Introduction to the definition of Natural Language Processing and its various sub-topics with practical examples.

We will introduce a practical problem statement of building an NLP-powered application for a movie production house and look at various components of it.

We will learn new NLP techniques and build features to the Streamlit app as we go

We will collect our dataset of Hollywood movie plots using the BeautifulSoup library.

This will give you confidence that you don't need to always look for existing datasets and building datasets from scratch is not all that hard.

We will also look at various options to collect datasets even without code, with scraping tools like Parsehub, etc.

We will build intuition to calculate the importance of a given word in a given movie plot which would later be called TF-IDF.

We will convert movie plots into vectors through which we will extract important keywords in a movie plot, summarize the movie plot, and also find similar movie plots using the TF-IDF algorithm.

We will use a pre-trained paraphraser that generates multiple alternate paraphrase sentences to a given sentence.

The goal is to find the most diverse sentence to the original sentence.

We will get introduced to sentence transformers that can convert a word, sentence, or a full document into a vector.

We will create 2-dimensional vectors from movie plot vectors and visualize them using the Plotly library.

We will explore the power of visualizations by plotting some descriptions of directors & movie plots on the same visualization

We will introduce word vectors and develop a practical project to convert Hollywood movie scripts to a different local/region.

Many times popular Hollywood movies are adapted and remade in countries like China and India.

We will see how we can use named entity recognition and word vectors to localize the names and locations to a different country.

We will use sentence transformers to convert movie plots into vectors.

We will revisit the topics of keyword extraction and similar movie plot retrieval from TF-IDF lecture and perform the same with sentence transformers and see significant improvement in the results when compared to TF-IDF.

Given an open-ended short descriptive answer that students will provide to a given question, we will see how we can evaluate them using traditional word matching as well as semantic matching using sentence transformers.

We will use topic modeling to identify different topics (genres) from the movie plots using sentence transformers.

We will take any new movie plot and associate it with one of the topics we created using topic modeling.

We will get introduced to advanced language models like GPT-3 and understand how SaaS businesses are built with GPT-3.

We will deploy an API with GPT-3 and use it with our Streamlit app.