Incremental Open World Reference Resolution

PROJECT NUMBER: 41

AUTHOR: Thomas Bennett, Computer Science | MENTOR: Tom Williams, Computer Science

ABSTRACT

—

Referring expression understanding and generation are critical for robots to communicate about the world around them. Recent advancements have enabled both incremental reference resolution (i.e., processing an utterance word by word in real-time as it is spoken) and open-world reference resolution (i.e., ability to resolve references both to known and previously unknown entities). In this work, we present I-POWER: the first algorithm for performing reference resolution both incrementally and in open-world environments

VISUAL PRESENTATION

—

AUTHOR BIOGRAPHY

—

Thomas Bennett is a junior in Computer Science at Mines. He works in the MIRRORLab under Tom Williams working on topics in Natural Language Processing. Mainly, he has been working on topics relating to Reference Resolution, however is looking into other topics in Natural Language Generation.

Sponsored by Phillips 66