Skip to main content Skip to navigation

Providing Insights for Open-Response Surveys via End-to-End Context-Aware Clustering

Project Overview

The document presents a novel framework that leverages generative AI to analyze open-response surveys in education, aiming to automate the analysis process and enhance the efficiency of formative assessments. By employing pre-trained natural language models to convert textual responses into semantic vectors, the framework facilitates the extraction of valuable insights and the generation of visual aids like word clouds. This innovation significantly reduces the time teachers spend on analysis and improves feedback mechanisms, allowing for more informed instructional decisions. However, the implementation of this framework also faces challenges, particularly in addressing the complexity of open-ended responses and ensuring that the contextual nuances of the data are accurately captured. Overall, the integration of generative AI in educational assessments shows promise in streamlining processes and enriching the feedback cycle for educators.

Key Applications

End-to-End Context-Aware Clustering Framework

Context: Analyzing open-response survey data in educational settings, primarily for formative assessments in classrooms.

Implementation: The framework uses pre-trained natural language models to encode survey responses into semantic vectors, which are then clustered using algorithms like k-means to extract insights.

Outcomes: Improved efficiency in analyzing open-ended responses, enhanced insights for teachers on student understanding, and the ability to visualize data through word clouds.

Challenges: Complexity of analyzing diverse and context-rich open-ended responses, potential biases in clustering, and the need for extensive computational resources.

Implementation Barriers

Technical Barrier

The complexity of accurately processing and analyzing open-ended responses due to their varied and nuanced nature.

Proposed Solutions: Utilizing advanced natural language processing techniques and pre-trained models to capture contextual semantics effectively.

Resource Barrier

The computational resources required for implementing the generative AI framework effectively.

Proposed Solutions: Developing optimized algorithms that can run efficiently on mobile and less resource-intensive platforms.

Project Team

Soheil Esmaeilzadeh

Researcher

Brian Williams

Researcher

Davood Shamsi

Researcher

Onar Vikingstad

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Soheil Esmaeilzadeh, Brian Williams, Davood Shamsi, Onar Vikingstad

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies