text mining (nlp) with python

Author: Ties de Kok (Personal Website)
Last updated: June 2020
Python version: Python 3.7
License: MIT License

Note: Some features (like the ToC) will only work if you run the notebook or if you use nbviewer by clicking this link:
https://nbviewer.jupyter.org/github/TiesdeKok/Python_NLP_Tutorial/blob/master/NLP_Notebook.ipynb

Introduction

This notebook contains code examples to get you started with Python for Natural Language Processing (NLP) / Text Mining.

In the large scheme of things there are roughly 4 steps:

  1. Identify a data source
  2. Gather the data
  3. Process the data
  4. Analyze the data

This notebook only discusses step 3 and 4. If you want to learn more about step 2 see my Python tutorial.

Note: companion slides

This notebook was designed to accompany a PhD course session on NLP techniques in Accounting Research.
The slides of this session are publically availabe here: Slides

Elements / topics that are discussed in this notebook:

Table of Contents

Primer on NLP tools (to top)

There are many tools available for NLP purposes.
The code examples below are based on what I personally like to use, it is not intended to be a comprehsnive overview.

Besides build-in Python functionality I will use / demonstrate the following packages:

Standard NLP libraries:

  1. Spacy
  2. NLTK and the higher-level wrapper TextBlob

Note: besides installing the above packages you also often have to download (model) data . Make sure to check the documentation!

Standard machine learning library:

  1. scikit learn

Specific task libraries:

There are many, just a couple of examples:

  1. pyLDAvis for visualizing LDA)
  2. langdetect for detecting languages
  3. fuzzywuzzy for fuzzy text matching
  4. Gensim for topic modelling

Get some example data (to top)

There are many example datasets available to play around with, see for example this great repository:
https://archive.ics.uci.edu/ml/datasets.php

The data that I will use for most of the examples is the "Reuter_50_50 Data Set" that is used for author identification experiments.

See the details here: https://archive.ics.uci.edu/ml/datasets/Reuter_50_50

Download and load the data

Can't follow what I am doing here? Please see my Python tutorial (although the zipfile and io operations are not very relevant).

In [1]:
import requests, zipfile, io, os
from tqdm.notebook import tqdm

Note: for tqdm to work in JupyterLab you need to install the @jupyter-widgets/jupyterlab-manager using the puzzle icon in the left side bar.

Download and extract the zip file with the data

In [2]:
if not os.path.exists('C50test'):
    r = requests.get("https://archive.ics.uci.edu/ml/machine-learning-databases/00217/C50.zip")
    z = zipfile.ZipFile(io.BytesIO(r.content))
    z.extractall()

Load the data into memory

In [3]:
folder_dict = {'test' : 'C50test'}
text_dict = {'test' : {}}
In [4]:
for label, folder in tqdm(folder_dict.items()):
    authors = os.listdir(folder)
    for author in authors:
        text_files = os.listdir(os.path.join(folder, author))
        for file in text_files:
            with open(os.path.join(folder, author, file), 'r') as text_file:
                text_dict[label].setdefault(author, []).append(' '.join(text_file.readlines()))

Note: the text comes pre-split per sentence, for the sake of example I undo this through ' '.join(text_file.readlines()

In [5]:
text_dict['test']['TimFarrand'][0]
Out[5]:
'Shares in brewing-to-leisure group Bass Plc are likely to be held back until Britain\'s Trade and Industry secretary Ian Lang decides whether to allow its proposed merge with brewer Carlsberg-Tetley, said analysts.\n Earlier Lang announced the Bass deal would be referred to the Monoplies and Mergers Commission which is due to report before March 24, 1997. The shares fell 6p to 781p on the news.\n "The stock is probably dead in the water until March," said John Wakley, analyst at Lehman Brothers.  \n Dermott Carr, an analyst at Nikko said, "the market is going to hang onto them for the moment but until we get a decision they will be held back."\n Whatever the MMC decides many analysts expect Lang to defer a decision until after the next general election which will be called by May 22.\n "They will probably try to defer the decision until after the election. I don\'t think they want the negative PR of having a large number of people fired," said Wakley.  \n If the deal does not go through, analysts calculate the maximum loss to Bass of 60 million, with most sums centred on the 30-40 million range.\n "It\'s a maxiumum loss of 60 million for Bass if they fail and, unlike Allied, you would have to compare it to the perceived upside of doing the deal," said Wakley.\n Bass said at the time of the deal it would take a one-off charge of 75 million stg for restructuring the combined business, resulting in expected annual cost savings of 90 million stg within three years.  \n Under the terms of the complex deal, if Bass cannot combine C-T with its own brewing business within 16 months, it has the option to put its whole shareholding to Carlsberg for 110 million stg and Carlsberg has an option to put 15 percent of C-T to Allied Domecq, which would reimburse Bass 30 million stg.\n Bass is also entitled to receive 50 percent of all profits earnied by C-T until the merger is complete, which should give it some 30-35 million stg in a full year. Carlsberg has agreed to contribute its interests and 20 million stg in exchange for a 20 percent share in the combined Bass Breweries and Carlsberg-Tetley business.\n C-T was a joint venture between Allied Domecq and Carlsberg formed in 1992 by the merger of their UK brewing and wholesaleing businesses.\n -- London Newsroom +44 171 542 6437\n'

Process + Clean text (to top)

Convert the text into a NLP representation

We can use the text directly, but if want to use packages like spacy and textblob we first have to convert the text into a corresponding object.

Spacy

Note: depending on the way that you installed the language models you will need to import it differently:

from spacy.en import English
nlp = English()

OR

import en_core_web_sm
nlp = en_core_web_sm.load()

import en_core_web_md
nlp = en_core_web_md.load()

import en_core_web_lg
nlp = en_core_web_lg.load()
In [6]:
import spacy
import en_core_web_md
nlp = en_core_web_md.load()

Convert all text in the "test" sample to a spacy doc object using nlp.pipe():

In [7]:
spacy_text = {}
for author, text_list in tqdm(text_dict['test'].items()):
    spacy_text[author] = list(nlp.pipe(text_list))

A note on speed: This is slow because we didn't disable any compontents, see this note from the documentation:

Only apply the pipeline components you need. Getting predictions from the model that you don’t actually need adds up and becomes very inefficient at scale. To prevent this, use the disable keyword argument to disable components you don’t need – either when loading a model, or during processing with nlp.pipe. See the section on disabling pipeline components for more details and examples. link

In [8]:
type(spacy_text['TimFarrand'][0])
Out[8]:
spacy.tokens.doc.Doc

NLTK

In [9]:
import nltk

We can apply basic nltk operations directly to the text so we don't need to convert first.

TextBlob

In [10]:
from textblob import TextBlob

Convert all text in the "test" sample to a TextBlob object using TextBlob():

In [11]:
textblob_text = {}
for author, text_list in text_dict['test'].items():
    textblob_text[author] = [TextBlob(text) for text in text_list]
In [12]:
type(textblob_text['TimFarrand'][0])
Out[12]:
textblob.blob.TextBlob

Normalization (to top)

Text normalization describes the task of transforming the text into a different (more comparable) form.

This can imply many things, I will show a couple of options below:

Deal with unwanted characters (to top)

You will often notice that there are characters that you don't want in your text.

Let's look at this sentence for example:

"Shares in brewing-to-leisure group Bass Plc are likely to be held back until Britain\'s Trade and Industry secretary Ian Lang decides whether to allow its proposed merge with brewer Carlsberg-Tetley, said analysts.\n Earlier Lang announced the Bass deal would be referred to the Monoplies and Mergers"

You notice that there are some \ and \n in there. These are used to define how a string should be displayed, if we print this text we get:

In [13]:
text_dict['test']['TimFarrand'][0][:298]
Out[13]:
"Shares in brewing-to-leisure group Bass Plc are likely to be held back until Britain's Trade and Industry secretary Ian Lang decides whether to allow its proposed merge with brewer Carlsberg-Tetley, said analysts.\n Earlier Lang announced the Bass deal would be referred to the Monoplies and Mergers"
In [14]:
print(text_dict['test']['TimFarrand'][0][:298])
Shares in brewing-to-leisure group Bass Plc are likely to be held back until Britain's Trade and Industry secretary Ian Lang decides whether to allow its proposed merge with brewer Carlsberg-Tetley, said analysts.
 Earlier Lang announced the Bass deal would be referred to the Monoplies and Mergers

These special characters can cause problems in our analyses (and can be hard to debug if you are using print statements to inspect the data).

So how do we remove them?

In many cases it is sufficient to simply use the .replace() function:

In [15]:
text_dict['test']['TimFarrand'][0][:298].replace('\n', '').replace('\\', '')
Out[15]:
"Shares in brewing-to-leisure group Bass Plc are likely to be held back until Britain's Trade and Industry secretary Ian Lang decides whether to allow its proposed merge with brewer Carlsberg-Tetley, said analysts. Earlier Lang announced the Bass deal would be referred to the Monoplies and Mergers"

Sometimes, however, the problem arrises because of encoding / decoding problems.

In those cases you can usually do something like:

In [16]:
problem_sentence = 'This is some \u03c0 text that has to be cleaned\u2026! it\u0027s difficult to deal with!'
print(problem_sentence)
print(problem_sentence.encode().decode('unicode_escape').encode('ascii','ignore'))
This is some π text that has to be cleaned…! it's difficult to deal with!
b"This is some  text that has to be cleaned! it's difficult to deal with!"

An alternative that is better at preserving the unicode characters would be to use unidecode

In [17]:
import unidecode
In [18]:
print('\u738b\u7389')
王玉
In [19]:
unidecode.unidecode(u"\u738b\u7389")
Out[19]:
'Wang Yu '
In [20]:
unidecode.unidecode(problem_sentence)
Out[20]:
"This is some p text that has to be cleaned...! it's difficult to deal with!"

Sentence segmentation (to top)

Sentence segmentation refers to the task of splitting up the text by sentence.

You could do this by splitting on the . symbol, but dots are used in many other cases as well so it is not very robust:

In [21]:
text_dict['test']['TimFarrand'][0][:550].split('.')
Out[21]:
["Shares in brewing-to-leisure group Bass Plc are likely to be held back until Britain's Trade and Industry secretary Ian Lang decides whether to allow its proposed merge with brewer Carlsberg-Tetley, said analysts",
 '\n Earlier Lang announced the Bass deal would be referred to the Monoplies and Mergers Commission which is due to report before March 24, 1997',
 ' The shares fell 6p to 781p on the news',
 '\n "The stock is probably dead in the water until March," said John Wakley, analyst at Lehman Brothers',
 '  \n Dermott Carr, an analyst at Nikko said, "the mark']

It is better to use a more sophisticated implementation such as the one by Spacy:

In [22]:
example_paragraph = spacy_text['TimFarrand'][0]
In [23]:
sentence_list = [s for s in example_paragraph.sents]
sentence_list[:5]
Out[23]:
[Shares in brewing-to-leisure group Bass Plc are likely to be held back until Britain's Trade and Industry secretary Ian Lang decides whether to allow its proposed merge with brewer Carlsberg-Tetley, said analysts.
  ,
 Earlier Lang announced the Bass deal would be referred to the Monoplies and Mergers Commission which is due to report before March 24, 1997.,
 The shares fell 6p to 781p on the news.
  ,
 "The stock is probably dead in the water until March," said John Wakley, analyst at Lehman Brothers.  
  ,
 Dermott Carr, an analyst at Nikko said, "the market is going to hang onto them for the moment]

Notice that the returned object is still a spacy object:

In [24]:
type(sentence_list[0])
Out[24]:
spacy.tokens.span.Span

Note: spacy sentence segmentation relies on the text being capitalized, so make sure you didn't convert it to all lower case before running this operation.

Apply to all texts (for use later on):

In [25]:
spacy_sentences = {}
for author, text_list in tqdm(spacy_text.items()):
    spacy_sentences[author] = [list(text.sents) for text in text_list]

In [26]:
spacy_sentences['TimFarrand'][0][:3]
Out[26]:
[Shares in brewing-to-leisure group Bass Plc are likely to be held back until Britain's Trade and Industry secretary Ian Lang decides whether to allow its proposed merge with brewer Carlsberg-Tetley, said analysts.
  ,
 Earlier Lang announced the Bass deal would be referred to the Monoplies and Mergers Commission which is due to report before March 24, 1997.,
 The shares fell 6p to 781p on the news.
  ]

Word tokenization (to top)

Word tokenization means to split the sentence (or text) up into words.

In [27]:
example_sentence = spacy_sentences['TimFarrand'][0][0]
example_sentence
Out[27]:
Shares in brewing-to-leisure group Bass Plc are likely to be held back until Britain's Trade and Industry secretary Ian Lang decides whether to allow its proposed merge with brewer Carlsberg-Tetley, said analysts.
 

A word is called a token in this context (hence tokenization), using spacy:

In [28]:
token_list = [token for token in example_sentence]
token_list[0:15]
Out[28]:
[Shares,
 in,
 brewing,
 -,
 to,
 -,
 leisure,
 group,
 Bass,
 Plc,
 are,
 likely,
 to,
 be,
 held]

Lemmatization & Stemming (to top)

In some cases you want to convert a word (i.e. token) into a more general representation.

For example: convert "car", "cars", "car's", "cars'" all into the word car.

This is generally done through lemmatization / stemming (different approaches trying to achieve a similar goal).

Spacy

Space offers build-in functionality for lemmatization:

In [29]:
lemmatized = [token.lemma_ for token in example_sentence]
lemmatized[0:15]
Out[29]:
['share',
 'in',
 'brewing',
 '-',
 'to',
 '-',
 'leisure',
 'group',
 'Bass',
 'Plc',
 'be',
 'likely',
 'to',
 'be',
 'hold']

NLTK

Using the NLTK libary we can also use the more aggressive Porter Stemmer

In [30]:
from nltk.stem.porter import PorterStemmer
stemmer = PorterStemmer()
In [31]:
stemmed = [stemmer.stem(token.text) for token in example_sentence]
stemmed[0:15]
Out[31]:
['share',
 'in',
 'brew',
 '-',
 'to',
 '-',
 'leisur',
 'group',
 'bass',
 'plc',
 'are',
 'like',
 'to',
 'be',
 'held']

Compare:

In [32]:
print('  Original  | Spacy Lemma  | NLTK Stemmer')
print('-' * 41)
for original, lemma, stem in zip(token_list[:15], lemmatized[:15], stemmed[:15]):
    print(str(original).rjust(10, ' '), ' | ', str(lemma).rjust(10, ' '), ' | ', str(stem).rjust(10, ' '))
  Original  | Spacy Lemma  | NLTK Stemmer
-----------------------------------------
    Shares  |       share  |       share
        in  |          in  |          in
   brewing  |     brewing  |        brew
         -  |           -  |           -
        to  |          to  |          to
         -  |           -  |           -
   leisure  |     leisure  |      leisur
     group  |       group  |       group
      Bass  |        Bass  |        bass
       Plc  |         Plc  |         plc
       are  |          be  |         are
    likely  |      likely  |        like
        to  |          to  |          to
        be  |          be  |          be
      held  |        hold  |        held

In my experience it is usually best to use lemmatization instead of a stemmer.

