ScopeIt: Scoping Task Relevant Sentences in Documents

Abstract

A prominent problem faced by conversational agents working with large documents (Eg: emailbased assistants) is the frequent presence of information in the document that is irrelevant to the assistant. This in turn makes it harder for the agent to accurately detect intents, extract entities relevant to those intents and perform the desired action. To address this issue we present a neural model for scoping relevant information for the agent from a large document. We show that when used as the first step in a popularly used email-based assistant for helping users schedule meetings 1 , our proposed model helps improve the performance of the intent detection and entity extraction tasks required by the agent for correctly scheduling meetings: across a suite of 6 downstream tasks, by using our proposed method, we observe an average gain of 35% in precision without any drop in recall. Additionally, we demonstrate that the same approach can be used for component level analysis in large documents, such as signature block identification. * Equal Contribution 1 We use Hedwig in lieu of the actual persona of the agent throughout this paper This work is licensed under a Creative Commons Attribution 4.

Utilizing coarse-grained data in low-data settings for event extraction

“…Our approach is similar to further pre-training in that we also utilize a relevant task's data (sentence and document binary classification data) that can be used in training the same model, but we actually train both tasks at the same time instead of in order. We achieve this by implementing a multi-task model with a hierarchical architecture inspired by ScopeIt [Patra et al, 2020].…”

“…The model in question has a multi-task architecture inspired by ScopeIt [Patra et al, 2020]. Our implementation can be found in github 1 .…”