This DecarboNet opinion mining service aims to annotate documents with sentiment related to climate change. The web service takes as input a document or set of documents, and outputs those documents as JSON documents of standoff annotations with opinion, term and URI information (see the API documentation for more details). Sentiment is classified into positive, negative and neutral polarity, as well as more fine-grained emotions such as fear, anger, joy etc. The person holding the opinion and the target of the opinion are also recognised if present (e.g. a particular scientist is worried about fracking). Climate-related terms are also matched and linked to the instance of that term in relevant Linked Open Data ontologies.
Note that the current version of the tool is tuned towards precision. This means that it may not find all instances of sentiment. It is also designed only to work for English texts, although a German version is also available.
This is a simple demo of the main web service API, with the JSON output translated into HTML to aid navigation and display. Please type or paste some English text into the box below, in order to see the opinions and terms found.
- For each sentence that contains sentiment, an emotion is displayed.
- Positive emotions are highlighted in green, while negative emotions are highlighted in red.
- The person holding the opinion, if explicitly mentioned, is highlighted in purple.
- The target of the opinion, if explicitly mentioned, is highlighted in yellow.
- You can click on the link to see the origin of any climate-related term.