Most models for this task determine if two mentions refer to
each other using a single function over a set of constraints
or features, but some recent approaches tend to use multitiers
methods where mentions are disambiguated gradually
in well-ordered tiers which apply each, a specific function
[55,57]. It is obvious that in a Big Data supply chain, such approaches
can be difficultly used without modification. Indeed,
analyzing billions of documents more than seven times is not
realistic. We note that (the direct) approach of [58] (direct) is
more scalable, but it is a pairwise disambiguation method.