Alibaba Introduces Secretarial Virtual Assistant for Meetings
Chinese tech giant Alibaba has launched a new virtual assistant to handle transcription and other secretarial duties at meetings and conferences. The Tingwu AI builds on Alibaba’s speech recognition technology, though not the same vein as its Tmall Genie voice assistant for smart speakers.
Tingwu, Mandarin for “listen well,” performs real-time transcriptions of conferences and analyzes the conversation to produce summaries and action items once complete. Tingwu’s speech recognition is 98% accurate, according to China’s State Information Center Software Testing Center and uses machine learning to correct mistakes as it transcribes the speech, including noting and adding custom vocabulary. The AI employs a new voice recognition algorithm developed at Alibaba’s DAMO Academy and can distinguish 10 different voices in a single meeting, Alibaba claimed. The AI can understand English and 14 Chinese dialects in addition to standard Mandarin. Alibaba has already integrated the assistant into its DingTalk conference platform, using it to distribute notes after the end of the meeting.
“Underpinned by our end-to-end speech recognition solution, Tingwu’s accuracy and powerful features make it a perfect tool for businesses to host multi-location online and offline conferences and interviews,” Alibaba speech lab head Zhijie Yan said. “As enterprises continue to embrace the next generation of remote working, Tingwu is well-positioned to transform business meetings with productive and easy-to-use automated voice transcription capabilities.”
Voice AI for business meetings has quickly become an almost standard feature, boosted by the enormous spike in remote meetings during the COVID-19 pandemic. One of the big success stories in the U.S. has been Otter.ai. The startup went from just real-time transcription to bringing a suite of secretarial features to Google Meet, Microsoft Teams, Zoom and Cisco Webex after raising $50 million in February. There’s plenty of competitors, such as Australian voice AI developer Dubber’s Unified Call Recording for Zoom, and the money has been rolling in for startups. Observe.AI could raise $54 million, and Replicant could pick up $27 million in September. Meanwhile, Dialpad beat everyone with a $100 million raise in October, and Verbit closed a $60 million round in November. Alibaba has its eye on the international market for Tingwu as well.
“The functionality of Tingwu will continue to expand, as we are now developing a model that can process mixed-language speech,” Yan said. “We will also explore additional language options to prepare Tingwu for use in large multinational corporations and overseas markets.”