A problem with traditional sentiment analysis is that it lacks context; it typically only gives an overall sentiment value for the text. This isn’t particularly insightful as people can express different opinions (positive/negative) about different subjects within one body of text. For example: [Read More]
A Better Search Algorithm For Natural Language Processing
Improving the performance of natural language processing search algorithms.
A large part of my job involves natural language processing. What I repeatedly find myself doing is searching through large lists of regular expressions, and doing something based on the expressions that have been matched. [Read More]
Twitter Visuals For ITV's EU Referendum Coverage
My code has been run on live TV; has yours?
Blurrt were asked by Twitter to help them visualise the conversation around the EU referendum for ITV. [Read More]
Twitter Visuals For The Queen’s 90th Birthday
My code has been run on a huge screen in front of Buckingham Palace; has yours?
Blurrt were asked to join in with The Patron’s Lunch, a huge street party to celebrate The Queen’s patronage of over 600 charities and organisations on the occasion of her 90th birthday. [Read More]
Analysing Twitter Conversation About Jeremy Corbyn
Identifying and explaining areas of conversation around Jeremy Corbyn.
Identifying trends in users expressing opinion about Jeremy Corbin, and explaining these trends through real world data. Originally done as coursework for a Data Mining module at University. [Read More]