Facebook Analytics can help marketers and social sellers better understand how their posts are engaging their customers. Manually reading and tagging all the posts just doesn’t scale, and eventually everyone wants to aggregate the statistics for posts with similarities. Use cognitive computing and let Watson do the work for you!

In the code pattern Discover hidden Facebook usage insights, we show you how, with a single Jupyter Notebook, you can use the Watson Python SDK to enrich your Facebook data with:

  • Visual Recognition
  • Tone Analyzer
  • Natural Language Understanding

Once you’ve enriched the data, you can explore it with interactive charting using PixieDust.

What can cognitive computing do?

Facebook will provide you with many metrics about each of your posts and provides links to the articles or photos. Let’s say you have hundreds or thousands of of posts. Try doing the following yourself:

  1. Read each post and record the following impressions:
    • Tone (emotion, language style, social tone)
    • Keywords found in the text
    • Things mentioned, what type of thing (e.g., location), its name (e.g., a city)
  2. Follow the links to articles and now read all the articles and do all of the above for them.
  3. Follow the links to photos and note what you see in each one:
    • Colors
    • Objects
    • Create a type hierarchy to classify those objects

Now let’s be honest, how many thousands of posts do you think you could process by yourself? Maybe you spend a lot of time in Facebook doing this in your spare time, but to record everything and make it useful to your business is too much work.

Fortunately, this is what cognitive computing can do for you when you run a Jupyter Notebook using the Watson Python SDK.

On GitHub, you’ll find a developer journey that demonstrates how to take a CSV file with Facebook metrics and enrich it with attributes discovered with Watson Visual Recognition, Tone Analyzer, and Natural Language Understanding.

Once you have all that information in a pandas DataFrame, you can process it with Python code and visualize it with PixieDust.

Where does PixieDust fit in?

After you’ve let Watson do the hard work, let yourself take the glory with interactive charts. With PixieDust, charts are interactive, so you can change chart type, key attributes, groupings, metrics, aggregation, sorting, and more.

As you explore different combinations, you’ll uncover hidden insights in the data, find the best way to prove your findings, and share them with others.

In conclusion

Data science and cognitive computing are a powerful combination. With Watson and Data Science Experience, it is easy to harness the power and find hidden insights in your data. Be sure to check out https://github.com/IBM/pixiedust-facebook-analysis today!

Join The Discussion

Your email address will not be published. Required fields are marked *