Language modeling (to top)

Text is inherently structured in complex ways, we can often use some of this underlying structure.

Part-of-Speech tagging (to top)

Part of speech tagging refers to the identification of words as nouns, verbs, adjectives, etc.

Using Spacy:

In [33]:
pos_list = [(token, token.pos_) for token in example_sentence]
pos_list[0:10]
Out[33]:
[(Shares, 'NOUN'),
 (in, 'ADP'),
 (brewing, 'NOUN'),
 (-, 'PUNCT'),
 (to, 'ADP'),
 (-, 'PUNCT'),
 (leisure, 'NOUN'),
 (group, 'NOUN'),
 (Bass, 'PROPN'),
 (Plc, 'PROPN')]

Uni-Gram & N-Grams (to top)

Obviously a sentence is not a random collection of words, the sequence of words has information value.

A simple way to incorporate some of this sequence is by using what is called n-grams.
An n-gram is nothing more than a a combination of N words into one token (a uni-gram token is just one word).

So we can convert "Sentence about flying cars" into a list of bigrams:

Sentence-about, about-flying, flying-cars

See my slide on N-Grams for a more comprehensive example: click here

Using NLTK:

In [34]:
bigram_list = ['-'.join(x) for x in nltk.bigrams([token.text for token in example_sentence])]
bigram_list[10:15]
Out[34]:
['are-likely', 'likely-to', 'to-be', 'be-held', 'held-back']

Using spacy

In [35]:
def tokenize_without_punctuation(sen_obj):
    return [token.text for token in sen_obj if token.is_alpha]
In [36]:
def create_ngram(sen_obj, n, sep = '-'):
    token_list = tokenize_without_punctuation(sen_obj)
    number_of_tokens = len(token_list)
    ngram_list = []
    for i, token in enumerate(token_list[:-n+1]):
        ngram_item = [token_list[i + ii] for ii in range(n)]
        ngram_list.append(sep.join(ngram_item))
    return ngram_list
In [37]:
create_ngram(example_sentence, 2)[:5]
Out[37]:
['Shares-in', 'in-brewing', 'brewing-to', 'to-leisure', 'leisure-group']
In [38]:
create_ngram(example_sentence, 3)[:5]
Out[38]:
['Shares-in-brewing',
 'in-brewing-to',
 'brewing-to-leisure',
 'to-leisure-group',
 'leisure-group-Bass']

Stop words (to top)

Depending on what you are trying to do it is possible that there are many words that don't add any information value to the sentence.

The primary example are stop words.

Sometimes you can improve the accuracy of your model by removing stop words.

Using Spacy:

In [39]:
no_stop_words = [token for token in example_sentence if not token.is_stop]
In [40]:
no_stop_words[:10]
Out[40]:
[Shares, brewing, -, -, leisure, group, Bass, Plc, likely, held]
In [41]:
token_list[:10]
Out[41]:
[Shares, in, brewing, -, to, -, leisure, group, Bass, Plc]

Note we can also remove punctuation in the same way:

In [42]:
[token for token in example_sentence if not token.is_stop and token.is_alpha][:10]
Out[42]:
[Shares, brewing, leisure, group, Bass, Plc, likely, held, Britain, Trade]

Wrap everything into one function

Basic SpaCy text processing function

  1. Split into sentences
  2. Apply lemmatizer, remove top words, remove punctuation
  3. Clean up the sentence using textacy
In [43]:
def process_text_custom(text):
    sentences = list(nlp(text, disable=['tagger', 'ner', 'entity_linker', 'textcat', 'entitry_ruler']).sents)
    lemmatized_sentences = []
    for sentence in sentences:
        lemmatized_sentences.append([token.lemma_ for token in sentence if not token.is_stop and token.is_alpha])
    return [' '.join(sentence) for sentence in lemmatized_sentences]
In [44]:
spacy_text_clean = {}
for author, text_list in tqdm(text_dict['test'].items()):
    lst = []
    for text in text_list:
        lst.append(process_text_custom(text))
    spacy_text_clean[author] = lst

Note: that this would take quite a long time if we didn't disable some of the components.

In [45]:
count = 0
for author, texts in spacy_text_clean.items():
    for text in texts:
        count += len(text)
print('Number of sentences:', count)
Number of sentences: 58973

Result

In [46]:
spacy_text_clean['TimFarrand'][0][:3]
Out[46]:
['Shares brew leisure group Bass Plc likely hold Britain Trade Industry secretary Ian Lang decide allow propose merge brewer Carlsberg Tetley say analyst',
 'Earlier Lang announce Bass deal refer Monoplies Mergers Commission report March',
 'share fall news']

Note: the quality of the input text is not great, so the sentence segmentation is also not great (without further tweaking).

Direct feature extraction (to top)

We now have pre-processed our text into something that we can use for direct feature extraction or to convert it to a numerical representation.

Feature search (to top)

Entity recognition (to top)

It is often useful / relevant to extract entities that are mentioned in a piece of text.

SpaCy is quite powerful in extracting entities, however, it doesn't work very well on lowercase text.

Given that "token.lemma_" removes capitalization I will use spacy_sentences for this example.

In [47]:
example_sentence = spacy_sentences['TimFarrand'][0][3]
example_sentence
Out[47]:
"The stock is probably dead in the water until March," said John Wakley, analyst at Lehman Brothers.  
 
In [48]:
[(i, i.label_) for i in nlp(example_sentence.text).ents]
Out[48]:
[(March, 'DATE'), (John Wakley, 'PERSON'), (Lehman Brothers, 'ORG')]
In [49]:
example_sentence = spacy_sentences['TimFarrand'][4][0]
example_sentence
Out[49]:
British pub-to-hotel group Greenalls Plc on Thursday reported a 48 percent rise in profits before exceptional items to 148.7 million pounds ($246.4 million), driven by its acquisition of brewer Boddington in November 1995.
 
In [50]:
[(i, i.label_) for i in nlp(example_sentence.text).ents]
Out[50]:
[(British, 'NORP'),
 (Greenalls Plc, 'ORG'),
 (Thursday, 'DATE'),
 (48 percent, 'PERCENT'),
 (148.7 million pounds, 'MONEY'),
 ($246.4 million, 'MONEY'),
 (Boddington, 'ORG'),
 (November 1995, 'DATE')]

Pattern search (to top)

Using the build-in re (regular expression) library you can pattern match nearly anything you want.

