An Automaton-Based Filtering System for Streaming MusicXML.
Ai-Ti Chiu, Jia-Lien Hsu · SWWS · 2006
Abstract—In the research field of music information re-trieval, a precise and efficient query/filter system are criticalto provide flexible services and applications. Meanwhile, theMusicXML is considered as one of the most popular data formatfor better structured description on music objects. With theemerging streaming data model, we propose an efficient filtersystem against streaming MusicXML.User queries are specified in XPath expressions. Three keyissues are investigated in this system, including query notes ofcrossing measures, polyphonic music query, and the order of querynotes. We propose an automaton-based method to efficientlyresolve user queries against streaming MusicXML. I. I NTRODUCTION In the researches of processing XML, most researchersfocus issues on XML filtering and XML stream processing.In [1], [2], [4], [5], the corresponding authors proposed theirown data structure to represent user queries. According tothe proposed data structure, There are two types of works,automaton-based methods and index-based methods. Mostresearches apply automaton-based method, because it allowsmore complicated user queries. Automaton-based systemalso queries/filters more complicated XML documents. Aslong as user queries are transformed into an automaton,the incoming XML documents are navigated through theautomaton to identify the matched queries and users.In [2], single XPath expression (user query) constructsa hierarchical pushdown transducer (HPDT). The authorsdefined pushdown transducer template for each location stepin a XPath expression. Then, lots of pushdown transducertemplate are combined into one hierarchical pushdown trans-ducer. In [1], XPush Machine has been proposed to processstreaming XML documents. They also focus issues on pro-cessing many XPath expressions in the same time. Each ofthe XPath expressions has many predicate expressions needto be evaluated.Our method is based on XPush Machine. However, XPushMachine is not robust for processing streaming MusicXMLdocuments. Fig. 1 shows the system architecture. XPushMachine is constructed based on user queries, and incom-ing streaming MusicXML is filtered by XPush Machine toevaluate the There are three key issues when apply XPushMachine to process streaming MusicXML, including “querynotes of crossing measures,” “polyphonic music query,” and“the order of query notes.”Query notes of crossing measures: An user query is apiece of music object, and notes in an user query may cross