CODEWITHSUNDEEP
datasetpythonscikit-learnnlp
Angelo
Angelo

Reputation:

Prepare data for scikit-learn

I am working on a small NLP project of authorship attribution: I have some texts from two authors and I want to say who wrote them.

I have some pre-processed text (tokenized, pos-tagged, ect.) and I want to load it into sciki-learn.

The documents have this shape:

Testo   -   SPN Testo   testare+v+indic+pres+nil+1+sing testo+n+m+sing  O
:   -   XPS colon   colon+punc  O
"   -   XPO "   quotation_mark+punc O
Buongiorno  -   I   buongiorno  buongiorno+inter buongiorno+n+m+_   O
a   -   E   a   a+prep  O
tutti   -   PP  tutto   tutto+adj+m+plur+pst+ind tutto+pron+_+m+_+plur+ind  O
.   <eos>   XPS full_stop   full_stop+punc  O
Ci  -   PP  pro loc+pron+loc+_+3+_+clit pro+pron+accdat+_+1+plur+clit   O
sarebbe -   VI  essere  essere+v+cond+pres+nil+2+sing   O
molto   -   B   molto   molto+adj+m+sing+pst+ind

So it's a tab separeted text file of 6 columns (word, end of sentence marker, part of speech, lemma, morphological information and named entity recognition marker).

Every file represents a document to classify.

What would be the best way to shape them for scikit learn?

Upvotes: 1

Views: 768

Answers (1)

Diego
Diego

Reputation: 832

The structure they use in scikit-learn example https://scikit-learn.org/stable/tutorial/text_analytics/working_with_text_data.html# is described here http://scikit-learn.org/stable/modules/generated/sklearn.datasets.load_files.html

Replace this

# Load some categories from the training set
if opts.all_categories:
    categories = None
else:
    categories = [
        'alt.atheism',
        'talk.religion.misc',
        'comp.graphics',
        'sci.space',
    ]

if opts.filtered:
    remove = ('headers', 'footers', 'quotes')
else:
    remove = ()

print("Loading 20 newsgroups dataset for categories:")
print(categories if categories else "all")

data_train = fetch_20newsgroups(subset='train', categories=categories,
                                shuffle=True, random_state=42,
                                remove=remove)

data_test = fetch_20newsgroups(subset='test', categories=categories,
                               shuffle=True, random_state=42,
                               remove=remove)

with your data load statements, for example:

# Load some categories from the training set
categories = [
        'high',
        'low',
]

print("loading dataset for categories:")
print(categories if categories else "all")

train_path='c:/Users/username/Documents/SciKit/train'
data_train = load_files(train_path, encoding='latin1')

test_path='c:/Users/username/Documents/SciKit/test'
data_test = load_files(test_path, encoding='latin1')

and in each of train and test directories create "high" and "low" subdirectories for your category files.

Upvotes: 1

Related Questions