CompletedAI / ML
Personalized News Recommendation System
Semantic news discovery shaped around the reader
The system transforms article text into sentence embeddings, organizes the corpus into topic clusters, and compares content against mood and interest profiles to produce personalized recommendations.
Project field / personalized-news
Context
The problem space.
A general news feed treats every reader the same. This project investigates how semantic representation and lightweight preference profiles can make discovery more relevant.
Confirmed work
What the evidence supports.
- Articles were collected from Indian Express pages with titles, descriptions and metadata.
- Sentence Transformers convert article text into dense semantic embeddings.
- K-Means groups articles into topic regions before recommendation.
- User interest and mood profiles are compared to content with cosine similarity.
PythonNLPSentence TransformersWeb ScrapingMachine LearningK-MeansCosine Similarity
System flow
- 01
Indian Express articles
- 02
Titles, descriptions & metadata
- 03
Sentence embeddings
- 04
K-Means topic clustering
- 05
Interest & mood profiles
- 06
Cosine similarity
- 07
Recommendations