Recurrent Conditional Random Field for Language Understanding

kaisheng yao, Baolin Peng, Geoffrey Zweig, Dong Yu, Xiaolong Li, and Feng Gao

Abstract

Recurrent neural networks (RNNs) have recently produced record setting performance in language modeling and word-labeling tasks. In the word-labeling task, the RNN is used analogously to the more traditional conditional random field (CRF) to assign a label to each word in an input sequence, and has been shown to significantly outperform CRFs. In contrast to CRFs, RNNs operate in an online fashion to assign labels as soon as a word is seen, rather than after seeing the whole word sequence. In this paper, we show that the performance of an RNN tagger can be significantly improved by incorporating elements of the CRF model; specifically, the explicit modeling of output-label dependencies with transition features, its global sequence-level objective function, and offline decoding. We term the resulting model a “recurrent conditional random field” and demonstrate its effectiveness on the ATIS travel domain dataset and a variety of web-search language understanding datasets.

Details

Publication typeInproceedings
Published inICASSP 2014
PublisherIEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)
> Publications > Recurrent Conditional Random Field for Language Understanding