Discovering Sound Concepts and Acoustic Relations In Text

TL;DR

This paper presents methods for extracting acoustic concepts from text and defining acoustic scenes, aiding in building an acoustic knowledge base and supporting acoustic event detection.

Contribution

It introduces novel pattern matching and neural network techniques for identifying sound concepts and modeling acoustic scenes from large text corpora.

Findings

Effective extraction of sound concepts from text

Successful prediction of acoustic scenes using neural networks

Potential applications in acoustic event detection

Abstract

In this paper we describe approaches for discovering acoustic concepts and relations in text. The first major goal is to be able to identify text phrases which contain a notion of audibility and can be termed as a sound or an acoustic concept. We also propose a method to define an acoustic scene through a set of sound concepts. We use pattern matching and parts of speech tags to generate sound concepts from large scale text corpora. We use dependency parsing and LSTM recurrent neural network to predict a set of sound concepts for a given acoustic scene. These methods are not only helpful in creating an acoustic knowledge base but in the future can also directly help acoustic event and scene detection research.