Learning Procedures from Text: Codifying How-to Procedures in Deep Neural Networks

Abstract

A lot of knowledge about procedures and how-tos are described in text. Recently, extracting semantic relations from the procedural text has been actively explored. Prior work mostly has focused on finding relationships among verb-noun pairs or clustering of extracted pairs. In this paper, we investigate the problem of learning individual procedure-specific relationships (e.g. is method of, is alternative of, or is subtask of) among sentences. To identify the relationships, we propose an end-to-end neural network architecture, which can selectively learn important procedure-specific relationships. Using this approach, we could construct a how-to knowledge base from the largest procedure sharing-community, wiki-how.com. The evaluation of our approach shows that it outperforms the existing entity relationship extraction algorithms.