Searching for economic effects of user specified events based on topic modelling and event reference
LE3 .A278 2015
Master of Science
Finding the relation between temporal events and economics is of crucial importance to businesses and investors. This thesis proposes a systematic solution in order to retrieve news articles of events as well as nding the economic e ects of these events. The system searches for news articles from a large collection of documents based on user queries. An approach combining probabilistic topic models and statistical language model is designed to retrieve the articles that the user speci ed. In this way, the developed model reaches a balance between generality and speci city that improves the retrieval performance to given user query. In order to nd the economic articles related to an user speci ed event, important features are extracted from the event. Term frequency inverse document frequency (TFIDF) and information gain theory are adopted for feature extraction. A new method called normalized feature referencing is introduced to reference from the event to the related economic articles. The experiment results show that the proposed method has su cient capability for event reference.
The author retains copyright in this thesis. Any substantial copying or any other actions that exceed fair dealing or other exceptions in the Copyright Act require the permission of the author.