Towards Efficient and Effective Semantic Table Interpretation
published: Dec. 19, 2014, recorded: October 2014, views: 22
Report a problem or upload filesIf you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.
This paper describes TableMiner, the ﬁrst semantic Table Interpretation method that adopts an incremental, mutually recursive and bootstrapping learning approach seeded by automatically selected ‘partial’ data from a table. TableMiner labels columns containing name dentity mentions with semantic concepts that best describe data in columns, and disambiguates entity content cells in these columns. TableMiner is able to use various types of contextual information outside tables for Table Interpretation, including semantic markups (e.g., RDFa/microdata annotations) that to the best of our knowledge, have never been used in Natural Language Processing tasks. Evaluation on two datasets shows that compared to two baselines, TableMiner consistently obtains the best performance. In the classiﬁcation task,it achieves signiﬁcant improvements of between 0.08 and 0.38 F1 depending on different baseline methods; in the disambiguation task, it outperforms both baselines by between 0.19 and 0.37 in Precision on one dataset, and between 0.02 and 0.03 F1 on the other dataset. Observation also shows that the bootstrapping learning approach adopted by TableMiner can potentially deliver computational savings of between 24 and 60% against classic methodsthat‘exhaustively’processestheentiretablecontenttobuildfeaturesfor interpretation.
Link this pageWould you like to put a link to this lecture on your homepage?
Go ahead! Copy the HTML snippet !