Understanding Sentiment Analysis: Malcolm De Leo #Netgain6 #MRIA


netgain mriaWelcome to my #Netgain6 MRIA live blogs. What happens at St. Andrews Conference Centre, gets blogged for all to read about. Each posting is published immediately after the speaker finishes. Any inaccuracies are my own. Any silly comments in [ ] are my own. Enjoy!

————————————————————————————————

Malcolm De Leo, Chief Evangelist, NetBase
Understanding Sentiment Analysis: Methodology and Relevance to Market Research

  • Do you trust social media research? Do you research opinions online before you buy a product? Is your answer the same for both?
  • 5 years ago there were 300 tweets per day and 200 million tweets per day now
  • 53% of people on twitter recommend companies or products in their tweets
  • The average consumer mentions specific brands over 90 times per week in conversations with friends, family, co-workers
  • Will research replace or augment? It doesn’t have to be one or the other.
  • Surveys remove and include biases all the time.
  • Challenge is not finding social media data because there is tons of it.
  • The world is no longer B to C, it’s now C to B. We can’t even keep up with the marketplace.
  • We are comfortable with numbers are removing bias but we still resist. We know social media is the wave of the future but we aren’t seizing the opportunity to control it.
  • Consider NLP versus text analytics vs keyword searching when doing sentiment analysis. Precision is key.
  • Covert passion – mentioning a hashtag. Overt passion – saying you love something about the hashtag
  • Market research companies wear no clothes – we can measure any of them anytime we want.
  • Your suggestion box is public – people will talk about you online no matter whether you want them to

4 responses

  1. What does De Leo mean by “consider NLP versus text analytics”? Natural Language Processing (NLP) is used in text analytics. As I see it, there’s no “versus” between the two.

    1. I believe he was talking about straight text searches vs grammatical configurations. For example, “stupid phone” vs “My phone is something that I think is stupid.” Text analytics will identify a stupid phone in the first case but perhaps not the second.

    2. Annie, as I see it, any decent text analytics tool — and I class NetBase as one — will successfully decode both of your examples. The difference between NLP and text analytics is that text analytics (done right) will help you make contextually correct business decisions based on the information decoded, vis NLP, from the subject text. It’s a difference between hearing and understanding. Text analytics is a data analysis, decision-support adapted process; NLP isn’t, on its own, really.

    3. I’d say that however folks want to define text analysis or NLP, the only thing that matters to me is validity. I’d like to see more work put into demonstrating the validity of whatever system is used, and doing so in a valid way.

%d bloggers like this: