NLP APPLICATION IN THE HEBREW LANGUAGE FOR ASSESSMENT AND LEARNING

This chapter describes the natural language processing (NLP) and automated essay scoring (AES) systems for the Hebrew language. Hebrew has a particular orthography (vowels are not written explicitly) and rich morphology (large lexeme families) with relatively complex conjugation and derivation rules. This fact led to the development of an infrastructure for NLP, which, at the time, was not available in Hebrew. Various tools and processes constituting this infrastructure and the AES system built upon it are described. After screening for off-topic and “gibberish” essays, scoring the essays is based on lexical, morphological, grammatical, syntactic, and semantic features, which are grouped into factors. This chapter reports the validities of the system for various AES tasks, which turn out to be on par with human scoring. In addition to the AES system, there is a description of other uses of the NLP infrastructure. For example, it is used for giving feedback on students’ essays and to assess the reading difficulty of textbooks and reading comprehension passages. At the end of the chapter there is a brief discussion of open issues relating to AES in general and the Hebrew AES system in particular.