Jointly Learning to Parse and Perceive: Connecting Natural Language to the Physical World

This page contains supplementary material and data for the paper:

Jointly Learning to Parse and Perceive: Connecting Natural Language to the Physical World. Jayant Krishnamurthy and Thomas Kollar. Transactions of the Association for Computational Linguistics, 2013. PDF

The two data sets used in the paper are available here:

• Scene data set. Contains 15 images and 284 natural language object descriptions.
• GeoQA data set. Contains 10 maps (represented as latitude-longitude polygons) and 263 natural language questions.

Table of lexicon induction heuristics

Code for this project is available on GitHub