Share on Facebook Tweet on Twitter Share on LinkedIn Share by email
Generating Descriptions of Visible Objects

Speaker  Margaret Mitchell

Affiliation  Johns Hopkins University

Host  Lucy Vanderwende

Duration  01:08:48

Date recorded  6 May 2013

In this talk, I will be discussing the connection between vision and language, particularly focusing on how to generate human-like reference to visible objects. Some of the findings in this talk have been used to automatically generate descriptions of images (EACL 2012), order descriptive modifiers before a noun (ACL 11), and approximate human preferences for describing color and size (ENLG 2011, NAACL 2013). Evaluating automatically generated text offers interesting challenges, particularly when aiming to capture speaker variation, and I will spend some time discussing how best to measure the naturalness of generated descriptions against a corpus of human-produced descriptions.

©2013 Microsoft Corporation. All rights reserved.
> Generating Descriptions of Visible Objects