This Paper Had the Smartest Reviewers -- Flattery Detection Utilising an Audio-Textual Transformer-Based Approach

TL;DR

This paper introduces a multimodal approach combining audio and textual transformers to automatically detect flattery in speech, achieving high accuracy on a novel dataset for improved human-AI interaction naturalness.

Contribution

It presents a new multimodal dataset and a transformer-based model that effectively integrates audio and text for flattery detection, advancing the state of the art.

Findings

Audio-only detection achieves 82.46% UAR

Text-only detection achieves 85.97% UAR

Multimodal detection achieves 87.16% UAR

Abstract

Flattery is an important aspect of human communication that facilitates social bonding, shapes perceptions, and influences behavior through strategic compliments and praise, leveraging the power of speech to build rapport effectively. Its automatic detection can thus enhance the naturalness of human-AI interactions. To meet this need, we present a novel audio textual dataset comprising 20 hours of speech and train machine learning models for automatic flattery detection. In particular, we employ pretrained AST, Wav2Vec2, and Whisper models for the speech modality, and Whisper TTS models combined with a RoBERTa text classifier for the textual modality. Subsequently, we build a multimodal classifier by combining text and audio representations. Evaluation on unseen test data demonstrates promising results, with Unweighted Average Recall scores reaching 82.46% in audio-only experiments, 85.97% in text-only experiments, and 87.16% using a multimodal approach.