How to extract chunks with multiple patterns from pos tagged sentences?

Question

Given an input sentence that is pos tagged using pos_tag function in nltk :

[('Veer', 'NNP'), ('Singh', 'NNP'), ('Rathore', 'NNP'), ('auctioned', 'VBD'), ('his', 'PRP$'), ('gigantic', 'JJ'), ('house', 'NN'), ('in', 'IN'), ('New', 'NNP'), ('York', 'NNP'), ('.', '.')]

I need to extract the phrases which follow a certain pattern. For example, 'NNP NNP' or 'JJ NN'. There can be 'n' no. of patterns that we might want to extract. For example, here we need 2 patterns namely 'NNP NNP' and 'JJ NN'.

The output that I want for the above inputted sentence is a list of the phrases like :

output :

['Veer Singh Rathore', 'gigantic house', 'New York']

I have tried something like this :

> grammar = (''' Chunk:{<JJ><NN>|<NNP>+<NNP>} ''')
> 
> def pos_and_chunking(question):
>     words = word_tokenize(question)
>     pos_words = pos_tag(words)
>     chunkParser = RegexpParser(grammar)
>     chunked_phrases = chunkParser.parse(pos_words)
>     chunked_phrases.draw()
>     for subtree in chunked_phrases.subtrees():
>         print(subtree)

But the output I am getting is in the form of a tree.

Output :

(S (Chunk Veer/NNP Singh/NNP Rathore/NNP) auctioned/VBD his/PRP$ (Chunk gigantic/JJ house/NN) in/IN (Chunk New/NNP York/NNP) ./.) (Chunk Veer/NNP Singh/NNP Rathore/NNP) (Chunk gigantic/JJ house/NN) (Chunk New/NNP York/NNP)

How can this be resolved?

I referred this link for Chunking : https://www.codespeedy.com/chunking-rules-in-nlp/

Answer 1

If you expect a simple list of tokens with tags for your grammar, then you can simply flatten the tree: ignore the structure, and just filter out the tokens with tags, and then use the resulting list.