Fun With IBM Personality Insights

In this project I use curl and Python 3 to play around with IBM Personality Insights. I used IBM's free Lite service, which provides 1,000 API calls per month at no cost, and deploy to the US South region.

I enabled a Lite Personality Insights service instance, which only lasts 30 days.

I followed the Getting started instructions and referred to the API reference.

The service name was automatically created: Personality Insights-ca

The Getting Started examples use curl to call methods of the HTTP interface. Ubuntu's version of curl installed by default is the correct version.

I used jq to pretty-print the returned JSON. Here is how to install it in Ubuntu:

$ sudo apt install jq

Step 1 of the Getting Started instructions has a serious error: instead of providing the apiKey, your username:password must be provided. Because I did not want to hard-code credentials into a program, I defined environment variables to hold this sensitive data.

export PI_USERNAME="999999-8888-7777-6666-12345678" # replace with your Personality Insights username
export PI_PASSWORD="zYxWv"                          # replace with your Personality Insights password

Curl

The IBM Personality Insights Getting Started instructions show a curl example, which I modified as shown. I then wrote a simple Python equivalent, shown in the next section.

export PATH_OF_TEXT_TO_ANALYSE=./profile.txt
curl -sSX POST --user "$PI_USERNAME:$PI_PASSWORD" \
  --header "Content-Type: text/plain;charset=utf-8" \
  --header "Accept: application/json" \
  --data-binary @"$PATH_OF_TEXT_TO_ANALYSE" \
  "https://gateway.watsonplatform.net/personality-insights/api/v3/profile?version=2017-10-13" | \
  jq .

Notice the leading @ character before $PATH_OF_TEXT_TO_ANALYSE. If you don't put that character there, the path is interpreted as the text analyze, instead of the filename of containing the text to analyze.

Here is the JSON output.

Python

I found the Python docstring for the PersonalityInsightsV3 constructor to be helpful. The Python version is more ambitious than the curl version; the Python version reads two blog postings, combines them, and submits them for analysis. Here is the resulting JSON.