A generative probabilistic framework for learning spatial language

Colin R. Dawson, Jeremy Wright, Antons Rebguns, Marco Valenzuela Escarcega, Daniel Fried, Paul R Cohen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

The language of space and spatial relations is a rich source of abstract semantic structure. We develop a probabilistic model that learns to understand utterances that describe spatial configurations of objects in a tabletop scene by seeking the meaning that best explains the sentence chosen. The inference problem is simplified by assuming that sentences express symbolic representations of (latent) semantic relations between referents and landmarks in space, and that given these symbolic representations, utterances and physical locations are conditionally independent. As such, the inference problem factors into a symbol-grounding component (linking propositions to physical locations) and a symbol-translation component (linking propositions to parse trees). We evaluate the model by eliciting production and comprehension data from human English speakers and find that our system recovers the referent of spatial utterances at a level of proficiency approaching human performance.

Original languageEnglish (US)
Title of host publication2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013 - Electronic Conference Proceedings
DOIs
StatePublished - 2013
Event2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013 - Osaka, Japan
Duration: Aug 18 2013Aug 22 2013

Other

Other2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013
CountryJapan
CityOsaka
Period8/18/138/22/13

Fingerprint

Semantics
Electric grounding
Statistical Models

ASJC Scopus subject areas

  • Artificial Intelligence
  • Human-Computer Interaction
  • Software

Cite this

Dawson, C. R., Wright, J., Rebguns, A., Escarcega, M. V., Fried, D., & Cohen, P. R. (2013). A generative probabilistic framework for learning spatial language. In 2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013 - Electronic Conference Proceedings [6652560] https://doi.org/10.1109/DevLrn.2013.6652560

A generative probabilistic framework for learning spatial language. / Dawson, Colin R.; Wright, Jeremy; Rebguns, Antons; Escarcega, Marco Valenzuela; Fried, Daniel; Cohen, Paul R.

2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013 - Electronic Conference Proceedings. 2013. 6652560.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Dawson, CR, Wright, J, Rebguns, A, Escarcega, MV, Fried, D & Cohen, PR 2013, A generative probabilistic framework for learning spatial language. in 2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013 - Electronic Conference Proceedings., 6652560, 2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013, Osaka, Japan, 8/18/13. https://doi.org/10.1109/DevLrn.2013.6652560
Dawson CR, Wright J, Rebguns A, Escarcega MV, Fried D, Cohen PR. A generative probabilistic framework for learning spatial language. In 2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013 - Electronic Conference Proceedings. 2013. 6652560 https://doi.org/10.1109/DevLrn.2013.6652560
Dawson, Colin R. ; Wright, Jeremy ; Rebguns, Antons ; Escarcega, Marco Valenzuela ; Fried, Daniel ; Cohen, Paul R. / A generative probabilistic framework for learning spatial language. 2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013 - Electronic Conference Proceedings. 2013.
@inproceedings{a2f0ae99a00c4c908d1efdaeac5fe95d,
title = "A generative probabilistic framework for learning spatial language",
abstract = "The language of space and spatial relations is a rich source of abstract semantic structure. We develop a probabilistic model that learns to understand utterances that describe spatial configurations of objects in a tabletop scene by seeking the meaning that best explains the sentence chosen. The inference problem is simplified by assuming that sentences express symbolic representations of (latent) semantic relations between referents and landmarks in space, and that given these symbolic representations, utterances and physical locations are conditionally independent. As such, the inference problem factors into a symbol-grounding component (linking propositions to physical locations) and a symbol-translation component (linking propositions to parse trees). We evaluate the model by eliciting production and comprehension data from human English speakers and find that our system recovers the referent of spatial utterances at a level of proficiency approaching human performance.",
author = "Dawson, {Colin R.} and Jeremy Wright and Antons Rebguns and Escarcega, {Marco Valenzuela} and Daniel Fried and Cohen, {Paul R}",
year = "2013",
doi = "10.1109/DevLrn.2013.6652560",
language = "English (US)",
isbn = "9781479910366",
booktitle = "2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013 - Electronic Conference Proceedings",

}

TY - GEN

T1 - A generative probabilistic framework for learning spatial language

AU - Dawson, Colin R.

AU - Wright, Jeremy

AU - Rebguns, Antons

AU - Escarcega, Marco Valenzuela

AU - Fried, Daniel

AU - Cohen, Paul R

PY - 2013

Y1 - 2013

N2 - The language of space and spatial relations is a rich source of abstract semantic structure. We develop a probabilistic model that learns to understand utterances that describe spatial configurations of objects in a tabletop scene by seeking the meaning that best explains the sentence chosen. The inference problem is simplified by assuming that sentences express symbolic representations of (latent) semantic relations between referents and landmarks in space, and that given these symbolic representations, utterances and physical locations are conditionally independent. As such, the inference problem factors into a symbol-grounding component (linking propositions to physical locations) and a symbol-translation component (linking propositions to parse trees). We evaluate the model by eliciting production and comprehension data from human English speakers and find that our system recovers the referent of spatial utterances at a level of proficiency approaching human performance.

AB - The language of space and spatial relations is a rich source of abstract semantic structure. We develop a probabilistic model that learns to understand utterances that describe spatial configurations of objects in a tabletop scene by seeking the meaning that best explains the sentence chosen. The inference problem is simplified by assuming that sentences express symbolic representations of (latent) semantic relations between referents and landmarks in space, and that given these symbolic representations, utterances and physical locations are conditionally independent. As such, the inference problem factors into a symbol-grounding component (linking propositions to physical locations) and a symbol-translation component (linking propositions to parse trees). We evaluate the model by eliciting production and comprehension data from human English speakers and find that our system recovers the referent of spatial utterances at a level of proficiency approaching human performance.

UR - http://www.scopus.com/inward/record.url?scp=84891093955&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84891093955&partnerID=8YFLogxK

U2 - 10.1109/DevLrn.2013.6652560

DO - 10.1109/DevLrn.2013.6652560

M3 - Conference contribution

AN - SCOPUS:84891093955

SN - 9781479910366

BT - 2013 IEEE 3rd Joint International Conference on Development and Learning and Epigenetic Robotics, ICDL 2013 - Electronic Conference Proceedings

ER -