2 of the 3 of us experienced having to learn a new language in a different environment. We are aware that this can be really tough especially if we only begin learning past a certain age.

Furthermore, the increasing demand and importance for Chinese-speaking workers in the US is inevitable. Government policies such as Lead with Languages and 1 million strong initiative are all signs that show the growing need to ease the language gap amongst citizens and foreigners. Mandarin remains the top 3 most spoken knowledge and more definitely needs to be done to ensure the language-learning resources available meets the growing requirements of the language.

This was what inspired us to work on the current resources available and target gaps that have been unmet.

What it does

Our Mandarin-learning platform allows users to scan any item to be identified in Mandarin. This can be done on the go, as and when the user wants to learn more about a certain item. Basically, we are a Mandarin dictionary for images but the hope is also to include more learning aspects to this function.

How I built it

Adapted IBM Watson's vision recognition to an IOS application that helps users learn Mandarin through their surroundings.

Challenges I ran into

We had trouble with implementing Google Cloud API . Our Mac #1 was unable to download pip and Mac #2 couldn’t figure out the path. We then seeked help from 2 mentors but the issue was not resolved.

We tried 2 alternative methods: Pillow + Tensor Flow (Alternate to Path) but this did not work as well. After seeking help with the speakers at Machine Learning intelligence workshop – they suggested for us to use IBM Watson’s Machine Learning System.

With further consultations with Ming Li, and very helpful ones we were able to follow through with majority of our project. When we finally managed to install the scanning system, we realized we did not have a very accurate database of pictures. This made the scanner predict the wrong item. We iterated by adding more pictures to the sytem to improve its accuracy.

Accomplishments that I'm proud of

We are all very proud of the fact that we failed so many times, over and over again but yet still managed to find a way to get what we want. We approached at least 6 different people and though we may not necessarily have gotten the issue solved every time, we learnt something new every step of the way.

What I learned

We learned that there is no hard and fast rule to solving any issue as there is always another API or language or system that can be served as an alternative route. It is always important to iterate and relook our ideas, critically analyse the situation and consider all other options available.

What's next for MandarENG

We envision for MandarENG to be the bilingual buddy that every one has in future. This buddy will be there to assist users with any requests (scans) on the go. The hope is to continue to build on the features that are currently in place (scanner) and provide users with a more all-rounded language-learning journey.

This will include voice recognition technology to encourage users to speak and practice the language and a social platform that would be the best place to make international companions. We aim to not just be a platform that performs translations but one that unites people and creates greater social value for the community.

Built With

Share this project: