At Appinions, we’re tackling a major issue for influence measurement – to distinguish the identities of individuals (influencers) who share the same name.
This is a critical aspect of influence analysis, as it establishes an identification method that clarifies an influencer’s identity and determines who the influencer is in real life (IRL).
Although the human eye can easily match the right name (in context) with the right person, this function is much harder for online platforms and technology to process. We call this issue disambiguation. To achieve the correct match, the platform must be able to detect when there are multiple identities that might be confused with each other (we’ll call this ambiguity), and when there are identities across multiple content platforms that might refer to the same individual IRL (we’ll call this personification).
The Ambiguity Challenge of Influence
Using a current pop culture example, when analyzing unstructured free text, influencers appear as named entities, or proper name mentions of individuals such as “Chris Brown.” This causes ambiguity because there is more than one real person with the same name.
Consider the following sentences extracted from different online sources:
- According to Hollywoodlife.com, Chris Brown contacted Rihanna following her emotional interview on “Oprah Winfrey’s Next Chapter,” apologizing for his actions and thanking her for respecting him on the program.
- Orbitz, which recently re-launched its iPhone app for speedier booking of air, hotels and cars, is seeing bookings via smartphones and tablets boom, said Chris Brown, Orbitz’s vice president of product strategy.
- With Buffalo’s stronger defense I [Chris Brown] would expect the Bills to defer almost every time they win the toss.
To the human eye, it is clear there are three different individuals named “Chris Brown” occurring in this content, because each item covers a different topic. The influence analyzer has to distinguish these identities using the context around the mentions. In computational linguistics, this problem is called cross document co-reference resolution.
Distinguishing between multiple influencers with the same names becomes even more challenging when you consider that these names can appear across multiple content sources as well, which brings us to personification.
The Personification Problem
When analyzing influence across multiple content platforms (e.g., free text news & blogs, blog comments, forums, Twitter, Google+), the same individual may be represented in multiple places. At Appinions, we refer to these representations as “personas”; the phenomenon of representing oneself in a social network is referred to as personification.
The entertainer Chris Brown has a Twitter persona and a Facebook persona, in addition to his named persona that appears in free text (as in example “a” above). These personas might generate content such as the following:
- [Twitter]: Congrats to my bro @YoungJeezy on his major boss move. SR Executive VP of Atlantic Recs! Way to boss up! #ItsTheWorld
- [Facebook]: This World Humanitarian Day I’m doing something good, somewhere, for someone else. Join me! #WHD2012 #IWASHERE http://thndr.it/PkC2Xe
If the analyzer determines that this content contributes to influence, then ideally we want the influence score to accumulate across all three personas that generated examples for Chris Brown (the entertainer).
This requires that we have a method for merging these personas into one profile. Merging can be based on persona attributes (e.g., the personas have the same name and the same location), or context (e.g., the content generated by/about the personas is very similar), or both.
Disambiguation Can’t Be Ignored
Disambiguation is a serious issue that influence measurement often fails to address accurately – after all, influence isn’t absolute. That’s why Appinions is a smarter influence marketing platform, helping users better identify the right influencer on the right platforms.
When we analyze unstructured free text at Appinions, we address the disambiguation of influencers with similar names in order to distinguish the influence of each person. To truly capture the entire influence of a given person, conducting an analysis across multiple content platforms (both unstructured and structured) requires merging these manifestations, or we risk painting an inaccurate picture of influence that lacks vital context.
Influence doesn’t live in a vacuum, and when you consider the millions of online and offline channels that an influencer can potentially impact, disambiguation is a crucial distinction that influence measurement platforms – and the brands that use them– can’t ignore.
Image credit: Rafael Anderson Gonzales Mendoza