Browse Prior Art Database

System and Method to predict financial instrument outcomes based on named events from news articles

IP.com Disclosure Number: IPCOM000247771D
Publication Date: 2016-Oct-06
Document File: 2 page(s) / 28K

Publishing Venue

The IP.com Prior Art Database

Abstract

Semi supervised system to detect domain specific events (litigation, acquisitions etc.) and predict financial metrics (stock price) based on these events for a given timerange.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 2

System and Method to predict financial instrument outcomes based on named events from news articles

Disclosed is a semi supervised system to automatically detect domain specific events (litigation, acquisitions etc.) by

Expanding financial metrics


Event detection by learning event-word distributions
Studying correlation of automatically extracted events from news with numerical aspects of entities (banks)

and predict metric (stock prices) values based on these events for given timerange.

Differences from Prior Work : ~~~~~~~~~~~~~~~~~~~~~~

             Our system is different from previous works in the field because of the following reasons

It works on openly available new articles from any source, without any external input


1.

like external agency's ratings on volatility to finance instrument but still identify domain specific events in news articles and quantify its impact.

Different mentions of the same event ("Apple wins Lawsuit", "Apple beats Samsung")


2.

are combined into a single coherent event. Since the system works at abstract level and learn the language model for each event type by its word distribution, It is able to capture events equally well.

The system learns the event importance from its language model and does not


3.

depend on mention count across articles.

The solution is a two step process

Identify Named Events from given corpus


1.

Identify correlation coefficient between Events and aspects for the entity.


2.


1. Identify Named Events steps: ~~~~~~~~~~~~~~~~~~~~~~~~~~


• From Subject Matter Experts (SME's) , get initial event related words (seed/event words)

• Expand event words using word2vector model from the news corpus (Paid News Corpu...