Developing NLP-based English learning applications for the Vietnam market

Overview

The client is a rising player in the SaaS sector in Vietnam. Most English learning apps on the market are simple and repetitive. They set an ambitious goal to develop a family of NLP-based English learning apps for Vietnamese people. Each application in the family targets a different market: adults, children, schools & English centres. 

They wanted to develop an NLP mobile app that would automatically provide feedback to learners and adapt the learning process to their pace, encouraging them to acquire a new language as effectively as possible.

Key features 

  • NLP-based shadowing and modeling exercises
  • Personalized learning content based on daily conversation context
  • Timely reports to review learning progress
  • Registration on the mobile app or via SMS 
  • Premium account activation through codes 
  • English learning games, flashcards, songs and videos, and coloring exercises based on different age groups
  • Parent/Teacher accounts to monitor and support children’s learning 

Technical Stack

  • Android Native, iOS
  • Swift
  • Spring, Java, Servlet
  • PostgresDB
  • Ocean Digital

Project Overview

  • Team: 10 people
  • Market: Vietnam
  • Development Time: 3 months
gemvietnam

Background

The adoption of Artificial Intelligence in the education sector is increasing rapidly, thanks to its power to provide a personalized learning experience. It can analyze the learners’ ability and level to offer them a suitable curriculum. The applications are deployed on mobile platforms, enabling users to learn English beyond classrooms, anytime and anywhere.

The project’s highlight is the application of NLP technology to support shadowing and modelling exercises. The app can analyze the user’s speech to identify spelling mistakes and suggest improvements. Therefore, it can help users fix their pronunciation in a much shorter time and improve their fluency

Challenges

The initial development phase for the first app was done in just two months. Then we continue to develop the following apps in the family.

The problems actually occurred after the app was launched. The app received a number of users’ feedback saying that it was hard to remember their login details, as there were too many login options to choose from (Google, Apple ID, Facebook, Phone number, Email, etc.).

Another problem arose from having too many learning functions available, which sometimes confused users. The Google AI service used to read the sample speech also brought up the cost significantly.

Solutions

In order to resolve the login issue, our team realized that having fewer login options would bring a much more seamless experience for the user. The app now only has 2 login options.

The same applies to learning functions. We analyze the data and realize that we only need to focus on improving and prioritizing the most frequently used functions to enhance customer experience.

Regarding the cost for Google AI speech, we realized that if we replace it with pre-recorded speeches at this point, it can bring down the cost. This was a crucial update as the app was relatively new, and applying a more cost-saving option could help reach the break-even point sooner.

Impacts

With the successful launch of the apps for adults and children, the client was impressed with our delivery speed and quality. The app supporting schoolwork was also ready to launch and prepare for partnership with schools in Vietnam.

Currently, on Google Play Store, the app for adults has more than 5000 downloads with a 4.3 stars rating. On App Store, it was rated 4.6 by users. Meanwhile, the one for kids has more than 1000 downloads on Google Play Store, with a 4.6 stars rating on Apple and Google Stores.