simple-processing-text

import nltk
# ubah jadi tipe nltk    
def convertToText(filename):    
    # raw/str -> token/list -> convert ke nltk.Text    
    raw = open(filename).read()
    # type(raw) == string    
    tokens = nltk.word_tokenize(raw)
    # type(tokens) == list    
    # token bisa berupa tanda-baca{?.,etc}, pos = {adverb,adj,}    
    return nltk.Text(tokens)

def get_context(keyword,filename):    
    nltktxt = convertToText(filename)
    return nltktxt.concordance(keyword)

def convert_to_wordpos(rawstr):
    sentences = nltk.sent_tokenize(rawstr) 
    sentences = [nltk.word_tokenize(sent) for sent in sentences]
    sentences = [nltk.pos_tag(sent) for sent in sentences]
    return sentences
    
# ubah jadi word dan tag pos nya
def text_preprocess(filename):    
    rawstr = open(filename).read()
    return convert_to_wordpos(rawstr)

# chunk by np-chunk
def get_chunk_np(sentence):
    grammar = "NP: {<DT>?<JJ>*<NN>}" 
    cp = nltk.RegexpParser(grammar) 
    result = cp.parse(sentence) 
    #     hasil parse ini, target grammar akan diberi tag (NP )
    return result

# chunk by tag-patterns
def get_chunk_tagpattern(sentence):
    pass

# chunk by regex
def get_chunk_regex(sentence):
    pass

def printAllNPChunk(sentences):
#    jika sudah di print, selanjutnya adalah ekstrak relation
#    caranya gimana ?
    for sen in sentences:
        print(get_chunk_np(sen))
        
def find_by_postag(sentences, tagkeyword):
    for sent in sentences:
        for wt in sent: 
            word = wt[0]
            tag = wt[1]        
            if (tag == tagkeyword):
                print(word)

# find NP in sentences
def findNP(senteChunk):    
    for ch in senteChunk:
        if (type(ch)!= tuple and ch.label() == 'NP'):                    
            print(ch)

example use of function find_by_postag(sentences,tagkeyword).

sentences = text_preprocess('kim-method-4-5.txt')
tagkeyword = 'CD'
find_by_postag(sentences, tagkeyword)

Result in the list of Cardinality words in the text.
Since it contains a number, probably contain information focus on us