Hacker News new | ask | show | jobs
by eurasiantiger 1361 days ago
The ”entity recognition” is absolutely useless.

I tried it with a short text describing several entities and their relationships, yet it only spotted the word ”Users”. Not ”company”, not ”account”, not ”subscription”… not exactly impressed.

1 comments

Named entity recognition is typically used to locate and classify named entities in text. So you'd want to have a text that mentions specific things - companies, people, locations, etc. Abstract things like "account" or "subscription" don't technically fall under "named entities" category.
So what good is the algorithm for knowledge management in IT if it cannot handle abstraction?
Can you show an example of the text you was trying to extract entities from, interested to see the problem.
I didn’t save it, but it was a generic specification text much like this one, but longer and more detailed and nuanced:

Users have personal user accounts. Each user account can belong into multiple companies. Each company can assign users access rights to system services according to predefined roles. Each system service has a unique identifier as well as one or more service group identifiers and namespace identifiers.