Contributor
Ananya

Neural Extraction Framework


Mentors
Tommaso Soru, Diego Moussallem, Ziwei XU
Organization
DBpedia
Technologies
python, sparql
Topics
machine learning, nlp, knowledge graphs, distant supervision
The Dbpedia property dbo:wikiPageWikiLink helps us to see for the articles which have the outlink to the Berlin_Wall article in their whole Wikipedia article i.e. the text of the Wikipedia article has the hyperlink present for the Berlin_Wall article. This forms an entity relationship between articles, where the Berlin_wall article is the subject entity and 299 articles are the base entities. However, it was found that only 9 out of these 299 articles had some other predicate relationship with the Berlin_Wall article. And the relationship of the other 290 articles is not clear. Currently, such relationships are extracted from tables and the infoboxes. Now for the rest 290 relationship extraction, we have to look out in the whole wikipedia article text (unstructured text) and find the relationship between the Berlin_Wall article and the rest of the 290 articles. The goal of this project is to develop a framework for predicate resolution of wiki links among entities. I will be using distant supervision for relation extraction.