Python nltk extract useful chunk

There is a lot of text information. How do we extract useful information?

For example:

JSON is a good boy

The expected information is JSON and a good boy.

First, we need to split sentences and determine the attributes of words:

You can use the followingCode:

 
Def ie_preprocess (document ):... sentences = nltk. sent_tokenize (document )... sentences = [nltk. word_tokenize (sent) for sent in sentences]... sentences = [nltk. pos_tag (sent) for sent in sentences]
 

Then we need to specify the types of information to be extracted:

That is, the syntax format:

 
Grammar = "NP: {<DT>? <JJ> * <NN> }"
 

DT is an attribute, JJ is an adjective, and NN is a noun.

 
CP = nltk. regexpparser (grammar)
 

Later use

 
Result = CP. parse (sentence)
 

Analyze statements

We will get a nltk. Tree. tree structure.

Then we use:

 
For N in chunked: If isinstance (n, nltk. Tree. Tree): If n. node = 'np ': A = N
 

This code gets the snippet we need

In reality, this method cannot remove non-English words.

We can add:

 
D = enchant. dict ("en_us ")
 

Such interpretation is used to delete and select.

I hope to help you.