I will not go into details about regular expressions but see here for a tutorial:
https://regexone.com/references/python

In [51]:
import re

TIP: Use Pythex.org to try out your regular expression

Example on Pythex: click here

Example 1:

In [52]:
string_1 = 'Ties de Kok (#IDNUMBER: 123-AZ). Rest of text...'
string_2 = 'Philip Joos (#IDNUMBER: 663-BY). Rest of text...'
In [53]:
pattern = r'#IDNUMBER: (\d\d\d-\w\w)'
In [54]:
print(re.findall(pattern, string_1)[0])
print(re.findall(pattern, string_2)[0])
123-AZ
663-BY

Example 2:

If a sentence contains the word 'million' return True, otherwise return False

In [55]:
for sen in spacy_text_clean['TimFarrand'][2]:
    TERM = 'million'
    if re.search('million', sen, flags= re.IGNORECASE):
        print(sen)
Analysts forecast pretax profit range million stg restructure cost million time
restructure cost million anticipate bulk million stem closure small production plant France
Cadbury drink business turn million stg trade profit million half entirely contribution Dr Pepper
Campbell estimate UK beverage contribute million stg operate profit million time
Broadly analyst expect pretty flat performance group confectionery business consensus forecast million stg operate profit
average analyst calculate beverage chip trade profit million
sale percent stake Coca Cola amp Schweppes Beverages CCSB operation Coca Cola Enterprises June million stg analyst want clear statement strategy company
far analyst company say shareholder expect return investment emerge market large far million Russian plant
Cadbury announce investment million stg build new plant Wrocoaw Poland joint venture China cost million
Net debt billion end fall million end result CCSB sale provide acquisition

Text evaluation (to top)

Besides feature search there are also many ways to analyze the text as a whole.

Let's, for example, evaluate the following paragraph:

In [56]:
example_paragraph = ' '.join([x for x in spacy_text_clean['TimFarrand'][2]])
example_paragraph[:500]
Out[56]:
'Soft drink confectionery group Cadbury Schweppes Plc expect report solid percent rise half profit Wednesday face question performance soft drink main question success relaunch brand say Mark Duffy food manufacture analyst SBC Warburg Competitor Sprite own Coca Cola see agressive market push rank fast grow brand Cadbury Dr Pepper Analysts forecast pretax profit range million stg restructure cost million time dividend penny expect restructure cost million anticipate bulk million stem closure small'

Language (to top)

Using the spacy-langdetect package it is easy to detect the language of a piece of text

In [57]:
from spacy_langdetect import LanguageDetector
nlp.add_pipe(LanguageDetector(), name='language_detector', last=True)
In [58]:
print(nlp(example_paragraph)._.language)
{'language': 'en', 'score': 0.9999970401265338}

Readability (to top)

Generally I'd recommend to calculate the readability metrics by yourself as they don't tend to be that difficult to compute. However, there are packages out there that can help, such as spacy_readability

In [59]:
from spacy_readability import Readability
In [60]:
nlp.add_pipe(Readability(), name='readability', last=True)
In [61]:
doc = nlp("I am some really difficult text to read because I use obnoxiously large words.")
print(doc._.flesch_kincaid_grade_level)
print(doc._.smog)
8.412857142857145
0

Manual example: FOG index

In [62]:
import syllapy
In [63]:
def calculate_fog(document):
    doc = nlp(document, disable=['tagger', 'ner', 'entity_linker', 'textcat', 'entitry_ruler'])
    sen_list = list(doc.sents)
    num_sen = len(sen_list)

    num_words = 0
    num_complex_words = 0
    for sen_obj in sen_list:
        words_in_sen = [token.text for token in sen_obj if token.is_alpha]
        num_words += len(words_in_sen)
        num_complex  = 0
        for word in words_in_sen:
            num_syl = syllapy.count(word.lower())
            if num_syl > 2:
                num_complex += 1
        num_complex_words += num_complex
        
    fog = 0.4 * ((num_words / num_sen) + ((num_complex_words / num_words)*100))
    return {'fog' : fog, 
            'num_sen' : num_sen, 
            'num_words' : num_words, 
            'num_complex_words' : num_complex_words}
In [64]:
calculate_fog(example_paragraph)
Out[64]:
{'fog': 13.42327889849504,
 'num_sen': 36,
 'num_words': 347,
 'num_complex_words': 83}

Text similarity

Using fuzzywuzzy

In [65]:
from fuzzywuzzy import fuzz
In [66]:
fuzz.ratio("fuzzy wuzzy was a bear", "wuzzy fuzzy was a bear")
Out[66]:
91

Using spacy

Spacy can provide a similary score based on the semantic similarity (link)

In [67]:
tokens_1 = nlp("fuzzy wuzzy was a bear")
tokens_2 = nlp("wuzzy fuzzy was a bear")

tokens_1.similarity(tokens_2)
Out[67]:
1.0000000623731768
In [68]:
tokens_1 = nlp("Tom believes German cars are the best.")
tokens_2 = nlp("Sarah recently mentioned that she would like to go on holiday to Germany.")

tokens_1.similarity(tokens_2)
Out[68]:
0.8127869114665882

Term (dictionary) counting (to top)

A common technique for basic NLP insights is to create simple metrics based on term counts.

These are relatively easy to implement.

Example 1:

In [69]:
word_dictionary = ['soft', 'first', 'most', 'be']
In [70]:
for word in word_dictionary:
    print(word, example_paragraph.count(word))
soft 2
first 0
most 0
be 7

Example 2:

In [71]:
pos = ['great', 'agree', 'increase']
neg = ['bad', 'disagree', 'decrease']

sentence = '''According to the president everything is great, great, 
and great even though some people might disagree with those statements.'''

pos_count = 0
for word in pos:
    pos_count += sentence.lower().count(word)
print(pos_count)

neg_count = 0
for word in neg:
    neg_count += sentence.lower().count(word)
print(neg_count)

pos_count / (neg_count + pos_count)
4
1
Out[71]:
0.8

Getting the total number of words is also easy:

In [72]:
num_tokens = len([token for token in nlp(sentence) if token.is_alpha])
num_tokens
Out[72]:
19

Example 3:

We can also save the count per word

In [73]:
pos_count_dict = {}
for word in pos:
    pos_count_dict[word] = sentence.lower().count(word)
In [74]:
pos_count_dict
Out[74]:
{'great': 3, 'agree': 1, 'increase': 0}

Note: .lower() is actually quite slow, if you have a lot of words / sentences it is recommend to minimize the amount of .lower() operations that you have to make.

Represent text numerically (to top)

Bag of Words (to top)

Sklearn includes the CountVectorizer and TfidfVectorizer function.

For details, see the documentation:
TF
TFIDF

Note 1: these functions also provide a variety of built-in preprocessing options (e.g. ngrames, remove stop words, accent stripper).

Note 2: example based on the following website click here

In [75]:
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.feature_extraction.text import TfidfVectorizer

Simple example:

In [76]:
doc_1 = "The sky is blue."
doc_2 = "The sun is bright today."
doc_3 = "The sun in the sky is bright."
doc_4 = "We can see the shining sun, the bright sun."

