Skip Ribbon Commands
Skip to main content

Maximising Learning Through Algorithms

Maximisinglearningthroughalgorthms-730.jpg

Creating a list of questions to help students reinforce and apply their learning is a crucial part of any course design. Now, scientists at the EEE have developed an algorithm to help instructors to automatically label written questions and thus optimise their curriculum.   

“Instructors normally generate a large number of questions that test the students’ ability to remember, apply and transfer what they have learnt. More often than not, however, they do not categorise these questions, which can lead to undesirable outcomes,” said Professor Andy Khong from the EEE. 
 
For their research, for example, the scientists had asked a subject matter expert to sort questions posed to students of an undergraduate electrical and electronic engineering course into those that helped the students to “remember”, “apply” and “transfer” their newfound knowledge respectively. 
 
The teacher discovered that most of the “transfer” questions, intended to help the students to learn how to use their knowledge to solve novel problems, were presented only during the final exam, when the students had no opportunity to learn from their mistakes to do better. 
 
To create their algorithm, the EEE researchers took 120 questions from the course, divided equally among “recall”, “apply” and “transfer” ones, and removed equations, mathematical symbols, diagrams, punctuation marks, numbers and non-unicode characters from them. They then stemmed the remaining words to obtain a list of root words for each question. 
 
They used the processed questions to train four combinations of word weighting and question labelling algorithms and validate the training. They then used another set of 30 questions to test the trained combinations’ reliability in labelling questions correctly. 
 
The researchers found that combining a modified version of term frequency-inverse document frequency – an algorithm used to identify the relevance of each unique word within a document – and extreme learning machine produced the most reliable results.   
 
Prof Khong said: “Our work can help instructors to not only label existing questions but also find new questions from external sources that align to intended learning outcomes. A novice instructor who takes over a course or teaching assistants tasked with refreshing course assignments would be able to use our work to create an optimal list of questions.”
 
 
 
Published on 1 Feb 2018​                    
Not sure which programme to go for? Use our programme finder
Loading header/footer ...