akeira™ - Virtual Assistant

Umesh Sachdev, Rajagopal Jayaraman, Zainab MillwalaInterspeech 2018

TL;DR

This paper introduces "akeira™," a commercially viable voice virtual assistant designed to enhance customer experience by providing natural, multi-turn conversational interactions. It aims to reduce call handling times, improve first-call resolution, and offer a more engaging, human-like customer service experience compared to traditional automated systems.

What Problem Does This Paper Address?

The paper addresses the limitations of existing automated customer service solutions, such as rigid IVR systems and basic chatbots, which often lead to frustrating customer experiences, long wait times, and escalations. Consumers today demand faster, more interactive, and anytime-anywhere access to solutions for their problems. The challenge is to create an automated system that can understand and respond to natural language queries in a way that mimics human interaction, providing a superior customer experience without the need for deep, pre-defined menu structures.

Methods

The core of "akeira™" is its natural language processing (NLP) engine, which allows it to understand and process queries in natural language. The system is designed to handle unstructured inquiries, meaning users don't need to navigate complex menus. It personalizes the brand experience by allowing customization of the virtual assistant's persona. The system is also highly scalable and compatible with existing IVR and mobile platforms. A key aspect is its ability to go beyond simple keyword matching by understanding complex grammar structures and even inferring user mood and emotions, enabling a more contextual and empathetic interaction.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Beyond Keywords: The system's strength lies in its advanced semantic processing, which moves beyond simple keyword matching to understand complex grammar and context, making interactions more natural.
  • Human-like Interaction: By incorporating elements like inflection, expressiveness, and emotions, "akeira™" aims to create a more endearing and connected human-machine relationship, differentiating it from robotic assistants.
  • Personalization is Key: The ability to personalize the virtual assistant to specific brand needs and customer service requirements is highlighted as a significant differentiator.
  • Efficiency Through Conversation: By enabling users to ask questions in their own words and receive direct answers, the system significantly reduces call handling time and improves first-call resolution rates.

Key Results

The paper emphasizes the benefits of "akeira™" rather than presenting specific quantitative results from a benchmark study. The key outcomes highlighted are:

  • Reduced Call Handling Time: By providing direct answers to natural language queries.
  • Improved First Call Resolution: By accurately understanding and addressing consumer issues.
  • Enhanced Customer Engagement: Through a more natural and personalized conversational experience.
  • Cost Reduction: By increasing call capacity and improving infrastructure utilization, reducing misroutes and transfers.
  • 24x7 Availability: Acting as a constant online research agent for consumers.
  • Support for Global Languages: Capable of supporting over 80 global languages and more than 100 accents/dialects.
  • Secure Authentication: Utilizes voice biometrics for security.

Limitations and Future Work

Not discussed in the paper.

Summary updated

BibTeX
@inproceedings{sachdev2018akeiratm,
  author = {Umesh Sachdev and Rajagopal Jayaraman and Zainab Millwala},
  title = {{akeira™ - Virtual Assistant}},
  booktitle = {Interspeech},
  year = {2018},
}
Figure