In Focus: Retail
I'm working on some ideas that combine a number of IBM Watson services - in a manner that is well suited for Retail Experimentation. Rapid Prototyping.
Below, I've imagined a scenario where a customer walks up to a Robot - like this one http://www-03.ibm.com/press/us/en/pressrelease/46045.wss
Input, Process, Output
Processing: Speech to Text; Natural Language Classifier (three models here - clothing type, gender and color), and then Text to Speech - so robot can confirm intent before taking action to help. (e.g. 'follow me')
Output: Robot Speech; Idea: once gender of the item is known, the output could be changed (e.g. woman's dress - female voice; men's socks, male)
Orange Dress Shirt for Man
(WIll Clean up code and NLC ground truths - and post to Github when I get time)
 "RECORDING ------------------ (beep) "
 "Recording COMPLETE --------- (beep) "
 "Saving WAV File"
 "Calling IBM Watson Speech To Text API"
No encoding supplied: defaulting to UTF-8.
 " I'm looking for an orange dress shirt for a man "
1: orange 0.48099900956318126
2: white 0.17950373664368793
3: HEX-FFFFFF 0.06698889359108752
1: dress-shirt 0.9727118189156493
2: shirt 0.004986392718949907
3: sportswear 0.0038523948807940196
1: mens 0.7285096577955334
2: undefined 0.2269006663145822
3: womens 0.02452439806600695
 "you are looking for a orange mens dress-shirt Is that right?"
Sky Blue Cocktail Dress
 " I would like a woman's sky blue cocktail dress "
1: blue 0.7941227362761096
2: white 0.04684205403631115
3: HEX-00BFFF 0.018449209673064116
1: dress 0.9553517820198002
2: sweaters 0.0066020380011610836
3: formalwear 0.005701485025781948
1: womens 0.9870002010334813
2: girls 0.0068963982305446335
3: boys 0.0026603848109799226
 "you are looking for a blue womens dress Is that right?"
You can get REALLY Accurate on colors - note the third color class is the EXACT Hex Color - HEX-00BFFF - http://www.colorhexa.com/00bfff
More to come on color combinations; exact color matching
OCT 2016 SUPPLEMENT
IBM Personality Insights team and Acxiom have collaborated in a project to investigate whether personality insights of individuals, as derived from IBM’s Watson Personality Insights technology, can improve the accuracy of the models that predict the consumption preferences of individuals as compared to the models that use demographic attributes alone. We hypothesized that peoples’ personality traits when combined with their demographic attributes will improve the accuracy of the models in predicting their consumption preferences. Our study confirmed our hypothesis. In this study, we examined 133 consumption preferences of about 785,000 individuals in the US.
Out of these 133 consumption preferences, we have noted that adding personality insights attributes to demographics has improved the prediction accuracy for 115 preferences (86.5%). For 23 of them, only using personality insights attributes provides even better prediction accuracy than using demographics only.
About this blog
This is an informal blog that explores tools, code and tricks that group members have developed to engage IBM Watson cognitive computing services - from the R Programming Language. Packages include RCURL to access Watson APIs - for services that include Natural Language Classifier and Speech to Text. THIS IS MY PERSONAL BLOG - it does not represent the views of my employer. Code is presented as 'use at your own risk' (it has lots of bugs)
Created: September 13, 2015English