How sarinasharif21/financial-news-ai works
This repository is a command-line financial-news analysis application. A user enters an article; the runtime preprocesses it for Reuters classification, then checks the predicted topic. Only selected financially important categories trigger IMDB-tokenized sentiment analysis and a Buy/Sell label; other categories produce an Indirect result. The graph distinguishes those domain stages from model and tokenizer assets. Training capabilities are documented in notebooks, but the only indexed runtime source is src/predict.py, so unsampled training details and links are not expanded.
Main entry points: predict.py CLI entry · sentiment_model.h5 Sentiment classifier · reuters_category_model.h5 Category classifier · tokenizer.pickle Sentiment tokenizer
Components 10
News analysis
- CLI entry
src/predict.py - Reuters preprocessing
src/predict.py - Category classifier
models/reuters_category_model.h5 - Topic relevance gate
src/predict.py - Sentiment preprocessing
src/predict.py - Sentiment classifier
models/sentiment_model.h5 - Suggestion decision
src/predict.py
Model assets
- Sentiment tokenizer
models/tokenizer.pickle - Reuters word index
Other components
- News user
Connections 11
- News user → CLI entry: enters article
README.md - CLI entry → Reuters preprocessing: preprocesses text
src/predict.py - Reuters preprocessing → Category classifier: classifies sequence
src/predict.py - Reuters preprocessing → Reuters word index: looks up words
src/predict.py - CLI entry → Topic relevance gate: checks category
src/predict.py - Topic relevance gate → Sentiment preprocessing: routes relevant text
README.md - Sentiment preprocessing → Sentiment tokenizer: tokenizes text
src/predict.py - Sentiment preprocessing → Sentiment classifier: prepares sequence
src/predict.py - Topic relevance gate → Suggestion decision: sets indirect result
src/predict.py - Sentiment classifier → Suggestion decision: provides sentiment
src/predict.py - CLI entry → News user: prints analysis
src/predict.py