Recursive Deep Models have been used as powerful models to learn \ncompositional representations of text for many natural language processing tasks. \nHowever, they require structured input (i.e. sentiment treebank) to encode sentences \nbased on their tree-based structure to enable them to learn latent semantics \nof words using recursive composition functions. In this paper, we present our \ncontributions and efforts for the Turkish Sentiment Treebank construction. We \nintroduce MS-TR, a Morphologically Enriched Sentiment Treebank, which was \nimplemented for training Recursive Deep Models to address compositional sentiment \nanalysis for Turkish, which is one of the well-known Morphologically Rich \nLanguage (MRL). We propose a semi-supervised automatic annotation, as a distantsupervision \napproach, using morphological features of words to infer the polarity of \nthe inner nodes of MS-TR as positive and negative. The proposed annotation model \nhas four different annotation levels: morph-level, stem-level, token-level, and \nreview-level. Each annotation level’s contribution was tested using three different \ndomain datasets, including product reviews, movie reviews, and the Turkish Natural \nCorpus essays. Comparative results were obtained with the Recursive Neural Tensor Networks (RNTN) model which is operated over MS-TR, and conventional machine learning methods. Experiments proved that RNTN outperformed the baseline methods and achieved much better accuracy results compared to the baseline methods, which cannot accurately capture the aggregated sentiment information.