Automatically Correcting Semantic Errors in Programming Assignments

Ben Trevett, Donald Reay, Nick Taylor

Research output: Contribution to conferencePaperpeer-review

224 Downloads (Pure)

Abstract

MOOCs allow thousands of students to be taught how to
program at scale and, using automatic error correction, personalized
feedback can also be provided at scale. Current NLP techniques have
been used to correct errors in programming assignments, however, these
mainly focus on syntactic errors and do not take advantage of NLU
techniques to understand the desired semantics. By understanding se-
mantics, there is a potential that the accuracy of error corrections can
be vastly improved as focus shifts to correcting errors specific to the de-
sired problem, rather than errors common across all programs. Each step
of the process also gives an opportunity to do surrounding work on NLU
techniques applied to source code.
Original languageEnglish
Number of pages4
Publication statusPublished - 18 Sept 2017
Event10th European Conference on Machine Learning & Principles and Practice of Knowledge Discovery in Databases 2017 - Skopje, Macedonia, The Former Yugoslav Republic of
Duration: 18 Sept 201722 Sept 2017
http://ecmlpkdd2017.ijs.si/

Conference

Conference10th European Conference on Machine Learning & Principles and Practice of Knowledge Discovery in Databases 2017
Abbreviated titleECML PKDD 2017
Country/TerritoryMacedonia, The Former Yugoslav Republic of
CitySkopje
Period18/09/1722/09/17
Internet address

Keywords

  • automatic error correction
  • natural language processing

Fingerprint

Dive into the research topics of 'Automatically Correcting Semantic Errors in Programming Assignments'. Together they form a unique fingerprint.

Cite this