CEO David Morgan explains: “Humans perform sentiment analysis all the time.  When you read a newspaper, you can tell whether an article is positive or negative, without thinking about it.  But a large company might be mentioned in 20,000 tweets a day.  For a person, that amount of data is overwhelming.  Our software can analyse it in a few seconds.”

The sophisticated software was developed by Professor Stephen Pulman and Dr Karo Moilanen from Oxford University’s Department of Computer Science.  TheySay applies over 68,000 rules to understand the structure of a text, and extract deep meaning.  As well as uncovering positive and negative sentiment, the software can detect fear, anger, surprise, humour, and speculation.

In the fortnight before the Scottish Independence Referendum (September 2014), TheySay analysed over 1.75 million tweets.   At first, people voting “Yes” (for independence) felt confident.   “No” voters (hoping to maintain the union) felt uncertain.  TheySay predicted that if Scots voted against independence, “Yes” voters would feel despair and anger, and “No” voters would experience euphoria.  After the result was announced, tweets showed exactly that pattern.

TheySay can also detect “intent to buy”, helping clients to understand their markets, project purchases, and evaluate advertising campaigns.

TheySay analyses any genre of text , from short sentences to hundred-page documents, in English or German.  Over the coming year, TheySay will keep a watchful eye on the news, and will analyse social media to provide insights into the US presidential elections.