Enhancing Customer Experience for Health Insurance: Tuning for Industry-Specific Vocabulary
When it comes to transcription, key terms are important for categorizing calls, identifying discussion topics, and assessing sentiment. Improving key term recognition can often have a greater impact on the delivery of actionable insights than generalized transcription accuracy improvement of a language model.
For example, one of the country’s largest health insurance providers — with nearly 15 million members — wanted to improve the outcomes of its automatic speech recognition (ASR) system. This included overall accuracy as well as an improved ability to identify industry-specific terms spoken by its customers and call center agents.
The challenge was that the provider’s ASR system had been automatically generating transcripts that did not include many key healthcare terms, such as “HMO”, “copays”, “colonoscopy”, and “rehab”, spoken by customers and call center agents. Without accurate transcriptions of industry-specific terms, the provider would not be able to improve its Net Promoter Score® (NPS®), which measures customer experience, predicts business growth, and reduces customer turnover.
The key to improving speech recognition for a specific enterprise is to tune the ASR system with audio captured from the enterprise’s telephone system. This ensures the tuning will result in better alignment between the ASR system and the specifics of what agents and customers are discussing, including their accents and speech patterns, and even noises and distortions peculiar to the telephone system.
By using Voci’s custom language models with its V-Blaze ASR system, the health insurance provider was able to quickly extract high-quality, actionable intelligence from vast quantities of call audio data. The provider could better understand what its customers and agents are saying during calls, as well as the sentiments and emotions they are expressing. This in turn has led to improved customer experience and reduced churn.
To find out more, click here to download the case study.