SUBMISSION PURPOSES, HEALTHCARE TRACK:
Summary and Inspiration
SKIMethods is a new way to read the literature. Inspired by the struggles of the graduate student trying to keep up with the literature while conducting their own research, SKIMethods uses machine learning to pull key points from papers and presents these highlights to users for their own use - all without having to pore through time-consuming text. For example, a cancer researcher will be able to use SKIMethods to extract protocols from a materials and methods section of an experimental paper that uses a cell line similar to theirs; a physician can pull highlights from a several case reports on experimental treatment of COVID patients; a graduate student no longer has to choose between spending time on their experiments and keeping themselves well-read.
Summary categorization by Google API "AutoML Natural Language"
SkiMethods uses a Google API, AutoML Natural Language (particularly Entity extraction), to create a new categorization/grouping of the scientific words found in the "method" section in papers that a user is interested in. A user will be able to evaluate which method to use for their research by going through these AI-made categories as decision factors.
Step 1: Upload texts of "method" section (in JSONL file format) to Cloud Storage (https://console.cloud.google.com/storage/browser/ivyhackssum;tab=objects?forceOnBucketsSortingFiltering=false&authuser=1&project=ivyhackssum&supportedpurview=project&prefix=&forceOnObjectsSortingFiltering=false)
Step 2: Select a CSV file, which is a list of GSC paths to JSONL files, on Cloud Storage (https://console.cloud.google.com/natural-language/locations/us-central1/datasets/TEN8197724500721139712/import?authuser=1&project=ivyhackssum&supportedpurview=project)
Reference: AutoML Natural Laungage Quickstart Guide (https://cloud.google.com/natural-language/automl/docs/quickstart)
What's next for SKIMethods
SKIMethods' most obvious application in research extends to almost any scientist, including physicians. In the current pandemic and beyond, staying aware of new methods to treat patients is essential, and we believe SKIMethods would be able to lessen the time burden that deeply reading the literature generally presents. Case reports are usually structured differently than basic science research papers, so SKIMethods would have to be trained on case-report-style layouts as well.
Log in or sign up for Devpost to join the conversation.