Mike Benson

Mike Benson

Solving natural language problems in cyber, intelligence, and influence operations.

Custom GPTs in ChatGPT

Overview

In this portfolio, I present a collection of custom GPT models that I developed to support students in mastering various topics in Cybersecurity, Computational Linguistics (CompLing), Human Language Technology (HLT), and Natural Language Processing (NLP). These models were created using OpenAI’s GPT technology and are designed to provide targeted assistance with learning key concepts and completing related assignments.

Project Goals

The primary goal of developing these custom GPTs was to create accessible, interactive tools that could help students grasp complex concepts in their coursework and enhance their problem-solving skills. By simulating tutoring sessions or providing specific task assistance, these GPTs bridge the gap between theoretical knowledge and practical application.

Technology Used

These custom GPTs were built using OpenAI’s platform, leveraging the capabilities of the GPT-4 architecture. The models were fine-tuned to focus on particular subject areas, incorporating domain-specific knowledge and instructional strategies aligned with the MSHLT program’s learning outcomes.

Custom GPT Models

  • Academic Site Builder: Assists graduate students in constructing academic websites, integrating both technical and design best practices.
  • Attempto Controlled English (ACE) GPT: Facilitates learning of ACE, a controlled natural language, useful for formal specifications.
  • CompLing Tutor: Provides targeted tutoring in Computational Linguistics topics, including parsing and grammar.
  • Intro to C Tutor: Supports students learning the C programming language with syntax guidance and problem-solving tips.
  • memm Classifier Helper: Aids in understanding and implementing Maximum Entropy Markov Models for sequence classification tasks.
  • MSHLT Advisor: Offers advice and guidance specific to the MSHLT program, including course selection and project planning.
  • MySQL Project Helper: Assists students in completing MySQL database projects, providing tips on query construction and optimization.
  • N-Gram Classifier Helper: Helps students with building and understanding N-Gram models for text classification tasks.
  • NCAE-C Instructional Designer: Provides guidance on designing instructional materials based on CAE Knowledge Units.
  • NLP Tutor: Provides tutoring in Natural Language Processing concepts, from tokenization to semantic analysis.
  • ParseTree to Tregex Converter: Converts parse trees into Tregex patterns, aiding in the syntactic analysis of texts.
  • Speech Technology Assistant: Supports students working on speech technology projects, with a focus on ASR and TTS systems.
  • StatNLP Tutor: Provides targeted help with statistical NLP topics, including probabilistic models and machine learning techniques.
  • TregexGPT: Aids in learning and applying Tregex, a tool for searching parse trees, particularly in linguistic research.
  • ZTN Tutor: Supports understanding and implementation of Zero Trust Network principles in cybersecurity.

Outcomes

The development and deployment of these custom GPT models resulted in several key outcomes:

  • Enhanced Learning: Students reported improved comprehension of complex topics, with the GPTs providing explanations tailored to their level of understanding.
  • Assignment Assistance: The models were instrumental in helping students complete assignments more efficiently and with greater accuracy, and in many cases, allowed them to exceed the assignment requirements.
  • Program Relevance: The GPTs reinforced the core learning outcomes of the MSHLT program by embedding fundamental algorithms, programming skills, and NLP concepts within the interactive sessions.

Future Work

Looking ahead, these custom GPTs will continue to evolve based on student feedback and advancements in GPT technology. Future iterations may include more sophisticated problem-solving capabilities, enhanced interactivity, and broader domain coverage.

Code and Resources

These custom GPTs are built using natural language instructions and hosted on ChatGPT. Premium ChatGPT users can explore them by clicking the individual GPT links provided in the Custom GPT Models section.

Professional Reflection

Creating these GPTs was a significant learning experience that deepened my understanding of both the technical and pedagogical aspects of AI-driven education tools. The project provided practical insights into fine-tuning large language models and highlighted the importance of aligning educational tools with specific learning outcomes.

Above and Beyond

The most relevant feature of this portfolio item is its demonstration of how AI can be customized to support education in highly specialized fields like Cybersecurity and NLP. These GPTs serve as both a proof of concept and a practical resource for students, illustrating the potential of AI to enhance traditional learning methods.