Text and social media analytics #IIeX


The state of text analytics today by Seth Grimes 

  • concept of extracting infromation from text dates back 60 years, it’s not new
  • bulk of information value is perceived as coming from data in relational tables as the data is structured and easy to mine and analyze


  • those embracing text analytics report having an epiphany moment
  • the insights are different from survey insights and they are very much valid
  • pay attention to emojis – the way people communicate changes
  • omni channel suggests all channels are similarly important so multichannel is probably a better term
  • for insights, technology drives methods, data science, data monetization, algorithms both cognitive and affective, API cloud services – volume, velocity, variety, language engineering, deep learning
  • do you currently need or expect to need to extract or analyze things – 85% want topics and themes, 80% want sentiment opinions attributes emotions intent, 80% want relationships
  • Word2Vec – word representation in vector space
  • Emoji – grammar with semantics that is visual
  • Facebook topic data – look for new methods to ensure vendors are staying up with the emerging capabilities

Uncovering emotional behavior drivers in text by David Johnson

  • organizations are still structured like the 1940s
  • people are data do not live by our boundaries
  • nearly all data seems to require enrichment and has an element of doubt
  • is it a lack of confidence?
  • have you ever taken a structured survey and felt that you communicated everything you wanted to in the way you wanted to? [why yes, all the time #sarcasm]
  • how much better are imaginative questions? more emotions, more sentiment classifications, more cognitive states mentioned
  • volume is not always important, the latent signal with emotions is far more important
%d bloggers like this: