Automatic Question Generation for HMM State Tying using a Feature Table

Generally, speech recognition systems use context-dependent subword units for recognition. To utilize these units, tree-based state tying is an unavoidable technique. This tying applies linguistic questions to group these units and ties them accordingly. These questions usually require an expensive resource-consumption manual procedure. Even though this problem has been addressed by data-driven automatic question generation systems, the quality of the systems greatly depends on the quality of the corpus. This paper presents a new approach to automatically assign questions using only a simple feature table which has three advantages. First, the requirement for expert knowledge is reduced to only a simple feature table instead of complex questions. Consequently, question extension or modification is effortless. Second, feature tables still preserve a benefit of tree-based state tying, i.e. incorporating linguistic knowledge into questions. Third, feature tables are commonly employed in phonetic studies for all languages. It is therefore a resource which is easy to access.