From Universal Dependencies to error tagging: Using dependency parsing to annotate errors in Greek learner sentences

Sammanfattning

This master’s thesis concerns the development of tools for automatic annotation oflearner data with a focus on Greek. Firstly, I developed a tool for automatic errorannotation, on an attempt to extract mainly morphological errors from the morphosyntacticannotation of parallel learner and standard data in the Universal Dependencies(UD) framework. The human evaluation of the tool proves the difficulty of this task,though also the tool’s suitability for a restricted domain of application. In addition,different models for syntactic parsing were trained, using different percentages of UDannotatedstandard and learner data. The primary aim of this second task was choosingthe most fitting training data for in-domain testing, as well as measure the effect thatthe particular nature of learner data, when occurring in the training set, can have onmodel performance. The quantitative and qualitative evaluation indicate a significantimprovement of performance on the learner test setting when learner data have been utilizedin the training process, and the best performing model across domains was the onetrained on both standard and learner data. Moreover, the addition of learner data to thetraining did not seem to have a negative effect on model performance overall. Lastly,a combination of the aforementioned morphosyntactic and error annotation methodswas attempted, in order to annotate novel learner data; the reported results show acompatibility of the employed methods, nevertheless a limitation of performance, andthus space for improvement. The importance of this work lies in the enhancement ofcomputational methods and automated tools used for morphosyntactic parsing and errortagging, as well as in the promotion of linguistic research on language learning, boostinga re-evaluation of learning and teaching methods