WIP - Work in Progress

Inspiration

With 210 countries and territories being quarantined due to the COVID-19 disease outbreak, a sheer number of both pupils and students are now acquainted with day-to-day online classes. However, during the past months, they have also encountered the weaknesses of the e-Learning experience. At the root of the problem, we found several factors:

  • reduced motivation;
  • difficulties in keeping track of one's progress;
  • lack of an efficient way to quickly comprehend the information that you don't understand;

After working with several apps and platforms we found ourselves unsatisfied and pursued our attempt regarding those issues, coming up with a pretty innovative solution.

What it does

We aim to create an online learning platform that immerses the users and helps to keep them engaged. The web app will feature several tools to aid with this:

  • a virtual classroom to upload and store learning materials and resources;
  • abbreviation fast hyper-link dictionary;
  • user analytics under the form of ratings and statistics in regards to each user’s focus level;
  • a pop-up quiz system;

These systems will come under ELLA (ELectronic Learning Assistant) and will provide all of this functionality as a complement to a teacher running their lecture to students, on any given subject.

But the project's most important feature consists of ELLA being a bot programmed to learn the lectures and materials to aid the student, providing answers to custom questions related to the subject, at the touch of a button.

Of course, Artificial Intelligent and content based Chatbots have long been used allowing a form of interaction between mankind and a robot that is programmed to work independently from a human operator and to answer questions in a natural language. Amazon’s Alexa, Apple’s Siri, and Microsoft’s Cortana are just a few examples of chatbots most of us have probably taken advantage of.

Each student learns and absorbs things at a different pace and requires new teaching methods now and then. Consequently, as Artificial Intelligence is rapidly developing Chatbots could become an essential part of the e-Learning environment.

If the bot won't be able to respond, due to lack of information or limited understanding of the question, the person in need will be directed to a chat where the lecturer in charge of the course will provide an answer. The bot will then use the question and the lecturers answer to enrich its database, being able to accommodate support for future individuals having a similar dilemma and thus saving precious time for the teacher.

How we built it

During these hard times, as learning and collaboration move more and more online in response to the COVID-19 pandemic, developers need to build future proof solutions (microservices to enable distance learning), easy to plan, code and deploy to bring quality over a short period in an agile way.

As part of that learning, instructors need to be able to assess their students’ understanding of course material, reflect on their decision-making methods, learn from their answers so that we can train and build an AI entity that will be able to assist students 24/7, most of the time when students cannot get help from their teachers.

The typical chatbot can answer simple questions, such as small pieces of information. When a question falls outside of the scope of the pre-determined training set, the option is typically to tell the student that the question isn’t valid or it will offer to speak to a real teacher.

In this code pattern, we provide another option. If the student question is about something that is not previsioned in the training set, we use the search skill feature of Watson Assistant to pass the question on to the Watson Discovery service, which has been pre-loaded with the different school manuals. So now, instead of “Would you like to speak to a teacher?” we can return relevant sections of the subject manual to help solve the students’ problems.

"Cloud AI Architecture

To take it a step further, we use the Smart Document Understanding feature of Watson Discovery to train it on what text in the class manual is important and what is not. This improves the answers returned from the queries.

Further on, we plan to build a simple virtual classroom application to assess learner understanding using quiz forms. A major benefit of the app is its flexibility: this starter kit can easily be adapted into a short essay app, a grading app, or other educational tools, proving easy to plan, code, and future proof principles.

"General Architecture"

Loopback is an open-source tool for quickly building a data API for your applications. Whatever your specific application’s purpose, Loopback gets you quickly writing application logic instead of data-handling code.

React has been designed from the start for gradual adoption, and you can use as little or as much React as you need. Whether you want to get a taste of React, add some interactivity to a simple HTML page, or start a complex React-powered app.

React has proven to be an invaluable tool in creating the following - Here you can see the homepage of the Electronic Learning Assistant, featuring a fully-functional PDF Viewer, upload document option, a rating system and some work setting up Google Analytics for a potential gathering of User Metrics, but more on this topic later.

"HomePage"

IBM Cloudant® is a distributed database that is optimized for handling heavy workloads that are typical of large, fast-growing web and mobile apps. Available as an SLA-backed, fully managed IBM Cloud™ service, Cloudant elastically scales throughput and storage independently.

In conclusion, the following code pattern will:

  • Create a student dialog skill in Watson Assistant;
  • Use Smart Document Understanding to build an enhanced Watson Discovery collection;
  • Create a Watson Assistant search skill that allows the Assistant dialog to post queries to Watson Discovery;
  • For further releases, we can have Watson analyze quiz responses to train himself on its own.
  • Implement the virtual classroom concept based on the architecture that we have created.

