ASP3IRE_SM_Promo.mp4
Analyze social media records to gain insights into children's environmental health
Author: Andrew Larkin
Affiliation: Oregon State University, College of Health
Summary
This github repository contains python scripts and custom classes for ingesting social media records into a Neo4j database as well as deep learning models and jupyter notebooks for and researching attitudes, perceptions, behaviors, and children's environmental health social media misinformation.
Why Social Media for Children's Health?
Children's environmental health issues can arise quickly and/or within overlooked and underserved populations. Social media analytics of public datasets allows for hard to reach voices to be heard, and for quickly identifying arising children's environmental health issues.
Repository Structure
Files are divided into five folders, with each folder corresponding to a unique stage of the research pipeline
- database setup - ingest records from X (formerly Twitter) and GIS datasets into a Neo4j database. Additional operations include processing social media records (e.g. georeferencing records)
- deep learning - multimodal models for extracting information from X text and imagery
- analysis - analyzing records for trends related to children's environmental health.
External Links
- Funding - NIH/NIEHS, GRANT13248774
- OpenStreetMap - https://www.openstreetmap.org/
- X (formerly Twitter) - https://twitter.com/home
- Advancing Science, Practice, Programming and Policy in Research Translation for Children’s Environmental Health (ASP3IRE) Center - https://health.oregonstate.edu/asp3ire
- NVIDIA Accelerator Research Program - https://www.nvidia.com/en-us/industries/higher-education-research/applied-research-program/
Related Publications
- Identifying children’s environmental health risks, needs, misconceptions, and opportunities for research translation using social media
- Integrating Geospatial Data and Social Media in Bidirectional Long-Short Term Memory Models to Capture Human Nature Interactions
- Measuring and modelling perceptions of the built environment for epidemiological research using crowd-sourcing and image-based deep learning models
Related Repositories
- https://github.com/larkinandy/GreenTweet_MultivariateBiLSTM - predict nature perceptions and use from Twitter records, and link to OpenStreetMap
- https://github.com/larkinandy/Portland_UrbanNature_Twitter - pilot study analyzing self-reported urban nature trends in Portland, OR from Twitter posts.