Sentiment Analysis & Text Generation

Animation by [3]

Sentiment analysis on Sarcasm dataset and text generation on William Shakespere sonnets have been implemened in this repo. Sentiment analysis is the process of analyzing digital text to determine if the emotional tone of the message is positive, negative, or neutral, however, in our case we are dealing with sarcastic and non sarcastic headlines from the news articles. Today, companies have large volumes of text data like emails, customer support chat transcripts, social media comments, and reviews. Sentiment analysis tools can scan this text to automatically determine the author’s attitude towards a topic. Companies use the insights from sentiment analysis to improve customer service and increase brand reputation. [1]

Text generation is a process where an AI system produces written content, imitating human language patterns and styles. The process involves generating coherent and meaningful text that resembles natural human communication. Text generation has gained significant importance in various fields, including natural language processing, content creation, customer service, and coding assistance. [2], in this project, I am using LSTM based model to generate text based on William Shakespere's sonnet dataset.

sentiment_analysis_and_text_generation 
├── __init__.py  
├── data
│   └── Sonnet.txt  
├── images 
│   └── ...
├── README.md  
├── requirements.txt  
├── sentiment_analysis
│   └── sentiment_analysis.ipynb 
├── text_generation
│   └── text_generation.ipynb
└── weights
    ├── meta.tsv  
    └── vecs.tsv

Overall structure of the repository.

Sentiment Analysis

The section involves downloading the Sarcasm dataset from the internet, preporcesing, training, inferencing:

preprocessing

1. Lower casing
2. Removing HTML tags and URLs from the text
3. Stop words removal
4. Lemmization
5. Dataset split
6. Tokenization
7. Sequence Padding

After preprocessing we have set some hyperparameters for training and test, those parameters will influcnece the performance metrics.

	General	Model
	Hyperparameters value Vocabulary size 5000 embedding dimension 100 maximum sequence length 200 truncation type post padding type post out of vocab token oov Training set size 80%	Hyperparameters value activation relu learning rate adaptive maximum epochs 30 loss binary cross entropy optimizer Adam metric Accuracy Dropout 0.3

Fig 1: LSTM based model for sentiment analysis

Installation

 git clone https://github.com/faizan1234567/sentiment_analysis_and_text_generation
 cd sentiment_analysis_and_text_generation

Create and activate Anaconda Environment

conda create -n nlp python=3.9.0
conda activate nlp

Now install all the required dependencies

pip install --upgrade pip
pip install -r requirements.txt

Installation done !

Text Generation

In the text generation step the william shakespere sonnet dataset has been used. The following steps covered.

1. Dataset loading
2. Lowercasing
3. Inspecting dataset
4. Tokenization
5. Padding to have uniform length sequences

Following the preprocessing, the model is define and trained with the following hypermeters:

	General	Model
	Hyperparameters value embedding dimension total_words + 1 maximum sequence length 11 truncation type pre padding type pre out of vocab token oov	Hyperparameters value activation relu maximum epochs 150 loss categorical cross entropy optimizer Adam Dropout 0.2

Acknowledgements

[1]. Amazon Web servies. (1978). What is ... Amazon. https://aws.amazon.com/what-is/sentiment-analysis/

[2]. Data camp. (1978). Text generation. Amazon. https://www.datacamp.com/blog/what-is-text-generation

[3]. C. (2022, January 5). Introduction to Natural Language Processing(NLP) — Part 1. Medium. https://medium.com/@chinmaychetan04/ introduction-to-natural-language-processing-nlp-part-1-2f2dfbc305d6