11/18/2020 Nikhil Sriram
My 2020 Springboard capstone project will be a quiz making program that will take factual and unambiguous statements and will question a user about them. The application will find, and be helped with finding keywords. Then a backend will generate confusing and fake alternatives. The fake alternative statements will then be the basis for a host of multiple choice questions that can be used to reinforce learning about the initial statement. Once this feature is implemented, larger collections of statements and larger collections of questions can be developed, and users will have a personalized learning environment to work within.
This tool can be useful to both students and educators alike. Knowledge testing is potentially a powerful way to reinforce the parts of a subject that must be necessarily memorized to empower deeper understanding. With Nugget-Quizzer, students and educators can focus their time and energy into the data entry of the useful information. With computer resources dedicated to generating the incorrect answers, more time can be spent on learning and memorization.
This app will utilize human data entry for the basic statements that will be the basis for the quizzes. Data will be collected from two primary external APIs. The first is WordsAPI, the second is Conceptnet IO. Words API has a free pricing plan that allows 2,500 requests, Conceptnet IO has a free online service and it’s data set is shared under a creative commons attribution-sharealike license.
The database involved in this project will involve two superstructures, a users dataset, and a quizzing dataset. The users data scheme is fairly easily understood. Each user will have necessary login and access/control restrictions. Each user will also have viewing and contribution access rules to various quizzes. Owners that create new quizzes will be the owner of that quiz and can invite other users to edit or view the quiz.
The quiz dataset will be built around truth statements. A truth statement will be linked to a keywords dataset. Each keyword and truth statement will be linked to a collection of fake alternatives to that keyword. When generating a quiz question, the fake alternatives will be plugged into the original statement, and the user will need to decide which statement is the correct one.
The fake alternatives will need to be found by traversing the semantic maps generated by WordsAPI and ConceptNet IO. The algorithm will need to successfully pick hyponyms, or words that are cousins to the original keyword. Synonyms will not be a good choice, because they could easily generate a fake alternative that is also a true statement. The quiz will need to have a clearly wrong and right answer.
The stretch goals for this project involve integrating different types of quiz questions. True and false, multiple choice, and fill in the blank type questions can easily be generated. Matching questions, step ordering questions will need a slightly different approach, Since these don’t require the external APIs these problems are not considered right now. Another stretch goal is to include data on the effectiveness of questions that use alternative words. The data should collect if the question was hard and if the question was fair. Questions that are unfair should be slated for deletion, questions that are easy should be deprioritized.
With the plan outlined above, this app can serve as a powerful reinforced training tool for educators and students alike.