Mining Numbers in Text for Various Kinds of Text Data

2014 Fiscal Year Final Research Report

• Project/Area Number: 24500162
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent informatics
• Research Institution: The University of Tokushima (2013-2014); The University of Tokyo (2012)
• Principal Investigator: YOSHIDA Minoru 徳島大学, ソシオテクノサイエンス研究部, 講師 (40361688)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: 数値情報抽出 / レイアウト解析

Outline of Final Research Achievements

We studied a method for extracting contexts (i.e., attributes or topics) of numbers written in text. Our goal is to develop a system that accept numbers as queries and returns appropriate data from the various kinds of text data such as Wikipedia, Twitter, etc. To achieve this goal, we proposed a method for extracting numbers and their contexts applicable both to unstructured texts (e.g., sentences) and semi-structured texts (e.g., tables). Our method uses unsupervised learning algorithms based on probabilistic generative models for texts to extract attributes and hierarchical topics from Web documents. We also proposed a method to extract corpus-specific number expressions from any kind of text data. For number expressions, we found a coding scheme that can be used both for indexing and probabilistic generative models.

Free Research Field

テキストマイニング