Analysing Guardian Coverage with ABSA

Being interested in learning about more machine learning techniques and also UK politics gave me an idea to analyse how closely public opinion was mirrored by UK news sites. The Guardian is one of the largest news sites in the country and happens to have a free API for pulling data on anything they’ve published. I was interested to see how well ABSA models could work with news articles, which are often very dense and require context in ways traditional ABSA uses like product reviews might not. Also on a personal level I was interested to investigate how true assumptions people interested in UK politics might make (e.g. “Farage is good at grabbing headlines” or “The Guardian as a left-leaning paper will generally give labour positive coverage”) actually were.

Getting the data was a simple process, involving …

Runnning ABSA …

This allowed me to generate results like …

Overall …

Skills:

• Natural Language Processing
• Machine Learning Models *