Following four annual C@merata evaluations at MediaEval, in which natural language phrases about music must be mapped to passages in MusicXML scores, we have developed a representation which captures the meaning of a text about music as a JSON feature structure. We have written a system to convert any text to this repres entation, comprising 23 stages of pre-processing, followed by application of the SpaCy statistical dependency parser. We have applied the system to the C@merata test queries with accurate results. Our approach can capture extremely detailed information, as well as general musical ideas.
|Number of pages||2|
|Publication status||Published - 2017|
|Event||ISMIR 2017: 18th International Society for Music Information Retrieval Conferences - Suzhou, China|
Duration: 23 Oct 2017 → 27 Oct 2017
Conference number: 18
|Period||23/10/17 → 27/10/17|