MINIPROJECT

INTRODUCTION
In the digital era, social media platforms like Twitter serve as influential tools for
communication and trend analysis. Twitter's real-time, concise format makes it ideal for
sharing views on diverse topics, including data science—a growing interdisciplinary field. As
discussions around data science increase, tweets offer valuable unstructured data reflecting
public sentiment. This project analyzes over 240,000 data science-related tweets using Natural
Language Processing (NLP), aiming to classify them as positive, negative, or neutral. The
VADER sentiment analysis tool, tailored for social media language, was used due to its ability
to interpret informal text, slang, and emojis effectively.

OBJECTIVES
The following are the objectives of this project, each aimed at systematically analyzing
sentiment from a large corpus of tweets related to data science. The project leverages
natural language processing techniques to uncover insights from unstructured social media
data:
1. Analyze sentiment in 240,000+ data science-related tweets using NLP.
2. Use VADER for rule-based sentiment classification.
3. Categorize tweets as Positive, Negative, or Neutral.
4. Implement the system using Python and NLTK libraries.
5. Gain practical experience in text analytics and social media monitoring.
6. Visualize results using bar charts for clear interpretation.

REQUIREMENTS
Hardware Requirements:

• A computer or laptop with at least an Intel i5 or Ryzen 5 processor.

• Minimum 8 GB of RAM recommended due to large dataset (240K+ tweets).

• At least 100 GB of free storage space, preferably on an SSD.

• Stable internet connection if you're working with clusters or downloading datasets.

ALGORITHM USED
This project uses Natural Language Processing (NLP) and rule-based sentiment analysis to
classify 240,000+ data science-related tweets. Given the informal, slang-heavy nature of
tweets, a lightweight approach was chosen to ensure efficiency without compromising
accuracy.

1
1. VADER Sentiment Analysis:
VADER, a rule-based tool designed for social media text, calculates sentiment using lexical
rules and assigns four scores: positive, negative, neutral, and compound (ranging from -1 to
+1).

• Classification logic:
o Positive: compound ≥ 0.05
o Negative: compound ≤ -0.05
o Neutral: between -0.05 and 0.05
Its ability to interpret emphasis, punctuation, and emojis makes it ideal for
Twitter data.

2. Text Preprocessing:
Tweets were cleaned to remove noise using the following steps:

• Lowercasing
• Removing URLs, special characters, digits, and stopwords
• Stripping extra whitespace

3. Natural Language Toolkit (NLTK):

4. Data Visualization:
Bar charts were used to visualize the sentiment distribution (positive, negative, neutral),
providing an intuitive overview of public opinion.

By combining rule-based analysis with effective preprocessing and visualization, the project
achieved efficient, accurate sentiment classification of large-scale Twitter data.

DATA ANALYTICS LIFE CYCLE STEPS

1. Discovery:
Defined the goal of analyzing public sentiment using a dataset of tweets
(data_science.csv), focusing on tweet text.
2. Data Preparation:
Cleaned tweets using Pandas—converted to lowercase, removed URLs, mentions,
hashtags, digits, and punctuation. A cleaned_text column was created.
3. Model Planning:
VADER was selected as the sentiment analyzer for its suitability to social media
content and ease of use.
4. Model Building:
Applied VADER’s SentimentIntensityAnalyzer to classify tweets as Positive,
Negative, or Neutral based on the compound score.
5. Communicate Results:
Used bar charts to visualize sentiment distribution and highlight public opinion on
data science.

2
 6. Operationalize:
Stored processed results in memory for evaluation. Though academic, the workflow
supports future reuse or extension.

IMPLEMENTATION

3
4
5
6
CONCLUSION

Through this project, we successfully classified data science-related tweets into positive,
negative, and neutral sentiments using the VADER sentiment analyzer. The analysis revealed
that most tweets had a neutral tone, while positive sentiments were more common than negative
ones.
The approach required no manual labeling and worked efficiently across a large dataset. This
made VADER a practical and reliable tool for large-scale sentiment analysis.
Overall, the project demonstrates how automated sentiment analysis can be used to interpret
public opinion in real time. These insights can benefit companies, educators, and researchers
by identifying trends and understanding user engagement with data science topics.