Reading in a SQL file and Using Count Vectorizer to get word occurences

Question

I wanto to read a SQL file and use CountVectorizer to get word occurences.

I have the following code so far:

import re
import pandas as pd
import numpy as np
from sklearn.feature_extraction.text import CountVectorizer




df = pd.read_sql(q, dlconn)
print(df)

count_vect = CountVectorizer()
X_train_counts= count_vect.fit_transform(df)

print(X_train_counts.shape)
print(count_vect.vocabulary_)

This gives an output of 'cat': 1, 'dog': 0

It seems that it is taking just the name of the column animal and counting from there.

How do I get it to access the full column and get a chart that shows every word in the column and its frequency?

Answer 1

According to the CountVectorizer docs , the method fit_transform() expects an iterable of strings. It cannot handle a DataFrame directly.

But iterating over a dataframe returns the labels of the columns, not the values. I suggest you try df.itertuples() instead.

Try something like this:

value_list = [
    row[0]
    for row in df.itertuples(index=False, name=None)]
print(value_list)
print(type(value_list))
print(type(value_list[0]))

X_train_counts = count_vect.fit_transform(value_list)

Each value in value_list should be of type str . Let us know if that helps.

---

Here is a little example:

>>> import pandas as pd
>>> df = pd.DataFrame(['my big dog', 'my lazy cat'])
>>> df
             0
0   my big dog
1  my lazy cat

>>> value_list = [row[0] for row in df.itertuples(index=False, name=None)]
>>> value_list
['my big dog', 'my lazy cat']

>>> from sklearn.feature_extraction.text import CountVectorizer
>>> cv = CountVectorizer()
>>> x_train = cv.fit_transform(value_list)
>>> x_train
<2x5 sparse matrix of type '<class 'numpy.int64'>'
    with 6 stored elements in Compressed Sparse Row format>
>>> x_train.toarray()
array([[1, 0, 1, 0, 1],
       [0, 1, 0, 1, 1]], dtype=int64)
>>> cv.vocabulary_
{'my': 4, 'big': 0, 'dog': 2, 'lazy': 3, 'cat': 1}

And now you can display the word count of each row (each input string separately):

>>> for word, col in cv.vocabulary_.items():
...     for row in range(x_train.shape[0]):
...         print('word:{:10s} | row:{:2d} | count:{:2d}'.format(word, row, x_train[row,col]))
word:my         | row: 0 | count: 1
word:my         | row: 1 | count: 1
word:big        | row: 0 | count: 1
word:big        | row: 1 | count: 0
word:dog        | row: 0 | count: 1
word:dog        | row: 1 | count: 0
word:lazy       | row: 0 | count: 0
word:lazy       | row: 1 | count: 1
word:cat        | row: 0 | count: 0
word:cat        | row: 1 | count: 1

You can also display the total word count (sum of rows):

>>> x_train_sum = x_train.sum(axis=0)
>>> x_train_sum
    matrix([[1, 1, 1, 1, 2]], dtype=int64)
>>> for word, col in cv.vocabulary_.items():
...     print('word:{:10s} | count:{:2d}'.format(word, x_train_sum[0, col]))
word:my         | count: 2
word:big        | count: 1
word:dog        | count: 1
word:lazy       | count: 1
word:cat        | count: 1

>>> with open('my-file.csv', 'w') as f:
...     for word, col in cv.vocabulary_.items():
...         f.write('{};{}\n'.format(word, x_train_sum[0, col]))

This should clarify how you can use the tools you have.