English Deep South Call-Center dataset

Home » Case Study » English Deep South Call-Center dataset

Project Overview:

Objective

The primary goal of our project, “Enhancing Customer Service through Voice Analytics,” is to develop a comprehensive dataset focused on the English Deep South dialect. This dataset will be instrumental in training advanced speech recognition models to better understand and process customer interactions in call centers. This enhanced understanding will lead to improved customer service experiences and more efficient call center operations.

Scope

Our project encompasses the meticulous collection and annotation of call center voice recordings. These recordings are derived from various sources, ensuring a rich and diverse dataset that accurately represents the unique linguistic characteristics of the English Deep South dialect.

Sources

Call Center Recordings: Direct collection of real-world call center interactions.
Scripted Conversations: Working with voice actors to simulate call center scenarios.
Public Domain Resources: Incorporating publicly available recordings to broaden our dataset.

Data Collection Metrics

Total Voice Recordings Collected: 25,000
Call Center Recordings: 15,000
Scripted Conversations with Voice Actors: 7,000
Public Domain Recordings: 3,000

Annotation Process

Stages

Dialect Identification: Annotating each recording with specific linguistic features of the English Deep South dialect.
Contextual Tagging: Classifying the recordings based on the context of the conversation – customer queries, responses, etc.
Metadata Documentation: Logging detailed metadata including call duration, speaker demographics, and linguistic nuances.

Annotation Metrics

Recordings with Dialect Labels: 25,000
Contextually Tagged Recordings: 25,000
Metadata Entries: 25,000

Quality Assurance

Stages

Annotation Review: Engaging linguistic experts to verify the accuracy of dialect annotations.
Data Cleansing: Rigorous screening to exclude any recordings that do not meet our quality standards.
Data Security and Privacy Compliance: Ensuring the confidentiality and integrity of the recordings, in line with regulatory requirements.

QA Metrics

Annotation Review Cases: 2,500 (10% of total)
Data Cleansing: Systematic removal of non-compliant or poor-quality recordings.

Conclusion

The “English Deep South Call-Center Dataset” is a pivotal resource for revolutionizing customer service in call centers. By providing a richly annotated and diverse dataset, it enables the creation of sophisticated speech recognition models tailored to understand the nuances of the English Deep South dialect. This initiative not only enhances customer interaction quality but also paves the way for more culturally inclusive technology solutions in the field of voice analytics.