UCLmogs

UCLmogs presents Yield, a tool that helps you navigate relevant financial news and turns it into clear portfolio signals, showing what impacts your holdings and how to adjust position sizes.

Comment

Inspiration

Financial news is messy and hard to act on. We wanted to turn raw articles into clear, actionable insights for your actual portfolio.

What it does

Yield finds relevant financial news and converts it into portfolio-specific signals, showing which holdings are impacted and how to adjust position sizes.

How we built it

We built a FastAPI backend connecting a Chrome extension and dashboard. Articles are processed using NLP: sentence-level sentiment (FinBERT), semantic embeddings, and theme matching. These are aggregated into stock-level signals with a confidence score, which feeds into a Kelly-style position sizing framework.

Challenges we ran into

Mapping language sentiment to real market impact was difficult, as financial text is often ambiguous. We also had to reduce noise from weak or conflicting signals while keeping the system responsive. It was also important to define clear and meaningful themes to ensure signals were interpretable and mapped correctly to stock exposures.

Accomplishments that we're proud of

We built an end-to-end system that goes beyond basic sentiment analysis by combining semantic understanding, theme-based reasoning, and portfolio-level impact with position sizing.

What we learned

We learned that sentiment alone isn’t enough to extract meaningful financial signals, context, relevance, and consistency are crucial. We also saw the importance of quantifying uncertainty, as noisy or conflicting information can easily lead to misleading conclusions without a proper confidence framework.

What's next for Yield

Next, we want to build a fully networked version of Yield with broader instrument coverage and live financial data integration. This would let us combine news sentiment with real-time market signals such as price and volume changes, helping us make the system more responsive and better grounded in actual market behaviour.

Built With

Share this project:

Updates