Calculate term frequency:

In [77]:
vectorizer = CountVectorizer(stop_words='english')
tf = vectorizer.fit_transform([doc_1, doc_2, doc_3, doc_4])
In [78]:
print(vectorizer.get_feature_names(), '\n')
for doc_tf_vector in tf.toarray():
    print(doc_tf_vector)
['blue', 'bright', 'shining', 'sky', 'sun', 'today'] 

[1 0 0 1 0 0]
[0 1 0 0 1 1]
[0 1 0 1 1 0]
[0 1 1 0 2 0]

TF-IDF (to top)

In [79]:
transformer = TfidfVectorizer(stop_words='english')
tfidf = transformer.fit_transform([doc_1, doc_2, doc_3, doc_4])
In [80]:
for doc_vector in tfidf.toarray():
    print(doc_vector)
[0.78528828 0.         0.         0.6191303  0.         0.        ]
[0.         0.47380449 0.         0.         0.47380449 0.74230628]
[0.         0.53256952 0.         0.65782931 0.53256952 0.        ]
[0.         0.36626037 0.57381765 0.         0.73252075 0.        ]

More elaborate example:

In [81]:
clean_paragraphs = []
for author, value in spacy_text_clean.items():
    for article in value:
        clean_paragraphs.append(' '.join([x for x in article]))
In [82]:
len(clean_paragraphs)
Out[82]:
2500
In [83]:
transformer = TfidfVectorizer(stop_words='english')
tfidf_large = transformer.fit_transform(clean_paragraphs)
In [84]:
print('Number of vectors:', len(tfidf_large.toarray()))
print('Number of words in dictionary:', len(tfidf_large.toarray()[0]))
Number of vectors: 2500
Number of words in dictionary: 21978
In [85]:
tfidf_large
Out[85]:
<2500x21978 sparse matrix of type '<class 'numpy.float64'>'
	with 410121 stored elements in Compressed Sparse Row format>

Word Embeddings (to top)

The en_core_web_lg language model comes with GloVe vectors trained on the Common Crawl dataset (link)

In [86]:
tokens = nlp("The Dutch word for peanut butter is 'pindakaas', did you know that? This is a typpo.")

for token in tokens:
    if token.is_alpha:
        print(token.text, token.has_vector, token.vector_norm, token.is_oov)
The True 4.70935 False
Dutch True 6.381229 False
word True 5.8387117 False
for True 4.8435082 False
peanut True 7.085804 False
butter True 7.466713 False
is True 4.890306 False
pindakaas False 0.0 False
did True 5.284421 False
you True 5.1979666 False
know True 5.160699 False
that True 4.8260193 False
This True 5.0461264 False
is True 4.890306 False
a True 5.306696 False
typpo False 0.0 True
In [87]:
token = nlp('Car')
print('The token: "{}" has the following vector (dimension: {})'.format(token.text, len(token.vector)))
token.vector
The token: "Car" has the following vector (dimension: 300)
Out[87]:
array([ 2.0987e-01,  4.6481e-01, -2.4238e-01, -6.5751e-02,  6.0856e-01,
       -3.4698e-01, -2.5331e-01, -4.2590e-01, -2.2277e-01,  2.2913e+00,
       -3.3853e-01,  2.3275e-01, -2.7511e-01,  2.4064e-01, -1.0697e+00,
       -2.6978e-01, -8.0733e-01,  1.8698e+00,  4.5562e-01, -1.4469e-01,
        1.6246e-02, -3.5473e-01,  7.6152e-01, -6.8589e-02,  1.2156e-02,
        9.0520e-03,  1.1131e-01, -3.0746e-01,  2.4168e-01,  1.1400e-01,
        4.3952e-01, -6.6594e-01, -7.3198e-02,  8.0566e-01,  1.1748e-01,
       -3.8758e-01,  1.0691e-01,  3.3697e-01, -1.3188e-01,  1.9364e-01,
        5.5553e-01, -3.4029e-01,  1.7059e-01,  4.0736e-01, -1.6150e-01,
        7.0302e-02,  6.7772e-02, -8.1763e-01,  3.0645e-01, -9.9862e-03,
        9.4606e-02, -5.9763e-01,  1.4192e-01,  1.4857e-01, -3.1535e-01,
        9.9092e-02,  2.0673e-01, -4.4041e-01,  2.1519e-01, -4.1294e-01,
        2.6374e-01, -1.5493e-01,  2.4739e-01,  4.2090e-01,  1.8768e-01,
        4.6904e-02,  9.6848e-02,  2.7431e-02,  1.0633e-01,  3.1926e-01,
       -7.6260e-01, -8.8373e-02,  3.7519e-01,  4.7369e-01, -7.3557e-01,
       -1.0760e-01, -2.6557e-02, -5.1079e-01, -1.8886e-01,  2.8679e-01,
        6.5798e-02,  5.7129e-01,  2.5056e-01,  7.3858e-02,  8.4700e-03,
        1.5158e-02,  7.3570e-01,  6.2549e-01,  5.1600e-02, -2.5802e-01,
       -8.1203e-02,  1.3731e-01,  1.8809e-01, -6.5871e-01, -2.2361e-01,
       -3.3318e-01,  1.5853e-01,  5.1523e-01,  5.0259e-01, -1.6894e-01,
       -8.6465e-02,  2.5036e-01, -1.7419e-01, -2.7723e-02,  1.1262e-01,
       -4.6449e-01,  1.6956e-01,  2.8931e-01, -1.3187e-01,  4.6368e-01,
       -2.9348e-01, -3.1244e-01,  6.5886e-01, -4.7842e-01,  1.4754e-01,
       -3.0646e-01,  4.3847e-01,  1.7684e-01, -1.1968e-01, -3.1002e-02,
       -1.2228e-01, -5.6424e-01,  1.5289e-01, -7.9389e-01, -3.6731e-01,
        1.6918e-01, -9.5210e-02,  1.6490e-01,  1.5936e-01,  1.2460e-01,
        3.8846e-01,  2.3019e-01, -1.3054e-01, -2.1932e-01, -2.6782e-01,
       -6.0745e-01, -3.4826e-01,  1.7656e-01, -1.0351e-01, -2.2750e-01,
       -1.6111e+00, -4.0504e-01,  1.0872e+00, -1.6391e-01,  6.5586e-02,
       -1.0632e-01, -1.4014e-01,  1.7712e-01,  7.1100e-01,  2.0313e-01,
       -5.0138e-01,  1.7291e-01, -5.3208e-02, -4.0668e-01,  1.4907e-01,
       -3.0631e-01,  4.6572e-01,  3.7977e-01, -1.3336e-01, -7.6937e-02,
        1.1803e-02, -1.1185e-01,  7.0364e-01, -6.8615e-02,  5.8586e-01,
       -5.9890e-01, -2.8104e-01,  4.9674e-01,  4.5867e-01,  1.6291e-01,
       -2.8317e-01,  3.8870e-01, -3.9882e-01,  2.2407e-01, -2.3704e-01,
       -2.3155e-01,  6.7882e-02,  8.3828e-01,  1.3231e-01,  2.9778e-01,
        1.8471e-01, -9.7415e-05, -6.9993e-01,  4.6959e-03, -3.5461e-01,
       -9.6413e-02,  1.0312e-01,  8.5293e-02, -2.6909e-01,  4.3886e-01,
        3.1275e-01,  2.2829e-01,  4.8072e-01,  1.8399e-01, -1.7628e-01,
       -4.8322e-01,  9.5676e-02, -2.4499e-01,  5.8915e-02, -3.9355e-02,
       -4.6954e-01, -2.6272e-01,  1.5462e-01, -1.8055e-01,  1.6881e-03,
        5.7027e-02, -6.7284e-02,  2.4853e-01,  3.5735e-01,  1.4325e-01,
       -4.9276e-01, -2.9321e-02,  5.1167e-02,  4.9620e-01,  3.7308e-01,
        4.0203e-01,  9.2905e-02,  7.4061e-01, -3.3765e-01, -3.5641e-01,
        6.1675e-01, -9.5517e-01, -2.7492e-01,  2.2079e-01, -2.8898e-01,
       -1.5504e-01, -3.1433e-01,  5.8383e-01, -2.6138e-02, -2.7755e-01,
       -4.7184e-02,  1.0504e-01, -4.2419e-01, -1.6414e-01, -5.0711e-01,
        4.2617e-01,  3.6889e-01,  4.3267e-01, -4.4480e-03,  5.6442e-01,
       -3.0964e-02,  7.7629e-02,  2.2218e-01,  1.2818e-01, -1.6235e-01,
       -2.2912e-01,  4.9174e-01, -5.1937e-01, -2.0793e-01, -3.6868e-01,
       -5.5714e-01, -1.9930e-01,  2.9782e-01,  7.5921e-02, -3.9895e-01,
        8.1692e-01, -1.0221e-01, -3.8049e-01,  1.9906e-01, -1.9875e-02,
        7.3431e-02, -1.3882e-01,  2.7914e-01, -4.5367e-01,  2.9227e-01,
       -5.5489e-01, -4.2121e-01,  5.5667e-01, -4.5230e-01, -1.1956e-01,
        1.3504e-01, -2.3580e-01,  7.4221e-01, -2.7890e-01, -5.4580e-02,
        3.1944e-01,  3.6717e-01,  1.3430e-01,  1.3629e-01, -9.7458e-02,
       -6.0310e-01, -1.7762e-01,  2.5910e-01,  3.3150e-01,  2.2701e-01,
        6.3664e-01,  1.5324e-01, -3.2894e-01, -3.6749e-01, -2.0328e-01,
       -1.1924e+00, -4.6395e-01,  6.6984e-01, -4.9404e-01,  4.4154e-01,
       -4.3699e-01,  2.3538e-01,  3.2135e-01,  2.6649e-01,  2.2438e-01],
      dtype=float32)

