• Publications
  • Influence
A Sensitivity Analysis of (and Practitioners’ Guide to) Convolutional Neural Networks for Sentence Classification
TLDR
A sensitivity analysis of one-layer CNNs is conducted to explore the effect of architecture components on model performance; the aim is to distinguish between important and comparatively inconsequential design decisions for sentence classification.
ERASER: A Benchmark to Evaluate Rationalized NLP Models
TLDR
This work proposes the Evaluating Rationales And Simple English Reasoning (ERASER) a benchmark to advance research on interpretable models in NLP, and proposes several metrics that aim to capture how well the rationales provided by models align with human rationales, and also how faithful these rationales are.
Closing the Gap between Methodologists and End-Users: R as a Computational Back-End
TLDR
This paper presents open-source meta-analysis software that uses R as the underlying statistical engine, and Python for the GUI, and a framework that allows methodologists to implement new methods in R that are then automatically integrated into the GUI for use by end-users, so long as the programmer conforms to the interface.
Attention is not Explanation
TLDR
This work performs extensive experiments across a variety of NLP tasks to assess the degree to which attention weights provide meaningful “explanations” for predictions, and finds that they largely do not.
Modelling Context with User Embeddings for Sarcasm Detection in Social Media
TLDR
This work proposes to automatically learn and then exploit user embeddings, to be used in concert with lexical signals to recognize sarcasm, and shows that the model outperforms a state-of-the-art approach leveraging an extensive set of carefully crafted features.
Semi-automated screening of biomedical citations for systematic reviews
TLDR
A semi-automated citation screening algorithm for systematic reviews has the potential to substantially reduce the number of citations reviewers have to manually screen, without compromising the quality and comprehensiveness of the review.
Deploying an interactive machine learning system in an evidence-based practice center: abstrackr
TLDR
The ongoing development of an end-to-end interactive machine learning system at the Tufts Evidence-based Practice Center is described and abstrackr, an online tool for the task of citation screening for systematic reviews is developed, which provides an interface to the machine learning methods.
A Corpus with Multi-Level Annotations of Patients, Interventions and Outcomes to Support Language Processing for Medical Literature
TLDR
A corpus of 5,000 richly annotated abstracts of medical articles describing clinical randomized controlled trials is presented and a set of challenging NLP tasks that would aid searching of the medical literature and the practice of evidence-based medicine are outlined.
Aggregating and Predicting Sequence Labels from Crowd Annotations
TLDR
A suite of methods for aggregating sequential crowd labels to infer a best single set of consensus annotations and using crowd annotations as training data for a model that can predict sequences in unannotated text are evaluated.
Learning to Faithfully Rationalize by Construction
TLDR
Variations of this simple framework yield predictive performance superior to ‘end-to-end’ approaches, while being more general and easier to train.
...
1
2
3
4
5
...