Fano Labs AI converts 30,000 calls to text in one night

An interview with Singtao Daily

Fano Labs AI converts 30,000 calls to text in one night
January 11, 2022
News Articles

Converting phone recording to text 30,000 audio files in one night

When you call the customer service hotline, you will always hear "In order to ensure the quality of service, the content of the conversation may have been recorded." When there is a dispute in communication between the two parties, or a violation of laws and regulations must be investigated, companies or regulatory agencies often spend a lot of time re-listening the content. Fano Labs uses cutting-edge artificial intelligence technologies such as speech recognition (ASR) and natural language processing (NLP) to convert telephone answers and recordings into text. It can be used to convert 30,000 voice files into text in one night to distinguish how many People speak and analyze content. Some regulatory agencies have used the service, and technology can help find who is the real scammer.

Recognise Cantonese, English and Mandarin

Dr. Miles Wen, the founder and CEO of Fano Labs, and Dr. Albert Lam, the chief scientist and chief technology officer, were previously engaged in R&D and teaching at the University of Hong Kong, and later moved to the Science Park to commercialize R&D. Miles pointed out that the voice analysis system "Callinter" uses voice recognition technology, natural language processing and big data analysis, and uses massive voice data for deep learning. It can accurately recognize Mandarin, English, Cantonese and Sichuan dialect. The system can analyze the answering calls, and can distinguish how many people have spoken, and sort them into text. Hong Kong people are accustomed to mixing Chinese and English. He pointed out that even if Chinese and English are mixed, and Mandarin plus English, the accuracy of the system is as high as 85 to 95%.

In recent years, there have been stock trading groups on social media and communication software from time to time, including the "sing high bulk" scam, which induces investors to buy specific stocks at high prices, and then scammers sell them at high prices or "sell" them for profit.  Miles said that through the system, even the voice can be used to distinguish how many people are talking, and analyzing the content of the words helps to find out who is the real liar in the "sing high bulk" scam.

This technology was previously won in the "Global RegTech Technology Challenge" held by the HKMA, and Fano Labs won the championship in the "Ethics and Customer Protection" category.