Dependency TreeBanks

Tesnière introduced the dependency trees, structural order, the concept of dependency and applied his representation concepts in a variety of languages such as French, Greek, Russian, Italian, and so on. In structural order, syntactic relations are presented in a hierarchical manner as opposed to the linear order. He uses “stemmas” to reflect hierarchy.

Today, dependency grammars are divided into two. One study from the tradition that applies this distinction to the dependency grammar is the Prague Dependency Treebank (PDT) developed by the Prague School of Functional and Structural Linguistics. A famous example for the other school of thought that displays linear order is the Penn Treebank that functioned between the years of 1989-1996, containing seven million annotated texts from American English.

In 2005, The Stanford Dependencies developed for the parsing of the English language and to be used in NLP studies and in Stanford Dependency Parser. Stanford Dependencies were acknowledged as the standard for the dependency analyses of English. However, the Stanford Dependency parser could not reach an adequate accuracy when it was used with other dependency schemes. In the following years, the Universal Dependency Treebank (UDT) project pioneered to develop treebanks for languages other than English by transforming the Stanford dependencies into a more inclusive annotation scheme for a diverse set of languages.

The developments in the dependency treebanking made it clear that Turkish language needed a Treebank of its own. The first Turkish language dependency treebank is METU-Sabanci Turkish Treebank. This treebank used a corpus that consisted of 7,262 sentences and included morphological and syntactic annotations. In 2016, this tree-bank was revisited under the name of ITU-METU-Sabancı Treebank (IMST) to reduce the inconsistencies of its earlier version. They succeeded to reduce inconsistencies by applying a new annotation scheme. As a last step, The Bogazici-ITU-METU-Sabancı Treebank (BIMST) is updated as the same corpus. Having a linguistic team of three people, they created a new annotation scheme for IMST and manually re-annotated the data of 5.635 sentences while introducing new dependency relations that were not present in IMST.

Video Lectures

For Developers

You can also see either Cython, Java, C++, C, Swift, Js, or C# repository.

Requirements

• Python 3.7 or higher
• Git

Python

To check if you have a compatible version of Python installed, use the following command:

python -V

You can find the latest version of Python here.

Git

Install the latest version of Git.

Pip Install

pip3 install NlpToolkit-DependencyParser

Download Code

In order to work on code, create a fork from GitHub page. Use Git for cloning the code to your local or below line for Ubuntu:

git clone <your-fork-git-link>

A directory called DataStructure will be created. Or you can use below link for exploring the code:

git clone https://github.com/starlangsoftware/TurkishDependencyParser-Py.git

Open project with Pycharm IDE

Steps for opening the cloned project:

• Start IDE
• Select File | Open from main menu
• Choose TurkishDependencyParser-Py file
• Select open as project option
• Couple of seconds, dependencies with Maven will be downloaded.

Cite

@INPROCEEDINGS{9259799,
author={A. {Kuzgun} and N. {Cesur} and B. N. {Arıcan} and M. {Özçelik} and B. {Marşan} and N. {Kara} and D. B. {Aslan} and O. T. {Yıldız}},
booktitle={2020 Innovations in Intelligent Systems and Applications Conference (ASYU)}, 
title={On Building the Largest and Cross-Linguistic Turkish Dependency Corpus}, 
year={2020},
volume={},
number={},
pages={1-6},
doi={10.1109/ASYU50717.2020.9259799}}