Last week, we delved deep into the fascinating world of various machine learning models and how they are applied in solving real-world problems. We discussed supervised learning, both in terms of classification and regression, as well as unsupervised learning, with a focus on feature extraction and clustering. Now, as we move into the seventh week of our course, we will focus on a specific aspect of data processing practices – Natural Language Processing (NLP).

Natural Language Processing – NLP

Natural Language Processing is a branch of machine learning that grapples with understanding and interpreting human language. The biggest challenge for machines and models alike lies in making sense of text data, which needs to be converted into a numerical form for our models to process.

Key takeaway: Converting text data into numerical form is vital for models to process it.

Goals in NLP

The primary goals in NLP include classifying, predicting, translating, and converting text. All these tasks necessitate the vectorization of text, which involves transforming our text data into a format that machine learning models can understand and learn from.

Key takeaway: Text vectorization is crucial for achieving various NLP tasks.

Key Definitions

As we continue our journey into Natural Language Processing, it’s important to familiarize ourselves with some key terms:

Tokens: These are the building blocks of any language – words, punctuation, or any collection of one or more characters..

Vocabulary: This refers to all distinct words in your dataset. It’s important to note that we often trim the dataset, meaning we may not keep every single word.

Corpus: Simply put, a corpus is the dataset you are working with. It could be a collection of documents, quips, paragraphs, and so on.

N-gram: An n-gram refers to n consecutive tokens. For example, “I like pizza” is a tri-gram. Even “I lik pizz@” is considered a tri-gram as these are still tokens.

Stopwords: These are words like ‘the’, ‘as’, ‘and’, etc., that appear in all documents. They are usually not effective at identifying patterns and are often removed during preprocessing.

Meta Description: Explore the world of Natural Language Processing (NLP), understanding its goals, preprocessing steps, and key definitions in our comprehensive guide.

Preprocessing text data

Before we can feed text data into our models, it needs to be preprocessed. This involves fixing punctuation, removing stopwords, and applying techniques like lemmatization and stemming to convert different forms of a word into a single instance.

Key takeaway: Text preprocessing helps clean up and streamline data for more effective machine learning.

Luckily, there is lots of data…

The abundance of text data available across various platforms provides an excellent resource for training our models. However, the task of building datasets involves proper preprocessing and conversion of this data into a ‘vectorized representation’ or ‘tokenization’.

Key takeaway: Proper data preprocessing and tokenization are integral steps in building effective datasets for NLP.


Vectorization is the process of converting text into a mathematical form, where each entry in the vector represents a specific attribute or concept related to the text.

Key takeaway: Vectorization allows us to represent text in a mathematical form that models can understand and learn from.

The main idea

Once our tokens are vectorized, they are ready to be used for training our models. However, it’s important to remember that our tokens must be cleaned and preprocessed first, to avoid wasting memory and time on unnecessary elements like stopwords and punctuation.

Key takeaway: Cleaned and preprocessed tokens are the foundation of effective NLP model training.

Putting Theory into Practice

Finally, see Github for an example of building a classifier that can distinguish between Tweets related to natural disasters, and otherwise normal Tweets 




This week we’ve taken a closer look at Natural Language Processing, understanding its goals, the importance of preprocessing, and the concept of vectorization. As we move forward, remember that the key to successful NLP lies in effectively converting text data into a form that our models can understand and learn from.

In our next session, we will be focusing on evaluating model performances. We’ll learn how to measure the success of our machine learning models and understand where improvements can be made. Stay tuned for a deep dive into model performance metrics and evaluation techniques.