Center for Language Engineering (Smart Bots)
Local Language Speech Interface for Banks
Design Development Implementation
A local language speech interface for banks via smart bots for customer engagement which utilizes speech recognition for local languages of Pakistan.
Tech
- Telephony Framework FreeSwitch
- Asterisk
- Dialog System Galaxy
- Ravenclaw
- Automatic Speech Recognition Kaldi
- Deep Learning Toolkits
Challenges
CLE in collaboration with Higher Education Commission (HEC) wanted to build a speech-based smart bot system for the banking sector. The aim was to create a coherent and streamlined customer engagement system for resolving consumer queries. Currently, the banks utilize call centers or dual-tone multi-frequency signaling (DTMF) systems. Overall costs, customer churn and human errors in query resolution can be significantly reduced with a speech-based smart bot.
Solution
- Ideation, Conception and Execution of Smart Bot for Engagement
- Speech Interface in English and Local Languages
- Active Noise Cancellation for Clear Voice Input
- Smart Bot Voice Analysis for Tone and Accent
- Context Analysis for Speech-to-Text
- Consumer Sentiment Analysis for Mindful Responses
- Easy Integration and Implementation on Existing System
OUR VALUE PROPOSITION
VF partnered with CLE to create a speech-based smart bot system for the various South Asian languages. This is a pioneering technology and a first in the field of local languages. There has been no integration and implementation of such local languages in any smart bot system in the world. The AI aspect is capable of analyzing sentiment via tone and accent of the user, which differs in the local languages in comparison to widely-spoken English. The smart bot can identify context via speech-to-text even in code-switching where the user speaks English and local language interchangeably within a single context.