Challenges we ran into

The main challenge we ran into while implementing the idea was the infrastructure requirements, costs, and integration, and that is all because of a lack of funds. Since we are in the developing phase, the traffic will be little to none while testing the integration and how ELLA is trained, but for different futures that the AI offers you need a paid plan. For example, we have the Search future, this enables IBM Watson to search the internet for references to a given subject, further training to have resulted from user-specific search queries, but, this is a PLUS paid plan, so it wasn't available for us to use.

Another thing that did not work out as we wanted it to was the implementation of state management using the Redux flux framework. It proved to be an overkill method of achieving state manipulation for our project, which is relatively small in scope on this front.

alt text

Any other challenges we have faced during the developing phase of the project were mainstream problems that every team has, such as remote communication, internet problems, technical debt, and so on.

Accomplishments that we're proud of

We are proud of the research we have done in the past two days and we are very proud of the solution we have compiled in the last 48 hours. It was not possible to do this working independently, the team effort was tremendous and all the advice from our mentors was priceless and worth the time of the discussion.

We did stick to best-practices guidelines in terms of coding despite the time-crunch, and we followed the Atomic Design Methodology architecture, using atoms, molecules, organisms, templates, and pages. This turned our initial prototype into a scalable, reliable, and easily expandable project by the time the new features were added.

This was the final front-end project structure:

"StructurePart1" "StructurePart2"

Moreover, we are proud of the support of the community, the support of our teachers, and our fellow college students. Their help and ideas were a good starting point.

What we learned

We learned more about Cloud Infrastructure, AI, Cloud Continuous Delivery, and AI Training, all within IBM Cloud premises. IBM Cloud offers a great variety of futures on the lite plan which is free. We consolidated remote working communication and we developed a bond between us as a team.

We've polished our react and ES6 skills, thus mastering the use of external component libraries to create a project such as this one in record time.

Lastly, we developed our agile ways of working, met exciting people, and exchanged different ideas on how virtual assistants will shape the future of education with other participants to the hackathon. We are grateful for the opportunity!

Impact on society

The impact that ELLA can have on students is proportional to the number of educational settlements that join the platform, thus students being able to access more and various information. It will increase the effectiveness of the student training by providing rapid answers to any inquiries. Leaving teachers with fewer questions but with more time. The additional quiz feature will provide a comfortable but competitive environment.

Although ELLA was created to facilitate higher education, it can be extremely useful for high schools, middle schools, and even primary schools. Following the business idea, the scale-up potential is extensive, ranging from the 101 accredited universities from Romania (our country of origin) to over 25000 world wide.

We as a group believe that humans do not stop learning after they finish school, but the learning extends to their professional life as well, thus a variant of the ELLA platform will be available for companies and the masses as soon as a comprehensive database of knowledge will be created. Companies will be able to use it for employees' continuous learning.

What's next for ELLA

As time passes, ELLA will grow and expand its knowledge as it gathers more and more information and feedback. We aim to help our fellow students and lecturers interact with one another through AI-based learning. ELLA will be your teacher, your companion, and your friend. The “know-it-all” kind of friend that nobody likes…

Moreover, our goal is to make kids of all ages fall in love with learning, and we know competitiveness is part of human nature, so we will develop a non-compulsory ranking system that will award users points and achievements for every task that they do successfully.

It's clear for us and we believe that we persuaded you to understand that our project is very useful during this unfortunate period, but its usefulness doesn't stop here and it can become a staple in modern times education.

ELLA can be further improved by finishing up the Google Analytics module started in this hackathon and featured in the image below. It would provide a fast feedback loop, as well as a good insight into the user's trends and preferences, which would influence the development of features going forward. Then, a select few metrics could be spooled into statistics to show the user, signaling potential improvement areas.

"GoogleAnalytics"

As we know, in the AI world nothing is impossible, so the sky's the limit for ELLA to grow into one of the best learning companions that everybody wished they had.

Business Plan Canvas

"Business Plan Canvas"

Summary of the business plan canvas

The problem

The main problem is online learning can sometimes be difficult. The student usually has questions about the subject during a course or while reading a course text and needs more clarification.

How we solve the problem

ELLA can become a useful tool for students to provide a reliable single source of truth for their information.

Expenses

General expenses are to be made to make the solution run. The actual implementation of the prototype is using the Lite plans that IBM Cloud is offering, making the implementation of the solution idea easy to develop with no financial resources.

For further usage and expansion, it is clear that funding is needed since most of IBM Watson's advanced training methods are available only to paid plans. The pricing plans are scalable horizontally or vertically depending on the customer needs.

Sources of income

The prototype that we have built is free of charge. The main source of income that we see, in our vision, is based on user donations. We don’t see fit to charge students.

+ 130 more
Share this project:

Updates