Difference between revisions of "User:Paal-Christian Salvesen Njoelstad"
m (Creating user page with biography of new user.) |
(No difference)
|
Latest revision as of 21:33, 19 September 2013
Personal biography: Lars Smørås Høysæter and Pål-Christian Salvesen Njølstad are both fifth-year MSc candidates in Computer Science at NTNU. They are currently working on a specialization project under the supervision of Jon Atle Gulla as a part of the SmartMedia project (http://smartmedia.idi.ntnu.no/). The task at hand is to process Norwegian news articles and determine an overall sentiment value per article. The system for processing articles and the performance of entity tagging is already in place (i.e. we know which persons / organizations are mentioned in the article, what the title is, what composes the introduction and main text). Our approach after having conducted a basic study of relevant literature is to consider each sentence in each article separately and determine 1) if it is subjective and 2) the sentiment of the sentence (positive, negative and magnitude - for instance on a [-1,1] scale) (2) contingent of the presence of subjectivity). After this has been done for all sentences we could average the sentiment of all subjective sentences (and maybe do some weighting - for instance with the title and introduction having a greater impact towards final sentiment value) to arrive at an aggregate sentiment value for the article.
Our approach for arriving at the sentence level subjectivity {0,1} and sentiment value [-1,1] is to rely on linguistics to help structure the sentences in a way suited for a semi-supervised learning algorithm to be able to arrive at rules for determining the sentence-level sentiment. For instance, we would need to be able to tag words according to class, maintain lists of 'positive' and 'negative' adjectives and adverbs, keep lists of 'subjective' verbs in addition to track conjunctions and negations. There are probably many other attributes of natural language that we need to account for in order to be able to (correctly) identify sentiment and we hope someone with a background in or familiarity with linguistics would be willing to help us with this and/or direct us to relevant literature.
About Pål: Pål has in parallel to his studies at NTNU completed a BSc in Economics at the Norwegian School of Economics (NHH). He has also had three internships in finance and related fields (investment bank ABG Sundal Collier and consultancy McKinsey & Company). Pål spent his third year at NTNU at Harvard University.
About Lars: Lars has experience with programming from the MTDT study program and the Artificial Intelligence specialization. In addition to this he is part of a startup company (Dogu AS) that delivers several software solutions and he has acquired a lot of experience being a part of this. (www.dogu.no).