The idea of this project has for some time been already in my head and recently I’ve found the technology that helped me in realizing it. As text processor I use TextBlob library for Python. The essence of the project is news analysis – what the world is talking about, what the most popular news are. But the more interesting is to study the attitude to the news, which can be expressed in the news article itself and definitely in social networks.
So I used the following resources for the project:
• BBC Front Page
• Reuters Top News
• Huffington Post
• NY Times Homepage
• Daily Mail
• WSJ World News
The result of the sentiment analysis of the most popular news from those resources you can see in the Bubble Chart, where the size of a bubble represents the frequency of a word or a phrase mentioning in the media (50 most frequently mentioned) and the tone of the bubble gives a flavor of the emotional coloring of the news, highlighting the given word or phrase.
By clicking on a particular bubble below you can find the links to the articles, where the chosen word was mentioned. Talking about social – here I realized the sentiment analysis of the most popular news topics (words and phrases in bubbles) in Twitter. To check it out you just choose option ‘Twitter’ in the source menu. And likewise in this case you can click on a particular bubble to see the Twits’, mentioning the word/phrase you are interested in.
I’m not really happy with results. If phrase extracting works really good but sentiment analysis has quite big error rate. In plans to use Neural Networks with TensorFlow as the sentiment analysis tool.