People talk about rivers online: from complaints about pollution to celebrations of wildlife, rivers provoke passionate social media comments. When people express their feelings about rivers on Twitter, we get a glimpse of how nature affects human wellbeing. The River Sentiment Dashboard displays social media sentiment about more than 450 rivers and other waterbodies in the Thames basin in England alongside data about their ecological status.

This prototype was developed by the University of Oxford and Thames21. The project was supported in 2021-2022 by the University of Oxford's ESRC Impact Acceleration Account.

We welcome your feedback! Please contact us on Twitter @RiverSentiment or send an email to helge.peters at ouce.ox.ac.uk

You can search for a river or look it up on a map. For each river, the dashboard shows how social media sentiment about the waterbody has shifted between positive and negative over time. The dashboard also shows which basic emotions Twitter users express when talking about the waterbody and which phrases they commonly use in their tweets. Using these data you can explore the link between human wellbeing and river water quality by comparing social media sentiment with ecological status. You can also explore the connection between the common phrases people use to describe the river with the reasons why it is not achieving good ecological status. If you want to compare between different rivers, you can use the colour-coded map at the bottom, which displays overall positive (blue), neutral (yellow), and negative (red) sentiment about all of the waterbodies in the Thames basin.

We mine the Twitter platform for tweets containing the names of 450 waterbodies in the Thames river basin management catchment. The total number of tweets collected is over 4 million tweets with a temporal resolution of January 1st 2008 to the present day. Datasets are aggregated at the water body scale and contain sentiment polarity, emotion detection, and common phrases.

Our method for sentiment analysis utilised an augmented version of the NRC Word-Emotion Association Lexicon and a polarized context cluster algorithm in order to determine the sentiment score of a given waterbody. The NRC Word-Emotion Association Lexicon is a dictionary of English words and their associations with eight basic emotions (anger, fear, anticipation, trust, surprise, sadness, joy, and disgust) and two sentiments (negative and positive). The polarised context cluster algorithm gives better results than a simple lookup dictionary approach and functions by clustering groups of words in a tweet (normally around 4) and checks whether valence shifters effect the overall sentiment in the word cluster. Data about the ecological status of waterbodies and the reasons for not achieving good status are from the Environment Agency.

Please visit our GitHub to learn more about our methods and data.