Integration of an Expert Teaching Assistant with Distance Learning Software

AAAI 1996

TL;DR

This paper introduces an Expert Teaching Assistant (ETA) integrated with a Remote Teaching Assistant (RTA) system to provide students with on-demand help. ETA uses a knowledge base derived from web pages to answer student queries, aiming to bridge the gap left by unavailable human TAs and limited course knowledge.

What Problem Does This Paper Address?

The paper addresses the challenge of providing consistent and accessible academic support in distance learning environments. Specifically, it tackles the problem of TA unavailability and the limitations of students' access to domain-specific knowledge. When human TAs are offline, students often struggle to get timely help, which can hinder their learning. The authors propose an Expert Teaching Assistant (ETA) as a solution, designed to act as a knowledgeable intermediary for students interacting through a multimedia communication platform (RTA). The goal is to create a system that can maintain knowledge bases and assist students effectively, even without direct human intervention.

Methods

The core of the ETA system involves constructing a knowledge base from existing web pages. This is achieved through a translator that processes HTML files. The translator extracts information from specific segments within HTML headers, identifying the document's context and defining data objects. This extracted information is then converted into facts and rules compatible with the CLIPS inference engine.

The system architecture includes:

  • HTML Database: Stores the web pages.
  • Translator: Parses HTML, extracts information, and generates CLIPS-compatible facts and rules.
  • Knowledge Base: Contains domain-specific rules and facts, along with general knowledge applicable to all sub-domains.
  • Domain Resource Files: Store sub-domain dependent rules and temporary relationships.
  • Inference Engine: Executes a two-stage process: an initial solution search followed by an iterative refinement stage to pinpoint the best solution.

ETA operates by receiving a sub-domain and keywords from the student. It then uses its knowledge base to find relevant information. The system is designed to help students who may have difficulty articulating their problems by interactively asking clarifying questions. The student client utilizes a graphical user interface with dialog boxes and radio buttons to gather responses.

Datasets

Not discussed in the paper. The knowledge base is described as being composed of manually written rules and facts, linking to a small database containing question and answer files, terms, and lecture notes.

Key Insights and Ideas

  • Leveraging existing web content: The system demonstrates a practical approach to building a knowledge base by repurposing readily available web pages, reducing the manual effort typically required for knowledge acquisition.
  • Interactive refinement for problem-solving: ETA's iterative questioning strategy is crucial for guiding students towards solutions, especially when their initial queries are vague or incomplete. This mimics a human TA's ability to probe for more information.
  • Bridging the gap in distance learning: The integration of ETA with RTA offers a tangible solution to the persistent problem of TA availability in online education, providing students with continuous support.
  • Usability for non-technical personnel: The design aims for a mechanism that allows non-technical individuals to maintain the knowledge bases, which is a significant advantage for scalability and ongoing system management.

Key Results

The paper reports on a prototype implementation of ETA successfully integrated with RTA during the Fall of 1995. This prototype was tested in an introductory computer class at UC Davis. The knowledge base was initially built from manually written rules and facts. The system successfully provided students with ranked lists of relevant URLs, each paired with a descriptive string. Double-clicking on a description would launch a browser to display the HTML document.

Limitations and Future Work

  • Limitations: The paper acknowledges that the initial knowledge base was composed of manually written rules and facts. While the translator aims to automate this, the effectiveness and completeness of the automated translation process are not detailed with quantitative results. The system's performance in terms of accuracy and efficiency is not explicitly quantified.
  • Future Work: Development is ongoing for the initial course and is planned to extend to other courses, including non-Computer Science disciplines. The authors suggest further development and refinement of the system.

Summary updated

BibTeX
@inproceedings{1996integration,
  title = {{Integration of an Expert Teaching Assistant with Distance Learning Software}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {1996},
}
Figure