In this paper, we focus on two important problems of social media text normalization, namely: vowel and diacritic restoration. For these two problems, we propose a hybrid model consisting both a discriminative sequence classifier and a language validator in order to select one of the morphologically valid outputs of the first stage. The proposed model is language independent and has no need for manual annotation of the training data. We measured the performance both on synthetic data specifically produced for these two problems and on real social media data. Our model (with 97.06% on synthetic data) improves the state of the art results for diacritization of Turkish by 3.65 percentage points on ambiguous cases and for the vowel restoration by 45.77 percentage points over a rule based baseline with 62.66% accuracy. The results on real data are 95.43% and 69.56% accordingly.

Multiword expressions (MWEs) present particular and distinctive semantic properties, hence their automatic extraction receives special attention from the natural language processing (NLP) and corpus linguistics community, and is still an active research area. Unfortunately, the creation of necessary resources for this task is quite rigorous and many languages suffer from the lack of these; as in the case for Turkish.This study presents our MWE annotations on recently introduced Turkish Treebanks, which focuses on annotating various types of linguistic units and expressions, including named entities, numerical expressions, idiomatic phrases, verb phrases with auxiliaries and duplications. The paper aims to provide a benchmark and pave the way towards further MWE extraction research for Turkish. To this end, the paper also introduces our experimental results with seven baseline approaches, a dependency parser and a previously introduced rule-based extractor on these annotated corpora. Our highest performances achieved over these resources are about 60% F-scores.