We introduce a novel paraphrastic augmentation strategy based on sentence-level lexically constrained paraphrasing and discriminative span alignment. Our approach allows for the large-scale expansion of existing resources, or the rapid creation of new resources from a small, manually-produced seed corpus. We illustrate our framework on the Berkeley FrameNet Project, a large-scale language understanding effort spanning more than two decades of human labor. Based on roughly four days of collecting training data for the alignment model and approximately one day of parallel compute, we automatically generate 495,300 unique (Frame, Trigger) combinations annotated in context, a roughly 50x expansion atop FrameNet v1.7.

We present a novel document-level model for finding argument spans that fill an event’s roles, connecting related ideas in sentence-level semantic role labeling and coreference resolution. Because existing datasets for cross-sentence linking are small, development of our neural model is supported through the creation of a new resource, Roles Across Multiple Sentences (RAMS), which contains 9,124 annotated events across 139 types. We demonstrate strong performance of our model on RAMS and other event-related datasets.

Lexically-constrained sequence decoding allows for explicit positive or negative phrase-based constraints to be placed on target output strings in generation tasks such as machine translation or monolingual text rewriting. We describe vectorized dynamic beam allocation, which extends work in lexically-constrained decoding to work with batching, leading to a five-fold improvement in throughput when working with positive constraints. Faster decoding enables faster exploration of constraint strategies: we illustrate this via data augmentation experiments with a monolingual rewriter applied to the tasks of natural language inference, question answering and machine translation, showing improvements in all three.

We designed and constructed a pipeline system for the Streaming Multimedia Knowledge Base Population (SM-KBP) 2019 evaluation. Our pipeline consists of a series of Information Extraction and Machine Translation components, supporting the population of knowledge graphs from a multilingual corpus (see Figure 1). Some of our key contributions include: a new dataset supporting event argument linking across multiple sentences; a novel algorithm for predicting such linkages, even when arguments are not co-referent with mentions in the same sentence as the event trigger; and a novel algorithm for supporting hierarchical typing of events, relations and entities.

We present a model for semantic proto-role labeling (SPRL) using an adapted bidirectional LSTM encoding strategy that we call "Neural-Davidsonian": predicate-argument structure is represented as pairs of hidden states corresponding to predicate and argument head tokens of the input sequence. We demonstrate: (1) state-of-the-art results in SPRL, and (2) that our network naturally shares parameters between attributes, allowing for learning new attribute types with limited added supervision.

We constructed a pipeline-based system for participation in the 2018 pilot NIST SMKBP evaluation. Our pipeline was assembled from a series of components representing contemporary approaches to each of the distinct Information Extraction and Machine Translation tasks required. Our goals for this system were: (1) to demonstrate an extant SM-KBP capability (can we run on the data and produce a valid output); and (2) to stand as a non-trivial baseline for contrasting against our future efforts for this task.