Adventures in Machine Learning

Unleashing the Power of NLTK: Techniques for Analyzing Unstructured Text Data

Introduction to NLTK

Language is a powerful communication tool that humans have used for centuries to express their thoughts, opinions, and emotions. In recent years, Natural Language Processing (NLP) has emerged as a field of study that aims to make it easier for humans and computers to interact with one another through language.

Python’s Natural Language Toolkit (NLTK) is a widely used open-source Python library that allows developers to process natural language text and analyze it using various techniques. In this article, we will explore the benefits of NLP using NLTK and the various techniques used in the preprocessing of unstructured data for analysis.

Preprocessing unstructured data for analysis

Data generated by humans is usually unstructured, making it difficult to analyze without preprocessing. The process of preprocessing involves cleaning, normalizing, and transforming unstructured data into a structured format that can be analyzed effectively.

Cleaning involves removing irrelevant information such as stop words, special characters, and numbers from the text. Normalizing involves converting data to a standard form by removing accents, expanding contractions, and converting text to lowercase.

Transforming involves breaking down the text into smaller units such as words, phrases, and sentences to make it easier to analyze.

Installing and setting up NLTK

NLTK can be installed using the Python package manager, pip. To install NLTK, open the terminal and type the following command:

pip install nltk

The next step is to download the necessary datasets and resources. To download all the resources, type the following command:

import nltk

The above code will open up a GUI to download the necessary resources like stop words, corpora, etc.


Tokenizing involves the process of breaking down the text into smaller units called tokens. Tokens are the basic building blocks of natural language processing and can be used to analyze the text at a granular level.

Tokenizing by word and by sentence

Tokenizing by word involves breaking down text into individual words, while tokenizing by sentence involves breaking down text into individual sentences. Let’s take a look at an example:

Text: “The quick brown fox jumped over the lazy dog.

The dog barked at the fox, but the fox kept running.”

Tokenizing by word:

[‘The’, ‘quick’, ‘brown’, ‘fox’, ‘jumped’, ‘over’, ‘the’, ‘lazy’, ‘dog’, ‘.’, ‘The’, ‘dog’, ‘barked’, ‘at’, ‘the’, ‘fox’, ‘,’, ‘but’, ‘the’, ‘fox’, ‘kept’, ‘running’, ‘.’]

Tokenizing by sentence:

[‘The quick brown fox jumped over the lazy dog.’, ‘The dog barked at the fox, but the fox kept running.’]

Importance of tokenizing for analysis

Tokenizing is an essential step in text analysis as it lays the foundation for other text analysis techniques. Texts, when tokenized into words or sentences, become much easier to compare, sort, and search.

By isolating keywords and important phrases through tokenization, researchers can analyze the sentiment, decipher the meaning of the text, and identify patterns or trends.


In conclusion, Natural Language Processing and NLTK can help you analyze and understand text data in a more effective and efficient way. By using the techniques discussed in this article, you can preprocess raw text data, tokenize it, and analyze it using various NLP techniques.

Armed with the right tools and techniques, you can extract valuable insights and make data-driven decisions based on your analysis of natural language text data.

Filtering Stop Words

Stop words refer to common words that are considered irrelevant in text analysis. These words, such as “the,” “and,” “in,” and “a,” occur frequently in text but do not add much value in terms of meaning.

Removing stop words is a common preprocessing step in natural language processing that helps to improve the effectiveness of text analysis.

Identifying and Removing Stop Words using NLTK

In NLTK, a stopword is defined as a word that is considered to be irrelevant in text analysis and can be filtered out. NLTK has a list of pre-defined stop words that can be filtered out.

To remove stop words in NLTK, you need to first import the list of stop words from the corpus module:

from nltk.corpus import stopwords

stop_words = set(stopwords.words(‘english’))

Next, you can use the stop_words set to remove the stop words from the text:

text = “The quick brown fox jumps over the lazy dog”

tokens = text.split()

filtered_tokens = [word for word in tokens if not word in stop_words]

filtered_text = ” “.join(filtered_tokens)



quick brown fox jumps lazy dog


Stemming refers to the process of reducing words to their base or root form. The purpose of stemming is to convert words with different inflections into a common base form to simplify analysis.

