Using Calais to filter searches for words with 2 meanings
Using Calais to filter searches for words with 2 meanings
Posted on: Fri, 06/27/2008 - 07:25
Hi, I have been playing with a few of the Calais tools in particular to see if it can distinguish words with multiple meanings. For example 'Zurich' could refer to the city of the fininacial services company, or 'Saga' the firm or ongoing calamity. Calais sometimes gets it right, sometimes gets it wrong, and other timesw misses the word altogether. Is there a view on trying to tighten up on this kind of situation or is this just always going to be a very difficult issue?

Comments
Michael -
Disambiguation will always be a challenge, but take a look at Release 3.1. It includes company and geographic disambiguation. See http://opencalais.com/blog.
Regards,
Michael -
The short answer is Yes and Yes.
"Is there a view on trying to tighten up on this kind of situation"?
Yes, we're constantly working on fine-tuning our system to better recognize elements, especially ambiguous ones.
For disambiguation, the system takes into account the context around each element, so for example Ford could be a Person name in one instance but a Company name in another - depending on the context.
Adding 'context' is a continuous effort while we explore new content types.
"or is this just always going to be a very difficult issue"?
This situation will indeed always remain a difficult problem, but over time you will see improvements as we add more and more 'knowledge' to the system.
Michal