Spring Symposium on UR and Community Engagement has ended
Back To Schedule
Tuesday, April 24 • 11:40am - 12:40pm
Determining Evironmental Adaptive Capacity of Building Parcels Through Natural Language Processing

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Climate Change has fundamentally altered how local governments approach disaster preparedness. A shifting climate exposes a greater variety of vital infrastructure to possible damage, which could have a cascading effect in an area’s ability to shelter and aid its citizens. As new disaster models are being developed it is necessary to evaluate the level of trauma buildings could be exposed to and the structures’ capacity to sustain that damage; a measurement known as adaptive capacity. Determining adaptive capacity at scale has proven to be difficult due to the historical difference in how buildings were constructed and when those regulations have been changed. Through the use of machine learning algorithms and neural networks, we have created a data pipeline to analyze legal corpus to determine its relevance in calculating the adaptive capacity of a series of land parcels. By utilizing Latent Dirichlet Allocation (LDA) models we are able to break a legal corpus down into its constituent topics and use network analysis to determine the relative difference between known relevant legal data. This data is then used in a feed-through neural network to further break down legal sentences into a vector space, using semantic analysis to determine the discrete values which determine the adaptive capacity of a land parcel. In this presentation we will visualize, compare and discuss the machine learning algorithms implementation in this data pipeline, the success rate in determining the relevant information, and future plans for incorporating other forms of disaster preparedness into the data pipeline.

Tuesday April 24, 2018 11:40am - 12:40pm PDT
125 Rhoades Robinson Hall