AIS teams up with Thailand AI Research Institute showcasing AI platform of Thai Speech Emotion Recognition in world-first tool, taking Thai companies to another level to deal with current challenges
AIS has joined the Thailand AI Research Institute (AIResearch) to showcase the success of Thai researchers after just two years to develop the first dataset for Thai Speech Emotion Recognition (Thai SER) in the world. This has been achieved by developing AI technology in combination with Deep Learning, applied to speech to understand emotions in sound to create Thai SER. It has been developed from the third-largest dataset in the world, after the first and the second which are in English. The data was derived from voice datasets generated by podcasts and YouTube, and can be deployed to increase service efficiency to better meet the needs of customers. Users can study more details and download the Thai SER data set for further development for free from the website https://airesearch.in.th from now onwards.
Mr. Sanjay Andrew Thomas, Chief Information Officer at AIS said that “AI technology for natural language processing is now being more widely used and no longer just in English or Mandarin. It is growing to encompass languages like Thai also.”
“For this reason, AIS as a digital service provider, sees the priority of developing Thai language AI structure, particularly in the matter of emotional recognition of the Thai language.”
“This project is the first of its kind in Thailand and only made possible due to the large size of the Thai language dataset which makes more accurate machine learning possible.Therefore the cooperation between AIS and AIResearch is the starting point of creating an ecosystem of Deep Tech in Thailand.”
“From here, it can grow strongly and be the equal of other countries. We have knowledgeable and expert human resources to create innovations which are beneficial to Thai people from every sector, to give them competitive opportunities on the world stage.”
Speech Emotion Recognition (SER) is a highly innovative application of AI/ML which is part of a group of exponential technologies like Cloud, IoT and 5G that are changing business models and driving reinvention.
“We now have the ability to detect voice emotion in a customer call and we can thereby take corrective actions and have our human and non-human agents deal even more empathetically with our customers.”
Breakthroughs like SER are how we will continue to dominate the Thai marketplace, maximise our customers’ experience and drive AIS towards becoming a Cognitive Enterprise.”
The Thai SER dataset can distinguish five emotions from voices, namely, anger, sadness, happiness, irritation and normalcy, with an accuracy of 70%, and it is 80% ready for use. The remaining 20% is for the basic model to give even more accurate analyses and be ready for use in real working environments in various companies, where it must be further developed in each context.
Examples of using the dataset for Thai SER
in the real world include:
- Analysis of the customer’s emotion from voice
- Ranking the level of customer service from the customer’s emotion
- Assigning the customer line to CSR, according to capability in dealing with various emotions
- Training and preparation for CSR, for real operations
- Added capabilities to understand situations and increase opportunities to close sales
Currently, the Thai SER dataset is ready for developers, startups, SMEs, service providers and other organisations to make additional studies and use the platform to design applications
and solutions supporting Thai language AI at no extra cost. It can be accessed from the website