Mastering spaCy/

...

/

Using Dependency Trees For Extracting Entities

Using Dependency Trees For Extracting Entities

Let's see how dependency trees can help in extracting entities.

We'll cover the following...

Previously, we extracted entities where the context provides obvious clues. Extracting the destination city from the following sentence is easy. We can look for the to + GPE pattern:

I want to fly to Munich tomorrow

But suppose the user provides one of the following sentences instead:

I'm going to a conference in Munich. I need an air ticket.
My sister's wedding will be held in Munich. I'd like to book a
flight.

Here, the preposition to refers to conference, not Munich, in the first sentence. In this sentence, we need a pattern such as to + .... + GPE. Then, we have to be careful what words can come in between "to" and the city name, as well as what words should not come. For instance, this sentence carries a completely different meaning and shouldn't match:

I want to book a flight to my conference without stopping at
Berlin.

In the second sentence, there's no to at all. Here, as we see from these examples, we need to examine the syntactic relations between words. Previously, we already saw how to interpret dependency trees to understand the relations ...