OpenWest 2014/Unstructured Data

Add Structure to Unstructured Data: Text Analysis and Speech to Text

by Craig Golightly

"A good chunk of what is produced each day on the internet and within companies is Unstructured Data (free form text fields, social media, voice comments) and too many people are getting buried by the sheer amount of this data. It doesn't store or sort nicely in its raw form, yet that is what most people do--they just store it and save it for "later". Join me for some real life examples and applications of tools to make sense of your Unstructured Data and turn it into something useful NOW."

-

Text analytic and speech to text

Consistency is not a human trait

100,000 tweets, 5 sorters, 44% agreement

• if accuracy is less than 100% = Fail
• if accuracy is greater than 0% = Win

Text analysis allows us to monitor the known and discover the unknown

Not just used to sort and search but FILTER!

Can monitor 100% of call center recordings? How to pick?

1) Maturity? how long has it been around

2) Features - do they add value or is it just cool

Can they find imperative (action) items. If they claim 100% it is a sales pitch. Uses Wikipedia logic to find categories sentiment must have positive, negative, neutral intent.

Speech (phoetic search) monitors - known (speech to text)

3) Speed & Scale - desktop or larger. Small footprint and speed.

4) Open Data - what format is data in text, proprietary, accessibility, metadata, measure of confidence

5) tune-ability, - does it work with our process?

6) Cost - not just licensing but data center and configuration cost

notes by Bethany