Assignment 4: News Headlines classification using Naive Bayes classifier

Problem statement: Given the headline of news, the objective is to find the category of the news. (Note: Use only headline as input to find the category ) For example:

Short description:

Headline: Why Keeping a Food Journal Is Better Than Going on a Diet

Date:

Link:

Authors:

Category: HEALTHY LIVING

Consider the following categories only: Business, Comedy, Sports, Crime, Religion, Healthy Living, Politics

2. Dataset: news_category_dataset.json
3. Classification Algorithm:

Naive Bayes

4. Features:

Train the classifier using the following features.

1. Bag-of-words
2. TF-IDF
3. Create your own custom feature vectors.

For example, feature vector can contains following features:

1. Current word (Unigram)
2. POS tag of current word
3. Position of the word
4. Length of the news instance

Here, a total of 3 models needs to be trained i.e., one model using

Bag-of-words features, one model using TF-IDF features and one model using Custom feature vectors.

For more information on feature selection, refer the following paper:

Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan: Thumbs up? Sentiment Classification using Machine Learning Techniques. In Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP 2002).

5. Evaluation:

Perform 3-fold cross-validation for each model and report

1. Overall precision, recall and F1-score
2. Category-wise precision, recall and F1-score

Implementation notes:

1. You can use existing libraries to implement Naive Bayes algorithm and other tasks such as feature extraction, POS tagging, cross-validation and calculation of metrics.