May I Ask Who's Calling? Named Entity Recognition on Call Center Transcripts for Privacy Law Compliance

TL;DR

This paper develops a specialized Named Entity Recognition system for call center transcripts, addressing unique challenges of spontaneous speech and recognition errors to improve privacy law compliance.

Contribution

It introduces a new annotated corpus, custom contextual embeddings, and a BiLSTM-CRF model tailored for call center conversation analysis.

Findings

Achieved state-of-the-art NER performance on call center transcripts.

Effectively handled noisy, spontaneous speech and recognition errors.

Demonstrated the model's potential for privacy compliance applications.

Abstract

We investigate using Named Entity Recognition on a new type of user-generated text: a call center conversation. These conversations combine problems from spontaneous speech with problems novel to conversational Automated Speech Recognition, including incorrect recognition, alongside other common problems from noisy user-generated text. Using our own corpus with new annotations, training custom contextual string embeddings, and applying a BiLSTM-CRF, we match state-of-the-art results on our novel task.