Word2Vec (to top)

Note: you might have to run nltk.download('brown') to install the NLTK corpus files

In [88]:
import gensim
from nltk.corpus import brown
In [89]:
sentences = brown.sents()
In [90]:
model = gensim.models.Word2Vec(sentences, min_count=1)

Save model

In [91]:
model.save('brown_model')

Load model

In [92]:
model = gensim.models.Word2Vec.load('brown_model')

Find words most similar to 'mother':

In [93]:
print(model.wv.most_similar("mother"))
[('father', 0.9847322702407837), ('husband', 0.9656894207000732), ('wife', 0.9496790170669556), ('friend', 0.9323333501815796), ('son', 0.9279097318649292), ('nickname', 0.9207977652549744), ('eagle', 0.9097722768783569), ('addiction', 0.9071668982505798), ('voice', 0.9051918983459473), ('patient', 0.8966456055641174)]

Find the odd one out:

In [94]:
print(model.wv.doesnt_match("breakfast cereal dinner lunch".split()))
cereal
D:\anaconda\envs\limpergPython\lib\site-packages\gensim\models\keyedvectors.py:877: FutureWarning: arrays to stack must be passed as a "sequence" type such as list or tuple. Support for non-sequence iterables such as generators is deprecated as of NumPy 1.16 and will raise an error in the future.
  vectors = vstack(self.word_vec(word, use_norm=True) for word in used_words).astype(REAL)
In [95]:
print(model.wv.doesnt_match("pizza pasta garden fries".split()))
garden

Retrieve vector representation of the word "human"

In [96]:
model.wv['human']
Out[96]:
array([-0.40230748, -0.36424252,  0.48136342,  0.37950972,  0.25889766,
       -0.10284429,  0.2539682 , -0.557558  , -1.0109891 , -0.30390584,
        0.02308058, -0.9227236 ,  0.06572528, -0.24571045, -0.17627639,
        0.35974783, -0.29690138, -0.25977215,  0.465111  , -1.5026555 ,
        0.23448153, -0.79958564, -0.6682266 , -0.51277363, -0.11112369,
       -1.4914587 , -0.30484447,  1.3466982 , -0.45936054,  0.02780625,
        0.31517667, -0.12471037,  0.46333146, -0.29451668,  0.28516975,
        1.3195679 ,  0.02986159,  0.27836317, -0.5356812 , -0.5574794 ,
        0.55741835, -0.3692916 ,  0.3067411 , -0.62016165,  0.6085465 ,
        0.6336735 ,  0.9925447 , -0.2553504 , -0.3593044 , -0.29228973,
       -0.05774796, -0.22645272, -0.594325  , -0.19128117,  0.13758877,
        0.58251387, -0.12266693, -0.33289537, -0.81493866,  0.64220285,
       -0.40921453,  1.7995448 ,  0.98320687,  0.66162825, -0.03371862,
        0.30391327,  0.30519032, -0.02499808,  0.46001107, -0.5412774 ,
       -0.14508785,  0.47390515, -0.01815019,  0.39801887,  0.33498788,
       -0.70357895,  0.80516887,  0.08044272, -0.70585257, -0.7256744 ,
       -0.95714486, -0.12571876, -0.20877206, -0.456315  ,  0.7478423 ,
       -0.25637153,  0.78873783,  0.24834621,  0.4455648 ,  0.1293853 ,
        0.17152755, -0.30077967,  0.5803442 ,  0.16445744,  0.60369337,
       -0.2301575 , -0.19547687, -0.36981392,  0.23723377,  0.24412107],
      dtype=float32)

Statistical models (to top)

"Traditional" machine learning (to top)

The library to use for machine learning is scikit-learn ("sklearn").

Supervised (to top)

In [97]:
from sklearn.model_selection import cross_val_score, KFold, train_test_split
from sklearn.pipeline import Pipeline
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn import metrics
import joblib
In [98]:
import pandas as pd
import numpy as np

Convert the data into a pandas dataframe (so that we can input it easier)

In [99]:
article_list = []
for author, value in spacy_text_clean.items():
    for article in value:
        article_list.append((author, ' '.join([x for x in article])))
In [100]:
article_df = pd.DataFrame(article_list, columns=['author', 'text'])
In [101]:
article_df.sample(5)
Out[101]:
author text
1325 LydiaZajc Canadian unit Wal Mart Stores Inc bow Canadian...
1516 MarkBendeich Australia labour market watchdog expect award ...
2121 SarahDavison Temperatures rise Hong Kong prepare midsummer ...
2226 SimonCowell British composite insurer Commercial Union Plc...
1914 PierreTran French defence electronic firm Thomson CSF soo...

Split the sample into a training and test sample

In [102]:
X_train, X_test, y_train, y_test = train_test_split(article_df.text, article_df.author, test_size=0.20, random_state=3561)
In [103]:
print(len(X_train), len(X_test))
2000 500

Train and evaluate function

Simple function to train (i.e. fit) and evaluate the model

In [104]:
def train_and_evaluate(clf, X_train, X_test, y_train, y_test):
    
    clf.fit(X_train, y_train)
    
    print("Accuracy on training set:")
    print(clf.score(X_train, y_train))
    print("Accuracy on testing set:")
    print(clf.score(X_test, y_test))
    
    y_pred = clf.predict(X_test)
    
    print("Classification Report:")
    print(metrics.classification_report(y_test, y_pred))

Naïve Bayes estimator (to top)

In [105]:
from sklearn.naive_bayes import MultinomialNB

Define pipeline

In [106]:
clf = Pipeline([
    ('vect', TfidfVectorizer(strip_accents='unicode',
                             lowercase = True,
                            max_features = 1500,
                            stop_words='english'
                            )),
        
    ('clf', MultinomialNB(alpha = 1,
                          fit_prior = True
                          )
    ),
])

Train and show evaluation stats

In [107]:
train_and_evaluate(clf, X_train, X_test, y_train, y_test)
Accuracy on training set:
0.8345
Accuracy on testing set:
0.72
Classification Report:
                   precision    recall  f1-score   support

