW-8500 and previously proposed intrinsic tasks of word similarity prediction, but also with downstream tasks (eg. Text Classification and Natural Language Inference).
Numeracy enhances the Literacy of Language Models
TLDR
A significant improvement in MWP for sentences containing numbers is found, that exponent embeddings are the best number encoders, yielding over 2 points jump in prediction accuracy over a BERT baseline, and that these enhanced literacy skills also generalize to contexts without annotated numbers.
Entity Linking to Knowledge Graphs to Infer Column Types and Properties
This paper describes our broad goal of linking tabular data to semantic knowledge graphs, as well as our specific attempts at solving the Semantic Web Challenge on Tabular Data to Knowledge Graph
Evaluating Opinion Summarization in Ranking
TLDR
This work discusses the evaluation of rankings of documents that aim to summarize the overall opinion expressed in product reviews, and three alternative metrics are adapted from previous similar works to evaluate opinion representativeness.
Are Online Reviews of Physicians Biased Against Female Providers?
TLDR
A new dataset of online patient reviews of male and female healthcare providers with respect to numerical ratings and language use is analyzed, finding that females consistently receive less favorable numerical ratings overall, even after adjusting for specialty.
IJCNLP-2017 Task 3: Review Opinion Diversification (RevOpiD-2017)
TLDR
The first run of RevOpiD shared task at IJCNLP-2107 is summarized and a new dataset is introduced that is annotated for the purpose of evaluating Opinion Mining, Summarization and Disambiguation methods.
BPE beyond Word Boundary: How NOT to use Multi Word Expressions in Neural Machine Translation
TLDR
This work observes that naively extending BPE beyond word boundaries results in incoherent tokens which are themselves better represented as individual words, and finds that Pointwise Mutual Information (PMI) instead of frequency finds better MWEs (e.g., New\_York, Statue of Liberty, neither .