Item: UNSTRUCTURED DATA PROCESSING: ASKING NEW QUESTIONS OF AVALANCHE ACCIDENT REPORTS AND OTHER NATURAL LANGUAGE DOCUMENTS
-
-
Title: UNSTRUCTURED DATA PROCESSING: ASKING NEW QUESTIONS OF AVALANCHE ACCIDENT REPORTS AND OTHER NATURAL LANGUAGE DOCUMENTS
Proceedings: Proceedings, International Snow Science Workshop, Whistler, BC, Canada, 2026
Authors:
- Bhavya Chopra [ Electrical Engineering and Computer Sciences Department, UC Berkeley, CA, USA ]
- Björn Hartmann [ Electrical Engineering and Computer Sciences Department, UC Berkeley, CA, USA ]
- Aditya Parameswaran [ Electrical Engineering and Computer Sciences Department, UC Berkeley, CA, USA ]
- Shreya Shankar [ Electrical Engineering and Computer Sciences Department, UC Berkeley, CA, USA ]
Date: 2026-09-28
Abstract: Statistical analyses are powerful tools for distilling patterns in avalanche accidents. Most such analyses rely on structured data — quantities that can be easily counted or measured — such as forecast danger ratings, burial times, or group sizes. But much of what the avalanche community records lives in unstructured text: accident narratives, crowd-sourced observation reports, and forecast discussions. Finding patterns in such sources has traditionally required manual search and review, which is time-intensive and limits the number of questions one can practically pursue. Recent advances in large language models (LLMs) have given rise to unstructured data processing: the ability to query and transform large collections of text documents using natural language prompts to define
Object ID: ISSW2026_P1.45.pdf
DOI: https://doi.org/10.15788/1790098815
Language of Article: English
Presenter(s): Bjoern Hartmann
Keywords: large language models, unstructured data processing, visualization, avalanche reports
Page Number(s): 1731 - 1738
-