A Semantic Text Processing System for Free-Write English Papers

Ryan Depascale, Stefan A. Robila

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

We present the development and evaluation of a semantic text processing system that evaluates student essays. The system processes n-many documents and suggests a letter grade, identifies papers that may need additional teacher action based on component/composite scores, and allows optional teacher input on features to generate the grade. The system was developed in Python using open-source libraries and is also available as open-source. Using a human-in-the-loop approach, expert teachers were interviewed as part of the design process. Assessing the documents on token, sentence, readability, dependency distance, and part of speech with user guided feature selection the system generated automated results where the true letter grade and machine letter grade corresponded exactly in 46% of papers and in a ±1 letter grade interval in 86% of papers. The program can be further extended to flag grades for potential human review based on user defined criteria with example code provided for papers marked as written above the high school level.

Original languageEnglish
Title of host publication2022 IEEE Integrated STEM Education Conference, ISEC 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages103-108
Number of pages6
ISBN (Electronic)9781665484299
DOIs
StatePublished - 2022
Event12th IEEE Integrated STEM Education Conference, ISEC 2022 - Virtual, Online, United States
Duration: 26 Mar 2022 → …

Publication series

Name2022 IEEE Integrated STEM Education Conference, ISEC 2022

Conference

Conference12th IEEE Integrated STEM Education Conference, ISEC 2022
Country/TerritoryUnited States
CityVirtual, Online
Period26/03/22 → …

Keywords

  • Automated Grading
  • Natural Language Processing
  • Open-Source.

Fingerprint

Dive into the research topics of 'A Semantic Text Processing System for Free-Write English Papers'. Together they form a unique fingerprint.

Cite this