Extracting positive descriptions and exploring landscape value using text analysis in the Cairngorms National Park

Being in natural places is known to influence well-being. Therefore, increasing attention is being paid to the diverse ways in which our emotions can be influenced by places, and this paper investigates this notion in a protected area - the Cairngorms National Park in Scotland. We assume that positive sentiment correlates with pleasant (hedonic) experiences, and extract positive emotions and objects associated with them, from textual descriptions written about places in the park. To do so, we first apply sentiment analysis and select descriptions with positive sentiment. Second, we filter the initial dataset to reduce the impact of prolific users and ambiguously positive words. From an initial set of 33790 descriptions we are left with 3031 positive descriptions. Third, we apply transducers to find semantic sequences for patterns detection. Finally, these sequences are annotated and assigned to one of five classes - activities, aesthetics, features, places and times. Features are by far the most common class, making up more than 50% of all unique nouns associated with positive emotions.