From online texts to Landscape Character Assessment: Collecting and analysing first-person landscape perception computationally

Inspired by the narrative nature of Landscape Character Assessment (LCA), we present a complete workflow to (i) build a collection of almost 7000 online texts capturing first-person perception of the Lake District National Park in England, and (ii) analyse these for sight, sound and smell perception. We extract and classify more than 28,000 descriptions referring to sight, almost 1500 to sound and 78 to smell experiences using text analysis. The resulting descriptions can be explored for the whole Lake District revealing for example, how traffic noise intrudes on experiences in the mountains close to transportation axes. Linking descriptions to LCA areas allow us to compare properties of different regions in terms of scenicness or tranquillity at a macro-level by identifying, for example, LCA areas dominated by descriptions of tranquillity or anthropogenic sounds. At a micro-level, we can zoom in to individual descriptions and landscape elements to understand how particular places are experienced in context. Local experts gave positive feedback about the utility of such information as a monitoring tool complementary to existing approaches. Our method has potential for use both in allowing comparison over time and identifying emerging themes discussed in online texts. It provides a scalable way of collecting multiple perspectives from written text, however, more work is required to understand by whom, and why, these contributions are authored.