Polynomial semantic indexing

Bing Ba, Kunihiko Sadamasa, Jason Weston, Yanjun Qi, David Grangier, Corinna Cortes, Ronan Collobert, Mehryar Mohri

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

Abstract

We present a class of nonlinear (polynomial) models that are discriminatively trained to directly map from the word content in a query-document or document-document pair to a ranking score. Dealing with polynomial models on word features is computationally challenging. We propose a low-rank (but diagonal preserving) representation of our polynomial models to induce feasible memory and computation requirements. We provide an empirical study on retrieval tasks based on Wikipedia documents, where we obtain state-of-the-art performance while providing realistically scalable methods.

Original languageEnglish (US)
Title of host publicationAdvances in Neural Information Processing Systems 22 - Proceedings of the 2009 Conference
Pages64-72
Number of pages9
StatePublished - 2009
Event23rd Annual Conference on Neural Information Processing Systems, NIPS 2009 - Vancouver, BC, Canada
Duration: Dec 7 2009Dec 10 2009

Other

Other23rd Annual Conference on Neural Information Processing Systems, NIPS 2009
CountryCanada
CityVancouver, BC
Period12/7/0912/10/09

Fingerprint

Semantics
Polynomials
Data storage equipment
Statistical Models

ASJC Scopus subject areas

  • Information Systems

Cite this

Ba, B., Sadamasa, K., Weston, J., Qi, Y., Grangier, D., Cortes, C., ... Mohri, M. (2009). Polynomial semantic indexing. In Advances in Neural Information Processing Systems 22 - Proceedings of the 2009 Conference (pp. 64-72)

Polynomial semantic indexing. / Ba, Bing; Sadamasa, Kunihiko; Weston, Jason; Qi, Yanjun; Grangier, David; Cortes, Corinna; Collobert, Ronan; Mohri, Mehryar.

Advances in Neural Information Processing Systems 22 - Proceedings of the 2009 Conference. 2009. p. 64-72.

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

Ba, B, Sadamasa, K, Weston, J, Qi, Y, Grangier, D, Cortes, C, Collobert, R & Mohri, M 2009, Polynomial semantic indexing. in Advances in Neural Information Processing Systems 22 - Proceedings of the 2009 Conference. pp. 64-72, 23rd Annual Conference on Neural Information Processing Systems, NIPS 2009, Vancouver, BC, Canada, 12/7/09.
Ba B, Sadamasa K, Weston J, Qi Y, Grangier D, Cortes C et al. Polynomial semantic indexing. In Advances in Neural Information Processing Systems 22 - Proceedings of the 2009 Conference. 2009. p. 64-72
Ba, Bing ; Sadamasa, Kunihiko ; Weston, Jason ; Qi, Yanjun ; Grangier, David ; Cortes, Corinna ; Collobert, Ronan ; Mohri, Mehryar. / Polynomial semantic indexing. Advances in Neural Information Processing Systems 22 - Proceedings of the 2009 Conference. 2009. pp. 64-72
@inproceedings{1e45b058c5cd43ad80bda8dbb123f934,
title = "Polynomial semantic indexing",
abstract = "We present a class of nonlinear (polynomial) models that are discriminatively trained to directly map from the word content in a query-document or document-document pair to a ranking score. Dealing with polynomial models on word features is computationally challenging. We propose a low-rank (but diagonal preserving) representation of our polynomial models to induce feasible memory and computation requirements. We provide an empirical study on retrieval tasks based on Wikipedia documents, where we obtain state-of-the-art performance while providing realistically scalable methods.",
author = "Bing Ba and Kunihiko Sadamasa and Jason Weston and Yanjun Qi and David Grangier and Corinna Cortes and Ronan Collobert and Mehryar Mohri",
year = "2009",
language = "English (US)",
isbn = "9781615679119",
pages = "64--72",
booktitle = "Advances in Neural Information Processing Systems 22 - Proceedings of the 2009 Conference",

}

TY - GEN

T1 - Polynomial semantic indexing

AU - Ba, Bing

AU - Sadamasa, Kunihiko

AU - Weston, Jason

AU - Qi, Yanjun

AU - Grangier, David

AU - Cortes, Corinna

AU - Collobert, Ronan

AU - Mohri, Mehryar

PY - 2009

Y1 - 2009

N2 - We present a class of nonlinear (polynomial) models that are discriminatively trained to directly map from the word content in a query-document or document-document pair to a ranking score. Dealing with polynomial models on word features is computationally challenging. We propose a low-rank (but diagonal preserving) representation of our polynomial models to induce feasible memory and computation requirements. We provide an empirical study on retrieval tasks based on Wikipedia documents, where we obtain state-of-the-art performance while providing realistically scalable methods.

AB - We present a class of nonlinear (polynomial) models that are discriminatively trained to directly map from the word content in a query-document or document-document pair to a ranking score. Dealing with polynomial models on word features is computationally challenging. We propose a low-rank (but diagonal preserving) representation of our polynomial models to induce feasible memory and computation requirements. We provide an empirical study on retrieval tasks based on Wikipedia documents, where we obtain state-of-the-art performance while providing realistically scalable methods.

UR - http://www.scopus.com/inward/record.url?scp=84858732366&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84858732366&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9781615679119

SP - 64

EP - 72

BT - Advances in Neural Information Processing Systems 22 - Proceedings of the 2009 Conference

ER -