Intro to supervised relation extraction

Relation Extraction(RE) is the method to finding and classifying semantic relations among entities mentioned in a text.

For example, suppose we have text:

Shaka khan born in Dhaka.

Here entities are Shaka Khan-> PERSON, Dhaka-> LOCATION.

and relation is: BIRTHPLACE.

So, how can we do this work in supervised classification?

So, let’s start………

Let’s say, our datasets is:

Sentence: Shaka khan born in Dhaka.
Tokens: ['Shaka', 'khan', 'born', 'in', 'Dhaka']
Entities: [B-PER, I-PER, O, O, B-LOC]

So, before feeding this data to classifier, we need to convert it to:

[PERSON] born in [LOCATION]

Then we feed this modified data to classifier and classifier will predict relation between entities, like BIRTHPLACE or NO-RELATION.

This image describe the whole process in briefly. For encoder, you can use any encoder like BERT, LSTM, GRU etc.