Search for: All records

Solving crossword puzzles requires diverse reasoning capabilities, access to a vast amount of knowledge about language and the world, and the ability to satisfy the constraints imposed by the structure of the puzzle. In this work, we introduce solving crossword puzzles as a new natural language understanding task. We release a corpus of crossword puzzles collected from the New York Times daily crossword spanning 25 years and comprised of a total of around nine thousand puzzles. These puzzles include a diverse set of clues: historic, factual, word meaning, synonyms/antonyms, fill-in-the-blank, abbreviations, prefixes/suffixes, wordplay, and cross-lingual, as well as clues that depend on the answers to other clues. We separately release the clue-answer pairs from these puzzles as an open-domain question answering dataset containing over half a million unique clue-answer pairs. For the question answering task, our baselines include several sequence-to-sequence and retrieval-based generative models. We also introduce a non-parametric constraint satisfaction baseline for solving the entire crossword puzzle. Finally, we propose an evaluation framework which consists of several complementary performance metrics. 

Transformer-based models have pushed state of the art in many areas of NLP, but our understanding of what is behind their success is still limited. This paper is the first survey of over 150 studies of the popular BERT model. We review the current state of knowledge about how BERT works, what kind of information it learns and how it is represented, common modifications to its training objectives and architecture, the overparameterization issue, and approaches to compression. We then outline directions for future research. 

Large Transformer-based models were shown to be reducible to a smaller number of self-attention heads and layers. We consider this phenomenon from the perspective of the lottery ticket hypothesis, using both structured and magnitude pruning. For fine-tuned BERT, we show that (a) it is possible to find subnetworks achieving performance that is comparable with that of the full model, and (b) similarly-sized subnetworks sampled from the rest of the model perform worse. Strikingly, with structured pruning even the worst possible subnetworks remain highly trainable, indicating that most pre-trained BERT weights are potentially useful. We also study the “good” subnetworks to see if their success can be attributed to superior linguistic knowledge, but find them unstable, and not explained by meaningful self-attention patterns. 

Machine learning has been suggested as a means of identifying individuals at greatest risk for hospital readmission, including psychiatric readmission. We sought to compare the performance of predictive models that use interpretable representations derived via topic modeling to the performance of human experts and nonexperts. We examined all 5076 admissions to a general psychiatry inpatient unit between 2009 and 2016 using electronic health records. We developed multiple models to predict 180-day readmission for these admissions based on features derived from narrative discharge summaries, augmented by baseline sociodemographic and clinical features. We developed models using a training set comprising 70% of the cohort and evaluated on the remaining 30%. Baseline models using demographic features for prediction achieved an area under the curve (AUC) of 0.675 [95% CI 0.674–0.676] on an independent testing set, while language-based models also incorporating bag-of-words features, discharge summaries topics identified by Latent Dirichlet allocation (LDA), and prior psychiatric admissions achieved AUC of 0.726 [95% CI 0.725–0.727]. To characterize the difficulty of the task, we also compared the performance of these classifiers to both expert and nonexpert human raters, with and without feedback, on a subset of 75 test cases. These models outperformed humans on average, including predictions by experienced psychiatrists. Typical note tokens or topics associated with readmission risk were related to pregnancy/postpartum state, family relationships, and psychosis.

 

We propose a triad-based neural network system that generates affinity scores between entity mentions for coreference resolution. The system simultaneously accepts three mentions as input, taking mutual dependency and logical constraints of all three mentions into account, and thus makes more accurate predictions than the traditional pairwise approach. Depending on system choices, the affinity scores can be further used in clustering or mention ranking. Our experiments show that a standard hierarchical clustering using the scores produces state-of-art results with gold mentions on the English portion of CoNLL 2012 Shared Task. The model does not rely on many handcrafted features and is easy to train and use. The triads can also be easily extended to polyads of higher orders. To our knowledge, this is the first neural network system to model mutual dependency of more than two members at mention level.