Machine Learning
Education Technology
Anytime Anywhere Learning with Continuous Assessment
Project Principal Investigator(s): Sujit Kumar Chakrabarti, Manish Gupta
Emergence of online mode of learning (MOOC, blended learning) promises unprecedented democratization and scale-up of education through technology. Today’s teachers and learners have access to virtually unlimited amount of learning material from the best of the best sources.
Education mostly gets its edge and credibility when complimented by valid, reliable, deep and fair assessment. In this project, we are in the process of building cutting edge technology to automate assessment (generation, conduction and evaluation) and hence make it continuous and integrate into a system for personalized learning. Automated software engineering, and more recently in machine learning, along with sound knowledge of pedagogical theory are the three main building blocks of this technology. With VideoKen, we have also been working on automated tagging, search, question/answering and recommendation of videos, and social learning to enable a more engaging learning experience while utilizing on-line content. These aspects will be combined to build a highly sophisticated learning platform that supports the journey of each learner in a personalized manner. All of this research will be driven by real world requirements and a goal of demonstrating a high level of societal impact. From an early stage, we will deploy our platform across several educational institutions in Karnataka, and will drive towards eventually supporting millions of learners in Karnataka and beyond.
Automated Evaluation of Objective Questions
Although there are a lot of off-the-shelf systems for administering and evaluating objective type questions, they have their own set of issues. Some of these issues are as follows:
- They mostly come bundled with other large systems, e.g. learning management systems.
- They mostly are web-based systems. While web based deployment has its advantages, there are some distinct disadvantages as well.
- Evaluation algorithms can’t be tweaked (easily).
- Data often gets locked in the system, thus making conduction of further research with the data a hard task.
- Extensibility is very heavyweight effort if not impossible in all such environment.
Our system is a locally installation system. It is stand alone and small enough to be accessible to customizations as per need without external dependencies. It has already been used successfully in a number of quizzes by the author and his team. Source code and more details: EvalObj
