We are seeking applications for a paid visiting student position to work at the intersection of robotics, perception, natural language understanding, and machine learning (particularly, deep learning). The visiting student will work primarily with TTI-Chicago faculty Matthew Walter and Mohit Bansal, and will have the opportunity to collaborate with other faculty at TTI-Chicago and the University of Chicago.
The broad goal of this work is to enable robots to learn semantically rich, metrically accurate models of their surroundings from their sensor streams (e.g., vision, RGB-D, LIDAR) and natural language descriptions. Existing solutions have shown promise [1], but are limited both in terms of language complexity (closed vocabulary, compact descriptions, shallow parsing, missing coreference) and in terms of model complexity (situated language, flat structure, and alignment). We propose to address these limitations by learning new models of deep learning for language-map translation, spatial part-of hierarchy (meronymy) [3] and co-occurrence, and non-situated utterances; and by building upon recent advancements in NLP [2] (e.g., syntactic and semantic parsing, coreference resolution, multimodal word embeddings with deep learning, and hierarchical ontologies) and in machine vision (scene classification, joint color- and depth-based segmentation, and deep object recognition).
The position will be based at TTI-Chicago, a computer science graduate institute located on the University of Chicago campus, with vibrant research groups in robotics, speech and language processing, machine learning, computer vision, and theory.
Applicants are expected to have strong qualifications in robotics and machine learning, and ideally also in natural language processing. Applicants should have a Bachelor's degree in a related field, and preferably be pursuing an MS or PhD degree.
The target start date is Spring or Summer 2015 and the duration is flexible (3–12 months). There are no citizenship requirements for this position. The position will remain open until filled.
Interested applicants should e-mail robotics-visiting@ttic.edu with the following:
1) brief cover letter that describes your interest in and
qualifications for the position;
2) curriculum vitae including publication list;
3) two or more letters of reference (preferably directly e-mailed to
robotics-visiting@ttic.edu), or the
names of contacts who can supply refrences on request.
[1] Matthew R. Walter, Sachithra Hemachandra, Bianca Homberg, Stefanie Tellex, and Seth Teller, "Learning Semantic Maps from Natural Language Descriptions." Proceedings of Robotics: Science and Systems (RSS), Berlin, Germany, June 2013.
[2] Chen Kong, Dahua Lin, Mohit Bansal, Raquel Urtasun, and Sanja Fidler, "What are You Talking About? Text-to-Image Coreference." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Columbus, OH, June 2014.
[3] Mohit Bansal, David Burkett, Gerard de Melo, and Dan Klein, "Structured Learning for Taxonomy Induction with Belief Propagation." Proceedings of the Association for Computational Linguistics (ACL), Baltimore, MD, June 2014.