For example, “jump,” “jumps,” “jumping,” and “jumped” can be reduced to their common base form “jump.”

Porter Stemmer and its Limitations

The Porter stemming algorithm is a popular stemming algorithm used in natural language processing. It uses a set of rules to reduce words to their base form.

While the Porter stemmer is widely used, it has its limitations. Firstly, the Porter stemmer algorithm can be too aggressive in its approach in reducing words to their base form.

This can result in over-stemming, where the stemmer reduces words too aggressively, resulting in the loss of important information. Secondly, the Porter stemmer algorithm is not effective for reducing irregular inflections that do not follow common English grammatical rules.

Comparison with


Lemmatizing, like stemming, aims to reduce words to their base form. However, the key difference between stemming and lemmatizing is that lemmatizing takes into account the context of the word to determine its root form.

Stemming, on the other hand, blindly applies a set of rules that are not context-dependent. For example, consider the following sentence:

“The mice were running in a field.”

The stem “run” from “running” and “mice” is “mic” using the Porter stemming algorithm.

In contrast, the lemma of “running” is “run” and the lemma of “mice” is “mouse,” reflecting the actual meanings of the words. In conclusion, NLTK provides powerful tools for preprocessing and analyzing unstructured text data.

Stop words can be filtered out, and stemming or lemmatizing can be applied to reduce words to their base form to simplify analysis. Careful consideration should be given to choosing the right technique for each specific task, as each technique has its strengths and limitations.

Tagging Parts of Speech

In natural language processing, parts of speech (POS) refer to the grammatical roles played by words in a sentence. Different parts of speech include nouns, verbs, adjectives, adverbs, prepositions, and conjunctions.

Identifying the part of speech of words in a sentence can be helpful in understanding the meaning of the text and carrying out advanced analysis.

NLTKs POS tagging and its accuracy

NLTK provides different ways of POS tagging, including the use of pre-trained models. The pre-trained models can be used to tag parts of speech in different languages.

The accuracy of POS tagging varies depending on the training data, the tagging algorithm, and the language being analyzed. In the English language, the accuracy of POS tagging can be as high as 95%, depending on the algorithm used.

Example of POS tagging in Real and Nonsensical Text

POS tagging can be used to analyze real and nonsensical text. In the example below, NLTK’s POS tagging is applied to real text and non-sensical text:

Real Text:

“John gave Mary a lovely bouquet of flowers for her birthday.”

[(‘John’, ‘NNP’), (‘gave’, ‘VBD’), (‘Mary’, ‘NNP’), (‘a’, ‘DT’), (‘lovely’, ‘JJ’), (‘bouquet’, ‘NN’), (‘of’, ‘IN’), (‘flowers’, ‘NNS’), (‘for’, ‘IN’), (‘her’, ‘PRP$’), (‘birthday’, ‘NN’), (‘.’, ‘.’)]

Nonsensical Text:

“The frabjous whiffling snicker-snackled the mimsy borogoves.”

[(‘The’, ‘DT’), (‘frabjous’, ‘JJ’), (‘whiffling’, ‘VBG’), (‘snicker-snackled’, ‘VBN’), (‘the’, ‘DT’), (‘mimsy’, ‘NN’), (‘borogoves’, ‘VBZ’), (‘.’, ‘.’)]

In the real text, NLTK’s POS tagging accurately identified the parts of speech of each word, including the proper nouns, nouns, verbs, adjectives, adverbs, prepositions, and conjunctions.

In the nonsensical text, NLTK’s POS tagging assigned parts of speech to each word, despite the words having no meaningful context.


