Inspiration
As the teacher to student ratio continues to decrease, teachers begin to work longer hours for less pay. We sought to make the job of teacher's easier. By assisting in the creation of tests by extracting text from sources, we seek to create a more streamlined process and allow the teachers to focus on personalized teaching.
What it does
The user would need to narrow down the source and subject area of the generated questions using query terms. Utilizing key word extraction on a specified text (found with previously mentioned query terms), we are able to generate multiple types of questions: true or false, fill in the blank, and short answer questions.
How we built it
We began by designing and creating a front end for our web application. We used CSS and HTML to create dynamic webpages responsive to varying screen sizes. Once our front end was created, we used NodeJS and Express to create a basic webserver to handle get and post requests. Combined with input boxes, we then were able to get data from the front end and conduct web scraping with the Google and Cheerio modules for web scraping. We then used the POS module to handle natural language processing in order to generate keywords and summaries for generating questions for our tests.
Challenges we ran into
We had never worked with natural language processing before, so we had a large learning curve to overcome.
Accomplishments that we're proud of
Effectively completing natural language processing.
What we learned
Instead of going the traditional route of utilizing languages like Java or Python, we tooled node.js for natural language processing.
What's next for QuickTest
The user's input in quality control sets the grounwork for further machine learning. We plan to use the data from the user either approving or disapproving of the generated question to more accurately predict questions in the future. Currently, the source of mining is from online pages. In the future, we plan to implement a simple, yet incredibly useful feature. We would make the source be textbooks which could be searched for by their ISBN number or by scanning the relevant pages. We'd also like to try different types of questions e.g. multiple choice questions.
Log in or sign up for Devpost to join the conversation.