Named Entity Generation Using Sampling-based Structured Prediction

The problem of Named Entity Generation is expressed as a conditional probability model over a structured domain. By defining a factor-graph model over the mentions of a text, we obtain a compact parameterization of what is learned using the SampleRank algorithm.

• Type: Journal
• Year: 2010
• Where: INLG
• Authors: Guillaume Bouchard
• Keywords: Compact Parameterization, Conditional Probability Model, Entity Generation, INLG 2010, Natural Language Processing