Comparing News Articles and Stock Prices
An imaginary search scenario where news articles are compared with historical stock prices for given stocks (around one hundred of them).
The stock's percent change is calculated by the stock's opening price prior to the article's publishing date and the next opening price.
Documents are searched with verb words belonging to "increase" (or similar terms) WordNet verb ontology in intransitive context
along with their subject complements.
Nouns (noun headwords in subjects) are grouped depending on their category in the WordNet ontology.
Similar ideas have been presented in research papers like AZFinText, which is more about using breaking financial news to predict stock prices.
This sample employs components such as a natural language parser and a semantic parser
that serve as core functions in the information extraction task.
The articles and the stock quotes are accessed through Yahoo Finance RSS and stock quotes API.
The test data set contains a small number of articles read during a one month period, mostly in energy/renewables sector.
appreciated strongly, evolving, expand rapidly, flourished, gains, go up, grew by 11%, increase significantly, jumped by 21.5%, magnified, prospering, soared...
Using WordNet's lexicographer files categories as a top-level ontology, for example:
N6 - nouns denoting man-made objects
N18 - nouns denoting people
Australian hotels expand rapidly to catch Chinese tourism boom
holidaymakers from China are increasing at double the rate of the tourism sector
N21 - nouns denoting possession and transfer of possession
Overall, earnings in both the utilities and transport segments grew by a low single digit growth rate, while energy earnings jumped 75.7%.