    AaronPressman       0.82      1.00      0.90         9
       AlanCrosby       0.55      0.92      0.69        12
   AlexanderSmith       0.86      0.60      0.71        10
  BenjaminKangLim       0.75      0.27      0.40        11
    BernardHickey       0.75      0.30      0.43        10
      BradDorfman       0.80      1.00      0.89         8
 DarrenSchuettler       0.58      0.78      0.67         9
      DavidLawder       1.00      0.60      0.75        10
    EdnaFernandes       1.00      0.67      0.80         9
      EricAuchard       0.86      0.67      0.75         9
   FumikoFujisaki       1.00      1.00      1.00        10
   GrahamEarnshaw       0.59      1.00      0.74        10
 HeatherScoffield       0.83      0.56      0.67         9
       JanLopatka       0.38      0.33      0.35         9
    JaneMacartney       0.33      0.60      0.43        10
     JimGilchrist       0.73      1.00      0.84         8
   JoWinterbottom       0.90      0.90      0.90        10
         JoeOrtiz       0.80      0.89      0.84         9
     JohnMastrini       0.83      0.29      0.43        17
     JonathanBirt       0.53      1.00      0.70         8
      KarlPenhaul       0.87      1.00      0.93        13
        KeithWeir       0.69      0.90      0.78        10
   KevinDrawbaugh       0.88      0.70      0.78        10
    KevinMorrison       0.33      1.00      0.50         3
    KirstinRidley       0.86      0.67      0.75         9
KouroshKarimkhany       0.54      0.88      0.67         8
        LydiaZajc       0.90      0.90      0.90        10
   LynneO'Donnell       0.89      0.73      0.80        11
  LynnleyBrowning       0.93      1.00      0.96        13
  MarcelMichelson       1.00      0.50      0.67        12
     MarkBendeich       0.86      0.55      0.67        11
       MartinWolk       0.57      0.80      0.67         5
     MatthewBunce       1.00      0.86      0.92        14
    MichaelConnor       0.83      0.77      0.80        13
       MureDickie       0.44      0.40      0.42        10
        NickLouth       0.83      1.00      0.91        10
  PatriciaCommins       0.80      0.89      0.84         9
    PeterHumphrey       0.35      0.89      0.50         9
       PierreTran       0.56      0.83      0.67         6
       RobinSidel       1.00      1.00      1.00        12
     RogerFillion       1.00      0.88      0.93         8
      SamuelPerry       0.78      0.50      0.61        14
     SarahDavison       1.00      0.29      0.44        14
      ScottHillis       0.44      0.44      0.44         9
      SimonCowell       0.91      1.00      0.95        10
         TanEeLyn       1.00      0.57      0.73         7
   TheresePoletti       0.80      0.73      0.76        11
       TimFarrand       1.00      0.77      0.87        13
       ToddNissen       0.60      1.00      0.75         9
     WilliamKazer       0.00      0.00      0.00        10

         accuracy                           0.72       500
        macro avg       0.75      0.74      0.71       500
     weighted avg       0.77      0.72      0.71       500

Save results

In [108]:
joblib.dump(clf, 'naive_bayes_results.pkl')
Out[108]:
['naive_bayes_results.pkl']

Predict out of sample:

In [109]:
example_y, example_X = y_train[33], X_train[33]
In [110]:
print('Actual author:', example_y)
print('Predicted author:', clf.predict([example_X])[0])
Actual author: AaronPressman
Predicted author: AaronPressman

Support Vector Machines (SVM) (to top)

In [111]:
from sklearn.svm import SVC

Define pipeline

In [112]:
clf_svm = Pipeline([
    ('vect', TfidfVectorizer(strip_accents='unicode',
                             lowercase = True,
                            max_features = 1500,
                            stop_words='english'
                            )),
        
    ('clf', SVC(kernel='rbf' ,
                C=10, gamma=0.3)
    ),
])

Note: The SVC estimator is very sensitive to the hyperparameters!

Train and show evaluation stats

In [113]:
train_and_evaluate(clf_svm, X_train, X_test, y_train, y_test)
Accuracy on training set:
0.9975
Accuracy on testing set:
0.83
Classification Report:
                   precision    recall  f1-score   support

    AaronPressman       0.89      0.89      0.89         9
       AlanCrosby       0.79      0.92      0.85        12
   AlexanderSmith       1.00      0.70      0.82        10
  BenjaminKangLim       0.67      0.36      0.47        11
    BernardHickey       1.00      0.50      0.67        10
      BradDorfman       0.78      0.88      0.82         8
 DarrenSchuettler       0.89      0.89      0.89         9
      DavidLawder       1.00      0.60      0.75        10
    EdnaFernandes       0.73      0.89      0.80         9
      EricAuchard       0.73      0.89      0.80         9
   FumikoFujisaki       1.00      1.00      1.00        10
   GrahamEarnshaw       0.83      1.00      0.91        10
 HeatherScoffield       0.80      0.89      0.84         9
       JanLopatka       0.67      0.44      0.53         9
    JaneMacartney       0.36      0.50      0.42        10
     JimGilchrist       0.88      0.88      0.88         8
   JoWinterbottom       1.00      0.90      0.95        10
         JoeOrtiz       0.82      1.00      0.90         9
     JohnMastrini       0.83      0.88      0.86        17
     JonathanBirt       0.80      1.00      0.89         8
      KarlPenhaul       0.93      1.00      0.96        13
        KeithWeir       0.83      1.00      0.91        10
   KevinDrawbaugh       0.82      0.90      0.86        10
    KevinMorrison       0.50      1.00      0.67         3
    KirstinRidley       1.00      0.56      0.71         9
KouroshKarimkhany       0.88      0.88      0.88         8
        LydiaZajc       1.00      1.00      1.00        10
   LynneO'Donnell       0.82      0.82      0.82        11
  LynnleyBrowning       1.00      1.00      1.00        13
  MarcelMichelson       1.00      0.67      0.80        12
     MarkBendeich       0.79      1.00      0.88        11
       MartinWolk       0.83      1.00      0.91         5
     MatthewBunce       1.00      0.86      0.92        14
    MichaelConnor       1.00      0.85      0.92        13
       MureDickie       0.60      0.60      0.60        10
        NickLouth       0.90      0.90      0.90        10
  PatriciaCommins       1.00      1.00      1.00         9
    PeterHumphrey       0.57      0.89      0.70         9
       PierreTran       0.60      1.00      0.75         6
       RobinSidel       1.00      1.00      1.00        12
     RogerFillion       1.00      1.00      1.00         8
      SamuelPerry       0.81      0.93      0.87        14
     SarahDavison       1.00      0.71      0.83        14
      ScottHillis       0.67      0.44      0.53         9
      SimonCowell       1.00      0.90      0.95        10
         TanEeLyn       0.83      0.71      0.77         7
   TheresePoletti       0.91      0.91      0.91        11
       TimFarrand       0.92      0.85      0.88        13
       ToddNissen       0.90      1.00      0.95         9
     WilliamKazer       0.27      0.30      0.29        10

         accuracy                           0.83       500
        macro avg       0.84      0.83      0.82       500
     weighted avg       0.85      0.83      0.83       500

Save results

In [114]:
joblib.dump(clf_svm, 'svm_results.pkl')
Out[114]:
['svm_results.pkl']

Predict out of sample:

In [115]:
example_y, example_X = y_train[33], X_train[33]
In [116]:
print('Actual author:', example_y)
print('Predicted author:', clf_svm.predict([example_X])[0])
Actual author: AaronPressman
Predicted author: AaronPressman

Model Selection and Evaluation (to top)

Both the TfidfVectorizer and SVC() estimator take a lot of hyperparameters.

