Reverie Language Technologies Sets New Benchmark in Automated Speech Recognition

Reverie Language Technologies has achieved significant improvements in the accuracy of its Automated Speech Recognition (ASR) models, validated using the comprehensive Kathbath dataset. Reverie's ASR solutions exceed industry standards and are tailored for diverse sectors. Their new models offer superior performance, privacy, and offline functionality while supporting multiple Indian languages.

Bengaluru, India, Friday 19th July 2024 – Reverie Language Technologies, a pioneer in language technology solutions specializing in Indian languages, has announced a significant advancement in their Automated Speech Recognition (ASR) model accuracy. The validation comes from the esteemed Kathbath dataset, curated by AI4Bharat.

Reverie benchmarked their ASR output with Google and Microsoft for Indian English, Hindi, Gujarati, Marathi, Tamil, Telugu, Bengali, Kannada, Punjabi, Malayalam, and Odia. The Kathbath dataset, known for its extensive speech data spanning 12 Indian languages, involves contributions from 1,218 individuals across 203 districts, making it a gold standard for ASR systems evaluation in India. According to Soham Bhattacharya, API Product Manager, this milestone reflects the labor and innovation of Reverie's voice R&D division, setting a new industry benchmark.

Reverie, with more than 15 years in the industry, continues to lead in ASR accuracy and offers customizable speech recognition models tailored for sectors including Banking, Finance, Insurance, e-commerce, Automotive, and Government. The company's solutions cater to both online and offline applications, ensuring data security and comprehensive functionality even in unreliable internet environments.

The company's innovation extends to real-world use cases in e-commerce and the automotive industry, enhancing customer engagement and in-car voice assistance with multilingual support. Reverie's ASR technology stands out for its natural language understanding and streamlined performance, contributing significantly to both user experience and safety.

(With inputs from agencies.)