Announcing PYLON 1.7.1 - Introducing Enhanced Sampling Support

Richard Caudle | 18th April 2016

Today we released version 1.7.1 of PYLON for Facebook Topic Data. This release includes some key features that will make it easier for you to build richer analysis results with PYLON.


If you try to record a large audience with PYLON (such as people discussing a box office movie launch) you can quickly hit recording limits, especially if you're sharing your allowances across your end customers. Sampling allows you to record a proportion of interactions that is representative for your audience while staying within your recording limits.

PYLON offers two levels of sampling to allow you to build accurate representative audiences.

Story-level sampling, supported by the interaction.sample target, allows you to record a percentage of stories that are posted and all the related engagements.

For example, this filter records 0.5 percent of stories that mention automotive topics and all of the related engagements on those stories:

(fb.topics.category in "Cars, Automotive" 
OR fb.parent.topics.category in "Cars, Automotive") 
AND interaction.sample <= 0.5

Engagement-level sampling, supported by the fb.sample target, allows you to sample stories and engagements at distinct rates.

For example, this filter records all the stories that mention the brand and only 10 percent of the related engagements:

fb.content contains_any "BMW" 
OR ( fb.parent.content contains_any "BMW" AND fb.sample <= 10 )

Read our blog post which introduces the new sampling features for more details.

Sampling is an optional add-on you can purchase for your package. Contact your account manager or sales representative to find out more.

Topic network graphs

As Facebook's topic graph is constantly growing it can be difficult to understand which topics your audience are engaging with.

One approach to this challenge is to generate a network graph of topics.


In this graph the nodes represent the topics being engaged with and the edges show which topics are related based upon them being mentioned together in stories. The weight of the edges show how often the pair of topics that are connected are engaged with together.

Being able to visualize topics in this way is a great help as it surfaces clusters of topics. You can use these clusters to focus on trends for your audience, and to remove noise from your recording.

Prior to this release you could generate a topic network graph by firstly analyzing the top topics in your index, and then using each of the topics as a query filter for subsequent analysis calls to find the co-occurring topics in stories.

In this release we have introduced two new targets to allow you to perform this analysis in one API call. The fb.topic_graph target returns the most frequently co-occurring topics in stories being posted. The fb.parent.topic_graph target returns the most frequently co-occurring topics in stories being engaged with. It is a straight-forward task to generate a network graph visualization from the result of analyzing these targets.

Take a look at our topic network graph blog post to find out more.

Link title analysis

Before today's release PYLON already allowed you to specify filter conditions for your recording and analysis calls based upon the title of links being shared.

Until now you have not been able to see the text of these link titles in your analysis. This release allows you to use the links.title target in your analysis.

For instance you can retrieve the titles of the top links being shared:


Link titles are usually much more revealing than URLs and are more meaningful for people consuming your analysis.

Previous post: New Tokenized Targets for PYLON Query Filters

Next post: Studying large audiences with PYLON for Facebook topic data