Spoken language understanding using finite state tagger and long-range dependency parsing

Spoken language understanding is aimed at the interpretation of signs conveyed by a speech signal. While data-driven methods reduce maintenance and portability costs compared with handcrafted parsers, the collection of word-level semantic annotations for training remains a time-consuming task. This paper has focused on building generative model “Finite State Tagger” from unaligned data, using expectation-maximization techniques to align semantic concepts. Moreover, to model the hierarchical semantic relations in different slot entities, this paper proposed a pipeline architecture using finite state tagger and long-range dependency parsing.