Detecting Interrogative Utterances with Recurrent Neural Networks

TL;DR

This paper investigates neural network architectures for detecting interrogative utterances, comparing regularization, context functions, gated activations, and multimodal inputs to improve classification accuracy on speech datasets.

Contribution

It introduces a comprehensive comparison of neural network components and methods for question detection in speech, including multimodal input integration.

Findings

Gated activation functions improve classification accuracy.

Regularization methods impact model performance.

Multimodal inputs enhance detection results.

Abstract

In this paper, we explore different neural network architectures that can predict if a speaker of a given utterance is asking a question or making a statement. We com- pare the outcomes of regularization methods that are popularly used to train deep neural networks and study how different context functions can affect the classification performance. We also compare the efficacy of gated activation functions that are favorably used in recurrent neural networks and study how to combine multimodal inputs. We evaluate our models on two multimodal datasets: MSR-Skype and CALLHOME.