ndexbio / gsoc_llm

GSOC 2024 LLM Project
MIT License
0 stars 0 forks source link

Advice on wrapping up the README #24

Open dexterpratt opened 3 months ago

dexterpratt commented 3 months ago

A short description of the goals of the project.

[Move these sections up to be after the project description]

Work Done

Describe the process, what you did to accomplish the goals of the project.

For example, your description might mention what was done related to each of your python scripts.

Save the description of the outputs for the current project state section. In this section, you might say, "To compare the output of the LLM with the REACH parser, I did XXX. The analysis scored the LLM by YYY" These scores helped guide my process of prompt development.

As you describe the process, briefly explain the things that a GSOC reviewer might not know. For example, what is INDRA, what is REACH?

Project Current State

Discuss what has been built and its state of operation.

Then discuss the results. Start with a qualitative description of the kinds of things that the LLM was able to do, give examples.

Then describe the results of the comparison to Reach output. Show examples.

Future Tasks

The obvious deliverables need to be mentioned: Uploading the knowledge graph to NDEx, making an application: Rather than building a standalone application, this code will become one of the first "app" extensions to the new Cytoscape Web application.

But also describe the improvements and further experiments that you would like to do.

Challenges and Lessons Learned

discuss the technologies that were new to you. which ones were the most interesting?

What new ideas, new methods, new knowledge were part of your summer experience? Pick one or two and explain what they are and why you were interested.

What are the approaches that you tried, didn't work, and how you solved the problem. Perhaps by fixing the problem, perhaps by taking a different approach.

How do you think this summer's work will influence your next steps in science and computing?


In general, think of how you would describe the project if you met someone at a conference. Or someone interviewing you. Or, as they suggest, like a blog post. Make it interesting. Make it accessible to someone who is a programmer or scientist but not familiar with knowledge graphs, parsers, prompt engineering, langchain, cancer pathways, etc.