Social Media Analytics: Sentiment Analysis

Comments: 5

  • Arun Singh says:

    Thanks for this great post. It does seem to me very amazing that analysis (math) can pick up word combination to give a good representation a feeling. I guess errors can come with the inability of a program to identify the nuances of the language, ie language style mostly like irony or humor… Still, it is quite impressive!

  • Hi Angela, great post. Your summary is exactly right. We’ve been working in the Social Media Analytics space for over 2 years now and continue to be amazed at the difficulty of the problem.

    One part of your post is the special challenge; that in including the blog posts themselves in the analytics and merging that information with the social channels is very, very difficult. Our SaaS tool SocialEars is just entering beta now and does exactly that. We scan the major social channels (TW, FB, LI, RSS) and also all of the blogs and articles that those channels point at (usually using short-urls). All of that is merged in real-time to identify ongoing conversations and those that are most influencing those conversations.

    The real difficulty for us was in making this topic-focused which we have succeeded in doing. We needed to employ some difficult mathematical pattern analysis to the results to reduce the noise level. Good thing is we’re launched the the cloud, so computing power is easily available for this challenge.

    In fact your blog article was discovered by SocialEars because we are analyzing the Social Analytics space for our own marketing purposes. We’re finding almost 10,000 articles and blogs a day specifically talking about Social Marketing and Analytics.

    Please get in touch. I think we have a lot to talk about.

  • Sentiment analysis is a challenge. I am always very interested to hear what others are doing for tracking and analyzing blog content.

  • It was a subtle but profound moment i realised that a word was not a number, and switched my focus to part of speech tagging. A wren and martin and a simple calculator is all I now need to make sense of the chaos out there. NLP is definitely far more efficient, especially in reading through text.

  • We’ve thought about adding a simple sentiment analysis feature to our social business analytics app ( which uses the number of people who like / unlike a page to determine whether there’s a positive or negative trend. It’s not text analytics for sure, but it’s an easy way to get a basic thumb-in-the-air perspective of fan sentiment.

Add a New Comment

Thank you for adding to the conversation!

Our comments are moderated. Your comment may not appear immediately.