Entity is a term used in natural language processing. Simply, it means a general term, like location or a person.
When the text analysis model is developed, it can recognize different terms. For instance, if we take an article about new iPhone, a text analysis algorithm can understand and extract terms like iPhone (=appliance), Apple (=organization), Tim Cook (=person), dates and etc. Here is a list of typical entities that Bitskout can recognize:

Custom Entities
More specialized algorithms can recognized legal terms, medical items and etc.
Here is an example from Amazon Machine Learning of a custom entity:

As you can see the algorithm can extract a insurance policy ID from text which allows you to detect whether the policy ID is present in a message or text or not.
Today, it is quite straightforward to build custom entities algorithm and there are a lot of providers for that. From Bitskout point of view, we can integrate the custom entities via external provider or run it on our premises (subject to evaluation).