fam
fam

Reputation: 711

How can I find relationship between two entities or words using NLP?

I am fairly new to NLP. I am finding difficulty in finding the relationship between two words. For example, Sarah spends $10. Max Spends $100. All of them spent some amount. In total, they spent $500.

Now, there are multiple persons spending some money, i.e., Sarah, Max, All of them, and They. And for some of them, there is an associated amount of money that they spend. I can train the NLP model to do NER and find out which represents a person and which word represents money, but how can I build a solution which tells me that $10 is associated with Sarah, $100 is associated with Max, $500 is associated with They.

One more thing, what is this type of problem called in NLP? I appreciate your help. It is also great if you can provide me just with the topics that I can research to explore this type of problem. Thank you!

Upvotes: 0

Views: 902

Answers (1)

ghchoi
ghchoi

Reputation: 5166

Machine Reading Comprehension (MRC)

MRC is very popular subtask of Natural Language Understanding (NLU). Check the popular benchmark dataset for this task; SQuAD. This task is to find an answer span for the given question. For example, if the question is "How much does Sarah spend?", a model should outputs

Sarah spends <answer>$10</answer>. Max Spends ...

Knowledge Graph Building (KGB)

You may first extract entities as node in KG.

Then you may extract the relation spend with a corresponding value.

Sarah  spend  $10
Max    spend  $100
they   spend  $500

In this case, the problem could be defined as Information Extraction or Relation Extraction.

Upvotes: 1

Related Questions