Microsoft Studying Multilingual Speech to Humanize Cortana – Investopedia



With the market for voice-powered gadgets forecasted by International Data Corp. to grow at a compound annual growth rate of 54.4% through 2020, companies are racing to improve the technology behind it.

Microsoft (MSFT) researchers are studying the speech of bilinguals in an effort to compose Cortana sound more human and communicate effectively. According to a report from CNBC, the company is particularly interested in code-mixing, a practice in which speakers proceed back and forth between different languages in a sentence or during a conversation. Dubbed Project Mélange, a team in India are looking at how virtual assistants can learn to reply to a user switching languages in the middle of a conversation. CNBC distinguished code-mixing often happens in societies where people speak more than one language. “Compartmentalization of mixed languages (by multilinguals) maintain gone absent with each coming generation,” Kalika Bali, a researcher at Microsoft, told CNBC in an interview. “So younger people exhaust mixed languages in increasingly phases of their lives.” Bali distinguished with Cortana, Microsoft’s digital voice activated assistant, has to understand how its users communicate and respond in kind. (See more: Amazon Releases Echo demonstrate)

The researchers are currently looking at the different things that proceed into code-mixing, from speech to understanding and recognizing several languages. The group is also investigating how the practice changes between generations and why people are prompted to switch languages during the same conversations. They found, for example, sometimes it’s done to interject humor or when a person is changing the topic. It could pick years for Cortana to switch between languages, distinguished CNBC. “Trying to solve just mono-lingual natural language processing and understanding has taken years. Now we’re talking approximately mixed stuff and where are we going to glean data from?” Bali said in the interview. “Because everything is data-driven. It just seemed like a very difficult problem to solve but we were indecent so excited approximately this that we just went on.”

The Redmond, Washington software maker’s push to compose Cortana more human-like comes a time when the company is trying to increase adoption of its voice-activated personal assistant that competes with the likes of Amazon’s (AMZN) Echo speaker powered by Alexa, Apple’s (AAPL) Siri and Samsung Electronics (SSNLF) Bixby. Earlier this month,Microsoft teamed up with Harman International to roll out a speaker that is powered by its virtual assistant. (See more: Microsoft, Harman Team for Rival to Amazon Echo)



Source link