A Context-based Approach for Dialogue Act Recognition using Simple Recurrent Neural Networks

A Context-based Approach for Dialogue Act Recognition using Simple Recurrent Neural Networks Chandrakant Bothe, Cornelius Weber, Sven Magg, and Stefan Wermter Knowledge Technology, Department of Informatics, University of Hamburg, Vogt-Koelln-Str. 30, 22527 Hamburg, Germany http://www.informatik.uni-hamburg.de/WTM/ {bothe,weber,magg,wermter}@informatik.uni-hamburg.de Abstract Dialogue act recognition is an important part of natural language understanding. We investigate the way dialogue act corpora are annotated and the learning approaches used so far. We find that the dialogue act is context-sensitive within the conversation for most of the classes. Nevertheless, previous models of dialogue act classification work on the utterance-level and only very few consider context. We propose a novel context-based learning method to classify dialogue acts using a character-level language model utterance representation, and we notice significant improvement. We evaluate this method on the Switchboard Dialogue Act corpus, and our results show that the consideration of the preceding utterances as a context of the current utterance improves dialogue act detection. Keywords: Dialogue Acts Detection, Recurrent Neural Networks, Context-based Learning 1. Introduction In natural language processing research, the dialogue act (DA) concept plays an important role. Its recognition, in most cases, is considered a lexical-based or syntax-based classification at utterance-level. However, the discourse compositionality is context sensitive, meaning that the DA of an utterance can be elicited from the preceding utter- ances (Grosz, 1982). Hence, classifying only utterances is not enough because their DA class arises from their con- text. For example, the utterance containing only the lexical entry ’yeah’ might appear in several DA classes such as Backchannel, Yes-Answer, etc. For certain DA classes, the utterances are short, and most of them share similar lexical and syntactic cues (Jurafsky et al., 1998). The aim of this article has two subgoals: first, we investi- gate the annotation process of DA corpora and review the modelling so far used for DA classification, and second, we present a novel model and compare its results with the state of the art. We propose to use context-based learning for the identification of the DA classes. First, we show the results without context, i.e., classifying only utterances. Including context leads to 3% higher accuracy. We use a simple re- current neural network (RNN) for context learning of the discourse compositionality. We feed the preceding and cur- rent utterances to the RNN model to predict its DA class. The main contributions of this work are as follows: - We provide detailed insight on the annotation and mod- elling of dialogue act corpora. We suggest to model dis- course within the context of a conversation. - We propose a context-based learning approach for DA identification. In our approach, we represent utterances by a character-level language model trained on domain- independent data. - We evaluate the model on the Switchboard Dialogue Act (SwDA1) corpus and show how using context affects the results. For the SwDA corpus, our model achieved an accu- 1Available at https://github.com/cgpotts/swda racy of 77.3% compared to 73.9% as state of the art, where the context-based learning is used for the DA classification (Kalchbrenner and …

Content truncated to 500 words