SANTI-Network Prototype of an Indonesian Multi-level Tagger

A multi-level tagger (MLT) to annotate Indonesian texts is crucial for corpus-based analyses. Prior to implementing the creation of such an MLT, a prototype is required as a model. The primary objective of this project is to develop a prototype of an Indonesian MLT that enables users to perform morphological, morphosyntactic (POS) and semantic annotation in a single pass using NooJ [1]. Currently, two NooJ-based Indonesian taggers are available: SANTI-morf (morphology) and SANTI-POS (morphosyntax). Note that no automatic semantic tagger is available for the Indonesian language; thus, semantic annotation functionality needs to be created and evaluated; these are subsidiary aims. Even though NooJ-based Indonesian morphological and POS (and later semantic) analysis systems are available, users cannot apply multiple tagging in one pass. The prototype is expected to demonstrate how to perform morphology-POS-semantic annotation in one pass. To fulfill the aims of this study, I implemented the following methodological procedure. I initially augmented selected SANTI-POS lexical entries (more than 400) with semantic analysis labels (using the USAS tagset). For this prototype, only the first semantic analysis label was considered. Then, the lexicon was included in NooJ resources, alongside SANTI-POS and SANTI-morf resources. A prototype was successfully created. The average precision, coverage and ambiguity rates of the prototype are 85%, 86% and 15%, respectively. Users can now tag a corpus with this MLT prototype and perform morphological-POS-semantic searches.