Lemmatizing involves the process of reducing words to their base form while still maintaining the context through the use of vocabulary and morphological analysis. Unlike stemming, which focuses on the reduction of words to their base form by getting rid of affixes, lemmatizing takes into account the context and the meaning of the word.

Comparison with Stemming

While stemming aims to reduce words to their base form by removing affixes, lemmatizing takes into account the context of the word in the sentence. For example, in the following sentence:

“I saw a 10-foot tall giraffe.”

Stemming would reduce “foot” to “foot” and “tall” to “tall,” resulting in the nonsensical sentence “I saw a 10 foot tal giraff.” In contrast, lemmatizing would convert “foot” to “feet” and “tall” to “tall,” resulting in the correct sentence “I saw a 10 feet tall giraffe.”

Importance of Obtaining Complete Words for Analysis

Obtaining complete words through lemmatizing is essential in natural language processing as it helps to maintain the integrity and context of the text. This is especially important when analyzing sentences with complex grammatical structure or in cases where the meaning of the text is dependent on the correct understanding of affixes and their meaning.

In conclusion, POS tagging and lemmatizing are essential techniques in natural language processing that help to simplify text analysis by reducing words to their base form and identifying the part of speech of each word in a sentence. Careful consideration should be given to choosing the right technique for each specific task, as each technique has its strengths and limitations.

Other Text Analysis and Visualization Techniques

In addition to pre-processing techniques such as tokenization, stemming, and lemmatization, NLTK offers several other text analysis and visualization techniques that allow for a more in-depth exploration of natural language data.

Concordance and Dispersion Plotting

The concordance function in NLTK displays the occurrence of a word in its context. The output presents each occurrence of the word in its respective line with some words of context around it.

With the concordance function, researchers can identify specific instances in which a particular word appears and analyze the surrounding context of the word. Dispersion plots, on the other hand, show the distribution of when specific words occur over the length of the text.

This technique can help to identify patterns in word usage and examine how frequently certain words appear in relation to others.

Frequency Distribution and Collocations

Frequency distribution is a useful technique for identifying frequently occurring words in a text. The NLTK provides a function for generating a frequency distribution of tokens.

A frequency distribution lists down how many times each token appears in the text. With this technique, analysts can identify popular keywords and their frequency to help them make informed decisions based on the text being analyzed.

Collocations, a method that’s often used in textual analysis, is a technique developed specifically to identify common phrases or word combinations which appear more often than would be expected by chance. Collocations provide an additional level of insight into the structure and usage of the language being analyzed.

Named Entity Recognition and visualization using NumPy and Matplotlib

Named Entity Recognition (NER) is the process of extracting named entities, such as people, locations, and organizations, from a text. NER uses algorithms to recognize and classify named entities in a text and can be useful in applications like information extraction and summarization.

After identifying named entities, researchers can use libraries such as NumPy and Matplotlib to visualize the results. For example, a word cloud can be generated with named entities, with the size of each named entity reflecting its frequency in the text.

Recap of Techniques

In conclusion, NLTK provides several techniques for natural language processing, including pre-processing techniques, POS tagging, and other text analysis and visualization techniques. Other techniques include concordance and dispersion plotting, frequency distribution, collocations, named entity recognition, and visualization techniques using NumPy and Matplotlib.

Researchers can use these techniques to process and analyze unstructured text data and gain valuable insights from it. In conclusion, NLTK is a powerful natural language processing tool that offers a range of techniques for analyzing and understanding unstructured text data.

Pre-processing techniques like tokenization, stemming, and lemmatization can make data more amenable to analysis. POS tagging, named entity recognition, and concordance and dispersion plotting can help identify parts of speech and named entities, and understand the contextual relationships between words.

Additionally, techniques like frequency distribution and collocation can help identify patterns of word usage. Visualization methods such as wordclouds and dispersion plots can bring data to life in an easily interpretable way.

Being able to analyze text data with these techniques can quickly reveal valuable insights that could otherwise be missed.