News analysis can be used as part of trading models to characterize firm behaviors over time and thus yield important strategic insights about rival firms.

A large number of companies use news analysis to help them make better business decisions. If provided a set of metrics such as sentiment and relevance as well as the frequency of news arrivals, it is possible to construct news sentiment scores for multiple asset classes from equities, fixed income and derivatives.

In practice, sentiment scores can be constructed at various horizons to meet the different objectives of quantitative trading strategies. Scores are generally constructed as a range of values which may range between 0 and 100, where values above and below 50 convey positive and negative sentiment, respectively. Based on such sentiment scores, it should be possible to generate a set of strategies useful for asset allocation, hedging, and order execution.

How it works

  1. User picks a set of companies of interest for comparison (browser)
  2. Python submits company cohort and appropriate query characteristics to AlchemyData
  3. Previously, AlchemyAPI will have parsed news articles for sentiment and named-entity recognition
  4. AlchemyData receives a query specifying characteristics for news articles appropriate to the given user returns source articles with requested data for analysis
  5. Python algorithm summarizes returned findings and passes to browser via JSON
  6. Boostrap renders findings (HTML) for the user

Challenges I ran into

Debugging proved the biggest challenge. The AlchemyData query api is a bit cryptic, but astonishingly functional.

Accomplishments that I'm proud of

We are very excited about what is possible with the AlchemyData news service. It was remarkably easy to do some very complex analyses of an enormous dataset, and we didn't even need a database.

What I learned

Python web dev, from scratch.

What's next for WhatsonNews

  1. A name that does not infringe on any IBM trademarks.
  2. Analyses of the effect of sentiment on stocks - using machine learning (perhaps Watson tradeoff analysis) to forecast resulting share price adjustments.

Built With

Share this project: