Generative and discriminative algorithms for spoken language understanding

TLDR

Spoken Language Understanding for conversational systems extracts concepts and relations from spontaneous speech, and as dialog systems grow more complex, it requires richer features such as a‑priori knowledge, long‑range dependencies, dialog history, and system belief. This study investigates generative and discriminative methods for sentence segmentation and concept labeling in SLU. The authors evaluate Finite State Transducers, Support Vector Machine sequence classifiers, and Conditional Random Fields on the ATIS and MEDIA corpora, comparing them on concept accuracy, generalization, and robustness to annotation ambiguities. Conditional Random Fields, which can incorporate non‑local non‑lexical features such as a‑priori knowledge, achieved the highest accuracy across both tasks.

Abstract

Spoken Language Understanding (SLU) for conversational systems (SDS) aims at extracting concept and their relations from spontaneous speech. Previous approaches to SLU have modeled concept relations as stochastic semantic networks ranging from generative approach to discriminative. As spoken dialog systems complexity increases, SLU needs to perform understanding based on a richer set of features ranging from a-priori knowledge, long dependency, dialog history, system belief, etc. This paper studies generative and discriminative approaches to modeling the sentence segmentation and concept labeling. We evaluate algorithms based on Finite State Transducers (FST) as well as discriminative algorithms based on Support Vector Machine sequence classifier based and Conditional Random Fields (CRF). We compare them in terms of concept accuracy, generalization and robustness to annotation ambiguities. We also show how non-local non-lexical features (e.g. a-priori knowledge) can be modeled with CRF which is the best performing algorithm across tasks. The evaluation is carried out on two SLU tasks of different complexity, namely ATIS and MEDIA corpora.