Hong Kong Economic Journal: AI presentation skill training platform: Observing emotions, contents & voice for detailed results in 10 minutes


AI presentation skill training platform: Observing emotions, contents & voice for detailed results in 10 minutes

By Hong Kong Economic Journal on July 20, 2021
The original article was published in the Hong Kong Economic Journal “StartupBeat”. View the story here

To speak confidently in front of a crowd, apart from repeated practice, it is equally important to know how to speak. Datality Lab, a research and data consulting start-up in Hong Kong, has invested over one million Hong Kong dollars to develop, an artificial intelligence (AI) speech training platform, which can analyze the performance of trainees from facial expressions, speech content, tone of voice and body movements, etc. Calculate your score and give a comment in as less as 10 minutes. Earlier this year, the platform cooperated with the City University of Hong Kong Business School to conduct the pilot and is now discussing with insurance companies to help the business and academic circles train speech professionals.

Traditionally, speech training is quite time-consuming. Roland, managing director of Datality Lab, said that in general one-on-one speech training, instructors need to analyze students’ performance in real time, which not only takes time but also costs a lot. The company has developed an intelligent speech training platform since 2019, hoping to objectively analyze the speech level with AI technology.

Before users log in to the platform or conduct training, they must pass face recognition authentication to ensure that the system analyzes the performance of the designated user; after that, they can select suitable scenarios according to their personal needs, such as interviews, sales reports, etc., then record practice videos, and upload them to Cloud for analysis.

Partner with City University of Hong Kong Business School for pilot and the result is positive
The platform uses Machine Learning and AI products of Amazon Web Services (AWS), including Amazon Rekognition, Amazon Comprehend, etc., it only takes 10 minutes to evaluate the speaker’s facial expressions, speech content, tone of voice and body movements, etc. There are also scores for confidence, fluency, credibility, energy level, and suggestions for improvement with each practice. The system uses decision science model to learn the tutor’s scoring method and simulate the scoring of real people. “We will also provide analysis of celebrity speeches, explaining what actions or content are made during the speech to attract the audience’s attention, and at the same time, we will analyze the mistakes of the speech. Teach how to avoid mistakes.” It is understood that the system has been recognized by some partners, such as Ms Cheung Wai Tsz, a senior cross-media media professional.

At the beginning of this year, cooperated with CityU Business School to test the system. About 50 post-graduate students of the school used the platform to prepare speeches for the presentation of their year final project, and conducted more than 300 recording practices in total. Analysis shows that if students practice repeatedly, their overall speech performance will be significantly improved. In addition to CityU, the company is discussing long-term cooperation with the Open University of Hong Kong, which will soon be renamed the “Hong Kong Metropolitan University”, and two insurance companies, and adding a customization plan, hoping to attract a wide range of academic and business circles’ application.

The company has been selected by the Hong Kong government’s Innovation and Technology Fund Enterprise Support Program, and has also received support from the Cyberport Incubation Program and the Creative Micro Fund. In June this year, it opened a branch in Singapore to enter the Southeast Asian market. Roland Leung pointed out that the system performs more than 170 calculations by observing 70 observation points of the user’s performance, such as head, forearm, palm, fingers, etc., which can adjust the evaluation criteria and the requirements of speech skills according to different national cultures and speech context to meet the needs of different markets.

It is planned to introduce wide-spread bilingual analysis by the end of the year
At present, the system is mainly in English. Roland Leung revealed that analysis in Mandarin and Cantonese will be added before the end of this year. Currently, the mobile application (App) is being developed. It is expected to be launched in the market in October this year. By then, only a smartphone will be needed to receive speech training.