It can be difficult to figure out what the best parameters are.

We can use GridSearchCV to help figure this out.

In [117]:
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import make_scorer
from sklearn.metrics import f1_score

First we define the options that should be tried out:

In [118]:
clf_search = Pipeline([
    ('vect', TfidfVectorizer()),
    ('clf', SVC())
])
parameters = { 'vect__stop_words': ['english'],
                'vect__strip_accents': ['unicode'],
              'vect__max_features' : [1500],
              'vect__ngram_range': [(1,1), (2,2) ],
             'clf__gamma' : [0.2, 0.3, 0.4], 
             'clf__C' : [8, 10, 12],
              'clf__kernel' : ['rbf']
             }

Run everything:

In [119]:
grid = GridSearchCV(clf_search, 
                    param_grid=parameters, 
                    scoring=make_scorer(f1_score, average='micro'), 
                    n_jobs=-1
                   )
grid.fit(X_train, y_train)    
Out[119]:
GridSearchCV(estimator=Pipeline(steps=[('vect', TfidfVectorizer()),
                                       ('clf', SVC())]),
             n_jobs=-1,
             param_grid={'clf__C': [8, 10, 12], 'clf__gamma': [0.2, 0.3, 0.4],
                         'clf__kernel': ['rbf'], 'vect__max_features': [1500],
                         'vect__ngram_range': [(1, 1), (2, 2)],
                         'vect__stop_words': ['english'],
                         'vect__strip_accents': ['unicode']},
             scoring=make_scorer(f1_score, average=micro))

Note: if you are on a powerful (preferably unix system) you can set n_jobs to the number of available threads to speed up the calculation

In [120]:
print("The best parameters are %s with a score of %0.2f" % (grid.best_params_, grid.best_score_))
y_true, y_pred = y_test, grid.predict(X_test)
print(metrics.classification_report(y_true, y_pred))
The best parameters are {'clf__C': 8, 'clf__gamma': 0.4, 'clf__kernel': 'rbf', 'vect__max_features': 1500, 'vect__ngram_range': (1, 1), 'vect__stop_words': 'english', 'vect__strip_accents': 'unicode'} with a score of 0.79
                   precision    recall  f1-score   support

    AaronPressman       0.89      0.89      0.89         9
       AlanCrosby       0.79      0.92      0.85        12
   AlexanderSmith       1.00      0.70      0.82        10
  BenjaminKangLim       0.67      0.36      0.47        11
    BernardHickey       1.00      0.50      0.67        10
      BradDorfman       0.78      0.88      0.82         8
 DarrenSchuettler       1.00      0.89      0.94         9
      DavidLawder       1.00      0.60      0.75        10
    EdnaFernandes       0.73      0.89      0.80         9
      EricAuchard       0.73      0.89      0.80         9
   FumikoFujisaki       1.00      1.00      1.00        10
   GrahamEarnshaw       0.83      1.00      0.91        10
 HeatherScoffield       0.82      1.00      0.90         9
       JanLopatka       0.60      0.33      0.43         9
    JaneMacartney       0.36      0.50      0.42        10
     JimGilchrist       0.88      0.88      0.88         8
   JoWinterbottom       1.00      0.90      0.95        10
         JoeOrtiz       0.82      1.00      0.90         9
     JohnMastrini       0.79      0.88      0.83        17
     JonathanBirt       0.80      1.00      0.89         8
      KarlPenhaul       0.93      1.00      0.96        13
        KeithWeir       0.83      1.00      0.91        10
   KevinDrawbaugh       0.82      0.90      0.86        10
    KevinMorrison       0.60      1.00      0.75         3
    KirstinRidley       1.00      0.56      0.71         9
KouroshKarimkhany       0.88      0.88      0.88         8
        LydiaZajc       1.00      1.00      1.00        10
   LynneO'Donnell       0.82      0.82      0.82        11
  LynnleyBrowning       1.00      1.00      1.00        13
  MarcelMichelson       1.00      0.67      0.80        12
     MarkBendeich       0.73      1.00      0.85        11
       MartinWolk       0.83      1.00      0.91         5
     MatthewBunce       1.00      0.86      0.92        14
    MichaelConnor       1.00      0.85      0.92        13
       MureDickie       0.60      0.60      0.60        10
        NickLouth       0.90      0.90      0.90        10
  PatriciaCommins       1.00      1.00      1.00         9
    PeterHumphrey       0.57      0.89      0.70         9
       PierreTran       0.60      1.00      0.75         6
       RobinSidel       1.00      1.00      1.00        12
     RogerFillion       1.00      1.00      1.00         8
      SamuelPerry       0.76      0.93      0.84        14
     SarahDavison       1.00      0.71      0.83        14
      ScottHillis       0.67      0.44      0.53         9
      SimonCowell       1.00      0.90      0.95        10
         TanEeLyn       0.83      0.71      0.77         7
   TheresePoletti       0.90      0.82      0.86        11
       TimFarrand       0.92      0.85      0.88        13
       ToddNissen       0.90      1.00      0.95         9
     WilliamKazer       0.27      0.30      0.29        10

         accuracy                           0.83       500
        macro avg       0.84      0.83      0.82       500
     weighted avg       0.85      0.83      0.82       500

Unsupervised (to top)

Latent Dirichilet Allocation (LDA) (to top)

In [121]:
from sklearn.decomposition import LatentDirichletAllocation

Vectorizer (using countvectorizer for the sake of example)

In [122]:
vectorizer = CountVectorizer(strip_accents='unicode',
                             lowercase = True,
                            max_features = 1500,
                            stop_words='english', max_df=0.8)
tf_large = vectorizer.fit_transform(clean_paragraphs)

Run the LDA model

In [123]:
n_topics = 10
n_top_words = 25
In [124]:
lda = LatentDirichletAllocation(n_components=n_topics, max_iter=10,
                                learning_method='online',
                                n_jobs=-1)
lda_fitted = lda.fit_transform(tf_large)

Visualize top words

In [125]:
def save_top_words(model, feature_names, n_top_words):
    out_list = []
    for topic_idx, topic in enumerate(model.components_):
        out_list.append((topic_idx+1, " ".join([feature_names[i] for i in topic.argsort()[:-n_top_words - 1:-1]])))
    out_df = pd.DataFrame(out_list, columns=['topic_id', 'top_words'])
    return out_df
In [126]:
result_df = save_top_words(lda, vectorizer.get_feature_names(), n_top_words)
In [127]:
result_df
Out[127]:
topic_id top_words
0 1 company pound million share group business bil...
1 2 official south people north northern police ko...
2 3 china kong hong beijing chinese deng year part...
3 4 market percent price year china trade bank sta...
4 5 percent year million profit analyst quarter sh...
5 6 gold bre canada toronto stock busang canadian ...
6 7 ford union oil strike car plant new worker pro...
7 8 tonne year million crop new world export add a...
8 9 company service new network computer internet ...
9 10 bank financial fund nomura company loan firm s...

pyLDAvis (to top)

In [128]:
%matplotlib inline
import pyLDAvis
import pyLDAvis.sklearn
pyLDAvis.enable_notebook()
In [129]:
pyLDAvis.sklearn.prepare(lda, tf_large, vectorizer, n_jobs=-1)
Out[129]:

Warning: there is a small bug that when you show the pyLDAvis visualization it will hide some of the icons of JupyterLab

Neural Networks (to top)

Interested? Check out the Stanford course CS224n (Page)!