PYLON 1.7.1 includes a package of enhancements and maintenance items. It was released on April 19, 2016.
Premium Feature: Story-Level Sampling
interaction.sample treats a story and all of its engagements together as a candidate to be potentially selected for random sampling. The entire group is either included or excluded from a sampled recording. This is good for use cases focused on content virality, where it’s important to include all engagements for an included story to understand which audiences the story resonates with.
For example, we could record 10% of all stories and all of the likes, shares and comments on those stories:
interaction.sample < 10
Or we could record a different sized sample for various subsets of the data to ensure that we have a representative sample from each group:
(fb.author.country ==“United States” and interaction.sample < 10) or
(fb.author.country ==“France” and interaction.sample < 50) or (…
Premium Feature: Engagement-Level Sampling
Other use cases are better suited to a sampling target which allows you to sample stories and engagements at different rates. For example in a reputation management use case, you might like to capture all stories but only 10% of engagements to manage your available indexing capacity. For this use case, we are introducing fb.sample.
It’s important to understand how to best use fb.sample on different interaction subtypes. When using fb.sample with an fb.all.content condition, the sampling rate applied to engagements is your specified sampling rate squared. For example:
fb.all.content contains “dogs” and fb.sample < 10
This rule will take 10% of stories mentioning dogs, and 10% of engagements on those 10% of stories (which is 1% of all relevant engagements), because fb.all.content filters both stories and engagements. A better practice is to specify which interaction.subtype you want the sampling rate applied to. So for example if you want to filter 100% of stories and 10% of engagements:
fb.content contains “dogs” or ( fb.parent.content contains “dogs” and fb.sample < 10)
Both sampling targets, interaction.sample and fb.sample, will be packaged together as a premium feature set. Customers with access to sampling can always use either target. Contact your account manager or sales representative to find out more.
Topic Edge Targets Network graphs are powerful tools for exploring the clusters of activity in a recording, but until now they have been very query intensive to build from PYLON analysis results.
This gives you all of the “edges”, or relationships, between topics which exist on the same interactions. So you can build a network graph from the response of a single analysis request when using fb.topic_graph and fb.parent.topic_graph as the analysis target. fb.topic_graph is populated on stories, and fb.parent.topic_graph is populated on engagements.
Link Title Analysis PYLON allows you to use links.title in both primary interaction filtering and secondary query filtering CSDL to return interactions where a keyword in the title or headline on the web page of a shared link. Prior to 1.7.1, this target was not available for frequency distribution analysis. In 1.7.1, you can now analyze this target to return the most prevalent page titles or headlines in your recording. These titles are truncated at 256 characters.
Blocking unpopulated Super Public posts In 1.7.1 we will introduce a rule that disallows Super Public posts from being available to customers if they contain no text, link, hashtag, VEDO tag, or topic. In other words, if only a media type such as “photo” is populated, this is not informative for filter iteration. So to help customers get the most value out of their Super Public query allowance, we will not cache these for retrieval. Some customers currently block these barebones posts themselves with query filters like “fb.content exists”.