Welcome to Porjai - Thai Dialect Voice Recognition

The Thai Dialect Speech Recognition project serves as the first milestone towards creating a virtual assistant for Thai people, using our own dialects! Voice interface is quite intuitive and common to many people. Since the COVID pandemic has started, e-commerce platforms or online markets have become a central place for sellers to buyers to meet and do transactions with minimal physical contacts. Unfortunately, many people cannot take advantages of online markets just because they are not familiar with technology. Imagine that if you can talk to an application, and it understands what you would like to sell. This is our project goal – namely we will create and train an AI model that performs speech-to-text tasks for Thai and common Thai dialects.

In order to teach the AI models to recognize Thai language, including Thai dialect, we have to collect many voice samples which will be used to trained AI models to recognize speech in Thai and Thai dialects. In this phase, the model will be tuned and trained to understand the sentences with e-commerce contexts. Currently, we have several applications and campaigns to collect the voice samples. You can contribute your voice samples or learn more about our applications.

Our Thai dialect speech recognition project is supported by PMU-C (https://www.nxpo.or.th/C/). Our research team consists of the researchers, faculty members and students from the CMKL University, Chulalongkorn University (CU), Khon Kaen University (KKU), Prince of Songkla University (PSU), and Chiang Mai Rajabhat University (CMRU). We also would like to thank AIS to support the cloud resources to host our applications for the data collection, and Friends in Need(of "PA") Volunteers Foundation, Thai Red Cross for helping initiating the project to help people in local communities.

Get to know Nong Porjai

Say Hello to Nong Porjai 👋

“Porjai” translated as “satisfied” is the first advanced Thai artificial intelligence (AI) virtual voice assistant that can automatically detect and understand speech from different Thai regional dialects. A group of researchers is currently working on making Nong Porjai becomes the one and only Thai speech virtual assistant you’ll need. However, currently, Nong Porjai is only just a ‘baby,’ which means Nong Porjai isn’t fully grown yet. She can only perform a very limited specific task to translate any of your Thai speeches into Thai texts. Even though, for the time being, we are still only in the beginning phase (Phase I) of developing Porjai but we hope to grow Nong Porjai into a brilliant AI Thai lady that serves all Thai people across all Thai E-commerce platforms in the near future.

How Nong Porjai grows? 🌱

We plan to implement Porjai in 3 phases. Think of it like how we grow up ;)

• Phase I – a cute Thai baby
In this baby phase, Porjai can only do a simple Thai Dialect speech recognition task => She can only translate your Thai dialectic voices into texts and recognize what you want to sell in your own dialect.

• Phase II – a Thai teenage girl
Now Nong Porjai becomes a smarter teenage girl. This time, she can make the dialect speech recognition understand the context more than just posting a product => She can understand better on selling/buying or even general context surrounding.

• Phase III – a brilliant AI Thai lady
This is a final form of Porjai. Now Porjai has grown into a brilliant AI Thai lady that can respond back to users in natural language in your own speech => She becomes more of a virtual assistant and can get smarter and smarter as time goes by.

Talk to Nong Porjai. Help her grow 😉

Nong Porjai can automatically recognize voices and get the product information which is the first step to list a product on the online platform. The developer team has selected four major regional local Thai dialects, from northern, northeastern, southern, and central part of Thailand. Currently, our goal is to make the model recognizes local Thai dialects to help enabling the use cases for those Thai people who aren’t familiar with technology to access e-commerce platforms.

To create Porjai—an automatic speech recognition system for Thai dialects, we have collected voice samples from Thai locals who speak their own dialects to train the speech recognition models. We will publish the Thai dialect corpus that has been acquired in this project so that anyone can access and use this data for further development. So, no matter who you are. No Matter where you are from. Don’t be shy, talk to Nong Porjai – Help her grow ;)

The Future of Nong Porjai 🙌

“The future of Nong Porjai depends on all of us to teach her”

Currently, Nong Porjai is only a cute little baby. In order to make her grows into a beautiful, brilliant lady in the near future, we need all your support – meaning that every step of the way. The university will also share the trained speech recognition model and open the Thai dialect voice corpus so that other researchers and the communities can extend the usage and improve the model. This project serves as an inclusive technology initiative for all Thais and hopes to drive AI adoption and future development for society.

Research Sponsor

Research Team