The Association for Computational Linguistics will be hosting their 50th Annual Meeting in Jeju Island, Korea July 8th – July 14th and we are pleased to share that VigLink’s Principal Data Scientist, Gabor Melli, will be presenting his recent paper entitled Identifying Untyped Relation Mentions in a Corpus Given an Ontology.
We are particularly excited because it is based on the research he has continued here at VigLink over the last nine months that has allowed us to continue to refine our link insertion solution and earn each of you more revenue each month.
Below is an overview of the paper his presentation will be based on.
The paper proposes a semi-supervised machine learning-based algorithm to the task of identifying relations between concepts that are stated in a document (such as a webpage). Knowing what is likely related to what in a document can for example help us at VigLink to know which product mentions refer to the same product (their coreference chains). Our approach is to leverage information found drawn from a well-organized database of concepts (an ontology) to heuristically create a labeled dataset (i.e. a semi-supervised approach) that can be used to train a statistical model. Finally, we demonstrated the effectiveness of the algorithm on a publicly available benchmark task.