GenAI Productivity: Ideas to project proposal 1

One of the ways I use Generative AI with students is to take basic ideas for projects, usually a title, and get these tools to greater ideas and start of a project proposal. This is with all the usual caveats 

• Check the references (if any);
• It is going to be basic, so extend it.

In this example I am going to use Co-pilot but the ChatGPT, etc can be used, employing a few basic prompt engineering basics: personas (who is the target audience?) and Templates (how do I want it to look?) to start this process.

Example: 

Project ideas for MSc Data Intelligence students (persona)  on a particular topic. The reply will include subheadings and relevant (hopefully) content for TITLE, INTRODUCTION, PROBLEM STATEMENT.

The prompt: "Taking the topic "Leveraging open-source tools to measure and present academics publications automatically from public domain data.". Give five innovative projects for a Master's level student
dissertation in Data Intelligence. Each project example will contain subheading and content for TITLE, INTRODUCTION, PROBLEM STATEMENT"

Copilot produces five ideas. Here is an example for the fifth one.

"Visualizing Research Trends with Topic Modeling

• Title: “Mapping Research Landscapes: Topic Modeling for Academic Trends”
• Introduction: Understanding research trends and thematic clusters is essential for informed decision-making. This project involves applying topic modelling techniques (e.g., Latent Dirichlet Allocation) to academic publications. By visualizing topic distributions, researchers can explore emerging areas and identify interdisciplinary connections.
• Problem Statement: How can we create an interactive visualization tool that allows users to explore research topics and their evolution over time?"

Okay, this is a nice start but let's make it a bit more like a project proposal and add in maximum word counts for each section extending the detail in the template. So the prompt used "Expand Project 5. For this project include the following Subheading in captialise and the for content of each subheading a max word count will be defined as a number and the subheading. Subheading and word count: TITLE 50; INTRODUCTION 100; STATEMENT OF THE PROBLEM 100; PURPOSE OF THE STUDY 100; ASSUMPTIONS AND HYPOTHESIS 100; DATA COLLECTION PROCEDURES 100; DATA ANALYSIS 100; DATA VALIDATION 100; ETHICAL ISSUES 100; REFERENCES 250"

Copilot went on to do the task but didn't produce a reference list. So a third prompt was produced to fix the issue "Repeat but add in citations through out and REFERENCES is a reference list of the citations"

So the final output from Copilot was:

"Title (50 words)

“Mapping Research Landscapes: Topic Modeling for Academic Trends”

Introduction (100 words)

In the rapidly evolving landscape of academic research, identifying emerging trends and understanding thematic clusters is crucial. Researchers, institutions, and policymakers need tools to navigate this vast sea of information effectively. This project aims to create an intelligent system that leverages topic modeling techniques to visualize and analyze research topics across various disciplines.

Statement of the Problem (100 words)

Despite the abundance of scholarly literature, researchers struggle to keep up with the latest developments. Traditional keyword-based search methods fall short in capturing nuanced relationships between research areas. How can we develop an automated approach that identifies and visualizes research trends, enabling informed decision-making?

Purpose of the Study (100 words)

The purpose of this study is to design and implement a topic modeling framework that extracts latent themes from academic publications. By doing so, we aim to provide researchers, students, and policymakers with an intuitive tool for exploring research landscapes, identifying gaps, and fostering interdisciplinary collaboration.

Assumptions and Hypotheses (100 words)

We assume that academic articles contain latent semantic structures that can be captured through topic modeling algorithms. Our hypotheses include:

1. Academic articles cluster around specific themes.
2. Topic modeling can reveal hidden connections between seemingly disparate research areas.
3. The resulting visualizations will enhance researchers’ understanding of the scholarly landscape.

Data Collection Procedures (100 words)

To build our dataset, we will collect scholarly articles from reputable open-access repositories (e.g., arXiv, PubMed). We will focus on articles published within the last five years to capture recent trends. Metadata (e.g., title, abstract, keywords) will be extracted for analysis.

Data Analysis (100 words)

We will employ Latent Dirichlet Allocation (LDA) and Non-Negative Matrix Factorization (NMF) algorithms to identify latent topics within the dataset. These methods will help us discover thematic clusters, assign articles to relevant topics, and create visual representations of the research landscape.

Data Validation (100 words)

To validate our results, we will compare topic assignments with manually curated topic labels. Additionally, we will assess the stability and coherence of identified topics using metrics such as perplexity and topic coherence.

Ethical Issues (100 words)

We recognize potential ethical concerns related to privacy, bias, and unintended consequences. We will ensure that data anonymization protocols are followed, address any biases introduced by the algorithms, and transparently communicate limitations to users.

References (250 words)

1. Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent Dirichlet Allocation. Journal of Machine Learning Research, 3, 993–1022.
2. Sievert, C., & Shirley, K. (2014). LDAvis: A method for visualizing and interpreting topics. Proceedings of the Workshop on Interactive Language Learning, Visualization, and Interfaces, 63–70.
3. Maaten, L. v. d., & Hinton, G. (2008). Visualizing Data using t-SNE. Journal of Machine Learning Research, 9, 2579–2605."

Summary:

It is a good start, and the student is expected to refine it; a good proposal requires a lot more detail. This is a quick technique that aims to mitigate the panic of a blank page, but still, enough needs to be done to make the work their own.

To read more go to part 2: https://llmapplied.blogspot.com/2024/06/genai-productivity-ideas-to-project.html

All opinions in this blog are the Author's and should not in any way be seen as reflecting the views of any organisation the Author has any association with. Twitter @scottturneruon