Personalised language learning experience, getting environmentally aware at the same time.


We've noticed that, when learning a new language, there are many programmes which teach vocab and grammar, so these skills are often learnt first, but yet people still lack the confidence to actually speak and our future life quality hinges largely on our ability to communicate. Our website addresses this.

Why we do what we what we do?

On a more factual note, it is projected that 2 billion more people will be living in this world by 2050. It is definitely not feasible to use the same way we had now to train teachers in order to deliver quality language education to the masses. Modern problems require modern solutions, and TalkPal addresses this precisely.

What it does

Our website gives you a way of practising speaking and gives you fast personalised feedback whilst you read interesting news articles. It acts as a buddy to listen to you and help you improve rapidly.

Users are able to choose a news article that interests them and record themselves reading it, or we can suggest something relevant to the environment.

They are then able to listen back to their recording and upload it to get feedback on which words were pronounced correctly and which words were not recognised. We are also currently adding a feature which reads the text back so the user can hear what they are aiming for.

How we built it

We build the project on Flask, a python microframework, using cloud computing solutions (can toggle to Google Cloud-Speech, Azure Speech services or Amazon AWS Transcribe) to decode the voice clips and evaluate your accuracy using a bespoke algorithm that our team has put much thought into it.

Challenges we ran into

The integration of frontend and backend. While we have some degree of experience in frontend and backend application development, we nevertheless ran into problems integrating the front-end to the back-end.

Integrating different bits of work produced by teammates into a single working solution.

Complying to the best practices in building a secure webapp. It is tempting to side step for quick and dirty development by coercing users to allow microphone access by the use of malicious scripts, but we understand this will backfire in a long run. Ultimately, we value the confidence and good faith that our users entrusted to us the very second they decided to try our products out.

Accomplishments that we're proud of

Put up a usable and functional product within 24 hours.

Learn a lot about developing practical applications using cloud computing solutions.

Building something that is practical today, potentially impactful in the future.

What we learned

The details of a voice clip, different encoding scheme for an audio file Signal processing of an audio file Cloud computing Flask Bootstrap framework Coding skills in Python, HTML Git Getting more familiar with Linux/Terminal/PowerShell commands Dependency management Best practices of writing codes, readability

Teamwork Collaboration and how to best convey our idea to teammates Patience

What's next for TalkPal

We made the fundamentals of the software modular and it is ready to integrate with other voice assistant or home solutions to deliver a more natural, in-person and seamless learning experience.

We looking to integrate it with Alexa, Google Assistant and Cortana to interpret the scores obtain by the learner and give appropriate correction, encouragement and feedback.

There will be a list collating the words you have not quite pronounced it right at the first try, so you can revise quickly if needed.

We also realised that learning is always more fun when you do it together. Soon, a community edition will be released, and users will have the chance to get competitive with their friends to see who can enunciate the most accurately, the one at the top of the leaderboard for three consecutive times will get a prize.

Our product will have the potential to work with major testing agencies, to deliver quality education in a new way to the humans of tomorrow.

Share this project: