Multi-level Query Analysis for NLT-based Synthesizer Interface

A synthesizer is an electronic musical instrument, which can produce a variety of tones, but requires long-term training and highly skilled knowledge to obtain the desired tone. Therefore we developed an interface for software synthesizers utilizing natural language processing so that even beginners can use it easily. This paper describes a new method to select a timbre from a database of presets by using word-level information from a sentence, and an additional function that applies a sound effect to the produced tone using a sentence-level information. We have been researching ways to use natural language processing for synthesizers to select timbre from words and phrases input by users. In this research, by analyzing input sentences at multi-level basis, it becomes possible to call out timbres with more variety than ever. In addition to calling timbre at the word level, the system performs language analysis at the sentence level and uses the results to apply certain sound effects to the synthesizer sounds. We plan to explore more appropriate relationships between sentences and effects to provide easy-to-use synthesizer interface.