Yesterday I analysed the sentiment of a potential travel booking for a trip.
I looked at the data a bit further. For each of the reviewers there was a bit more information available.
The had the review distribution, the home town, age bracket and gender. This could help me further decide if I want to use their review in my evaluation.
Annoyingly for thiis experiment a lot didn’t have a location, age or gender. Sadly, when i looked at the data on 42 of my 108 (39%)reviewers had published their age and gender. Almost everyone has some sort of profile picture though.
So I decided to try to add to my data by taking everyone who didn’t have a gender and age, grabbing their profile picture and putting through the Face API to get it’s guess at their age and gender.
For example, this profile had no age and gender and the face and running https://westus.api.cognitive.microsoft.com/face/v1.0/detect?returnFaceId=true&returnFaceLandmarks=false&returnFaceAttributes=age,gender with the correct keys/content it told me the subject was a 42.5 year old female.
Then i did a bit of extra culling by excluding any results that were 15 or younger – as eyeballing a few of the photos indicated some people used a picture of their children (or what looks to be their children).
This gave me extra data for 8 reviewers bringing to a 46% dataset that now have both gender and age.
Now i’ve enhanced my dataset to have additional sentiment and also added some extra information to the reviewers I just need to do something with that data.