Quantifying and mapping landscape value using online texts: A deep learning approach

Abstract

Current studies have quantified landscape value based on the quantity of social media, ignoring public preference, emotional tendency and intensity. Our objective is to quantify and map landscape value by developing a novel intelligent landscape value evaluation method based on online texts. We argue that online texts can be converted into quantities through deep learning technology, which allows the measurement of public preference, emotional tendency and intensity. We selected perception frequency, perception heat, perception preference and the value index as indicators to quantify landscape value, which is the product of perception heat and perception preference. We collected more than 14,000 online texts about the Orange Islands in Changsha city from 2010 to 2020, analysed the perception heat using Python to extract data, and analysed the perception preference using Baidu API sentiment analysis technology. The results show that the novel method has a wide range of data sources and represents the real public perception and value orientation, demonstrating that online texts with a wealth of mined information can be used to evaluate landscape value. The model calculating the landscape value overcomes the shortcomings of current methods that ignore public preference, emotional tendency and intensity. The results also show that the conservation mode of Orange Island is successful, and the conservation method suggested by the recommendation on the historic urban landscape is feasible. The study highlights the contribution of public preferences to landscape value and provides insight into the situations and variations of value types on Orange Island. The novel method is highly intelligent, time-saving, has low labour costs, and is applicable and can be applied to the evaluation of landscape value, especially in cases where value conflicts exist.