Lexion at 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)
One of the things we’re most proud of here at Lexion is the amazing natural language processing (NLP) talent we have on our team that’s helping us build the AI that powers our contract management system under the hood. One of these talented individuals includes Allison Hegel.
Allison recently presented a research paper at the 2020 Conference on Empirical Methods Natural Language Processing. The paper offers a new neural model for rewriting text in a controlled way: for example, taking a recipe for meat lasagna and rewriting it to be vegetarian. This is a challenge for existing state-of-the-art language models like GPT-3, which often stray from the original prompt and say things that are not factually accurate. It is especially difficult to generate coherent documents that are longer than a sentence. The paper provides a new machine learning task, a new dataset, and a new model that out-performs existing models at rewriting documents based on a constraint.
Paper link: https://www.aclweb.org/anthology/2020.emnlp-main.526/
Recorded talk link: https://slideslive.com/38939103/substance-over-style-documentlevel-targeted-content-transfer