A new type of deep contextualized word representation is introduced that models both complex characteristics of word use and how these uses vary across linguistic contexts, allowing downstream models to mix different types of semi-supervision signals.
It is found that transferring from entailment data is more effective than transferring from paraphrase or extractive QA data, and that it, surprisingly, continues to be very beneficial even when starting from massive pre-trained language models such as BERT.
This paper trains a naive model that makes predictions exclusively based on dataset biases, and a robust model as part of an ensemble with the naive one in order to encourage it to focus on other patterns in the data that are more likely to generalize.
It is shown that it is possible to significantly improve performance by using a modified training scheme that teaches the model to ignore non-answer containing paragraphs, which involves sampling multiple paragraphs from each document, and using an objective function that requires themodel to produce globally correct output.
An algorithm that extracts figures, tables, and captions from documents called “PDFFigures 2.0” that analyzes the structure of individual pages by detecting captions, graphical elements, and chunks of body text, and then locates figures and tables by reasoning about the empty regions within that text.
The convolutional neural networks trained in this work can consistently defeat the well known Go program GNU Go and win some games against state of the art Go playing program Fuego while using a fraction of the play time.
This work introduces a new dataset of 150 computer science papers along with ground truth labels for the locations of the figures, tables and captions within them and demonstrates a caption-to-figure matching component that is effective even in cases where individual captions are adjacent to multiple figures.
This paper proposes a method that can automatically detect and ignore dataset-specific patterns, which it hypothesize are likely to reflect dataset bias, and trains a lower capacity model in an ensemble with a higher capacity model.
IKE is a new extraction tool that performs fast, interactive bootstrapping to develop high-quality extraction patterns for targeted relations and is the first interactive extraction tool to seamlessly integrate symbolic and distributional methods for search.
This work uses a dataset of 1M+ images spanning 10k+ visual concepts to demonstrate webly-supervised concept expansion for two existing GPVs and proposes a new architecture, GPV-2 that supports a variety of tasks — from vision tasks like classification and localization to vision+language tasks like QA and captioning, to more niche ones like human-object interaction detection.