A scalable machine-learning approach for semi-structured named entity recognition

• Authors: Utku Irmak;Reiner Kraft
• Venue: Proceedings of the 19th international conference on World wide web
• Year: 2010

Named entity recognition studies the problem of locating and classifying parts of free text into a set of predefined categories. Although extensive research has focused on the detection of person, location and organization entities, there are many other entities of interest, including phone numbers, dates, times and currencies (to name a few examples). We refer to these types of entities as "semi-structured named entities", since they usually follow certain syntactic formats according to some conventions, although their structure is typically not well-defined. Regular expression solutions require significant amount of manual effort and supervised machine learning approaches rely on large sets of labeled training data. Therefore, these approaches do not scale when we need to support many semi-structured entity types in many languages and regions. In this paper, we study this problem and propose a novel three-level bootstrapping framework for the detection of semi-structured entities. We describe the proposed techniques for phone, date and time entities, and perform extensive evaluations on English, German, Polish, Swedish and Turkish documents. Despite the minimal input from the user, our approach can achieve 95% precision and 84% recall for phone entities, and 94% precision and 81% recall for date and time entities, on average. We also discuss implementation details and report run time performance results, which show significant improvements over regular expression based solutions.