Repository for "On Extracting Overlapping Named Entities from E-Commerce Product Reviews"

E-commerces such as Amazon, E-bay and many others receive from consumers thousands of product reviews every day with rich fine-grained information that, without the aid of Opinion Mining, would be impossible to make sense of. Common tasks in this domain, such as Aspect-Based Sentiment Analysis, Taxonomy Learning and Entity Linking, depend on extracting relevant entities from text with considerable performance. In this paper, we propose a set of entities that provides specific structuring for product reviews in order to enable richer insights for this domain. We also provide a new dataset on laptop reviews annotated with overlapping entities that is used to experiment with state-of-the-art named entity recognition models. Lastly, we perform experiments to understand if this approach generalizes well across different product categories.

Obtained through 5 x 2-fold cross validation. Download Trained Model

TBA