I did some experiments to look at the capabilities of AI2's DataVoyager AI tool for data analysis. It can locate and/or generate Python code and can resolve dependencies, but the environment does not support modern NLP libraries (transformers, spaCy), so only deterministic, non-ML methods are used for extraction and analysis. I had it analyze 1000 abstracts gathered by Harzing's Publish or Perish on robust AI.

https://asta.allen.ai/share/f0b030eb-8eb3-4a79-9421-33cca998250a

#research #AItools #DataVoyager #AI2 #science

I ran an experiment with AI2 Asta's DataVoyager tool, which is now available to the public. I uploaded a public Kaggle dataset of 10,000 Arxiv AI papers:
https://www.kaggle.com/datasets/yasirabdaali/arxivorg-ai-research-papers-dataset

My query, "Analyze the summary column and identify topic trends in this paper collection." and results are below:
https://asta.allen.ai/share/d7a53d94-cf6e-4c94-b28d-7c2cbf88df32

#research #AItools #DataVoyager #science