This document discusses using text to explore landscapes in three ways:
1) Macroanalysis of large text corpora to quantify how people perceive landscapes.
2) Close reading of text to deeply analyze landscape meanings.
3) User generated content like social media that provide scalable methods to understand valued landscapes.
However, computational text analysis at scale poses challenges regarding language, domain specificity, and identifying relevant landscape documents.
Original Description:
Original Title
WEBINAR GISCIENCE TEXT-BASED EXPLORATION OF LANDSCAPES
This document discusses using text to explore landscapes in three ways:
1) Macroanalysis of large text corpora to quantify how people perceive landscapes.
2) Close reading of text to deeply analyze landscape meanings.
3) User generated content like social media that provide scalable methods to understand valued landscapes.
However, computational text analysis at scale poses challenges regarding language, domain specificity, and identifying relevant landscape documents.
This document discusses using text to explore landscapes in three ways:
1) Macroanalysis of large text corpora to quantify how people perceive landscapes.
2) Close reading of text to deeply analyze landscape meanings.
3) User generated content like social media that provide scalable methods to understand valued landscapes.
However, computational text analysis at scale poses challenges regarding language, domain specificity, and identifying relevant landscape documents.
CHANCES, CHALLENGES, AND LIMITS FOR TEXT-BASED EXPLORATION OF LANDSCAPES
January 13, 2021
WHAT ARE WE TALKING ABOUT? GIS Science Perspective:
Landscape Starting Points
- Defined as an area, as perceived by - Much geographic information about
people, whose character is the result of landscape (as opposed to research) the action and interaction of natural focuses on well-structured spatial data. and/or human factors. - Space is characterized in terms of crisp - Definition is from the European objects or continuous fields. landscape convention. - Attributes have single value, representing an institutional viewpoint OTHER DEFINITIONS: and purpose. - Landscape is not a mere substrate, nor a bundle of (actual or potential) GLOBAL ECOLOGICAL LAND UNITS resources, but instead is invested with a framework of deep meaning. - A global dataset that maps ecosystem o Meaning matters for landscape using: - The unpacking and valuation of some o Bioclimate “cultural ecosystem services” not o Landforms readily amendable to biophysical or o Landcover monetary metrics have lagged behind, o Rock types and so has their mainstreaming into - Endpoint: description of individual policy. ecosystems THE AIM Example: - We aim to use TEXT to EXPLORE, - Assynt, Scotland = Cold wet hills on QUANTIFY, and COMPARE ways in carbonate sedimentary rock with which people perceive landscape and grassland, shrub, or scrub. its meaning(s) through 2 ways: o MACROANALYSIS o MICROREADING WHAT ABOUT VGI/UGC/SOCIAL MEDIA? AKA close reading - User-generated content have emerged as a scalable method to explore where and how landscapes are valued. CHALLENGE:
HOW CAN WE COMPUTATIONALLY ANALYZE
TEXTS AT SCALE TO EXPLORE LANDSCAPES?
A FEW WORDS ABOUT COLLECTIONS:
LANGUAGE MATTERS
- We work primarily with English and
German - It’s important to be clear
DOMAIN SPECIFIC CORPORA
- Means the user already know the sorts
of things the user expects people to write. - Include digitized Swiss Alpine Club yearbooks and scraped hiking descriptions. - Building general landscape corpora requires identification of geographically and thematically relevant documents and filtering of duplicates.