Dependency Parsing

Task Definition

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

Data Annotation

Preparation

1. Collect a set of sentences to annotate.
2. Each sentence in the collection must be named as xxxx.yyyyy in increasing order. For example, the first sentence to be annotated will be 0001.train, the second 0002.train, etc.
3. Put the sentences in the same folder such as Turkish-Phrase.
4. Build the project and put the generated sentence-dependency.jar file into another folder such as Program.
5. Put Turkish-Phrase and Program folders into a parent folder.

Annotation

1. Open sentence-dependency.jar file.
2. Wait until the data load message is displayed.
3. Click Open button in the Project menu.
4. Choose a file for annotation from the folder Turkish-Phrase.
   
5. For each word in the sentence, click the word, and link that word to its head. Select appropriate dependency tag for that dependency.
6. Click one of the next buttons to go to other files.

For Developers

Requirements

• Java Development Kit 8 or higher, Open JDK or Oracle JDK
• Maven
• Git

Java

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

java -version

If you don't have a compatible version, you can download either Oracle JDK or OpenJDK

Maven

To check if you have Maven installed, use the following command:

mvn --version

To install Maven, you can follow the instructions here.

Git

Install the latest version of Git.

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 UniversalDependencyParser will be created. Or you can use below link for exploring the code:

git clone https://github.com/olcaytaner/UniversalDependencyParser.git

Open project with IntelliJ IDEA

Steps for opening the cloned project:

• Start IDE
• Select File | Open from main menu
• Choose UniversalDependencyParser/pom.xml file
• Select open as project option
• Couple of seconds, dependencies with Maven will be downloaded.

Compile

From IDE

After being done with the downloading and Maven indexing, select Build Project option from Build menu. After compilation process, user can run UniversalDependencyParser.

From Console

Go to UniversalDependencyParser directory and compile with

 mvn compile 

Generating jar files

From IDE

Use package of 'Lifecycle' from maven window on the right and from UniversalDependencyParser root module.

From Console

Use below line to generate jar file:

 mvn install

Maven Usage

    <dependency>
        <groupId>io.github.starlangsoftware</groupId>
        <artifactId>UniversalDependencyParser</artifactId>
        <version>1.0.2</version>
    </dependency>

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}}