Turkish Touch to Artificial Intelligence

by İTÜ | May 06, 2020
Istanbul Technical University Turkish Natural Language Processing (DDI) Laboratory offers solutions to researchers working in this field and develops solutions. The research group, which is founded by Assoc. Dr. Gülşen Eryiğit, also its coordinator, contributes to many sectors in our country.

 
Istanbul Technical University Natural Language Processing Group, which specializes in this area, is the oldest and largest language technology group in Turkey. Established in 2000, the group still operates within the ITU Faculty of Computer and Informatics. Under the coordination of Assoc. Dr. Gülşen Eryiğit, the group currently has 3 faculty members, 5 research assistants and nearly 30 doctoral, postgraduate and volunteer undergraduate students. The center has conducted 12 scientific research projects so far. Faculty members and researchers of the group provide academic support during the suggestion and implementation stages of 8 TEYDEB projects. The research group works like an R&D base.

We directed our questions regarding the studies and objectives of the research group, to Assoc. Dr. Gülşen Eryiğit, Coordinator of Istanbul Technical University Natural Language Processing Group.

itu-ddi-lab-2

You are working on natural language processing projects in Turkish. What is the importance of Turkish?

Natural language processing, which is one of the most popular fields of artificial intelligence, brings together opportunities for researchers and students who will find solutions to these problems, along with many new problems in Turkish. Turkish, which is an articulated and formally rich language, is very different from English, which is the most studied language in natural language processing. For this reason, when new techniques developed on English are tried to be applied directly to Turkish, they cannot fully operate as they do in English. Turkish draws attention not only of national researchers but also of international researchers due to its different structure. The algorithms developed on Turkish are similarly examples for many languages ​​with morphological richness. Turkish ranks 20th in the world with 83 million speakers. Turkish languages ​​stand out as the languages ​​spoken by more than 200 million people in an area of ​​approximately 12 million square kilometers.

What would you say about the economic dimension of the projects carried out under the roof of the laboratory and the market target?

Natural Language Processing is seen as one of the most challenging areas of artificial intelligence all over the world. Following the most up-to-date techniques in the world, ITU DDI Group has determined its most important mission as producing solutions to the additional challenges it presents in the field of DDI as of the structure of Turkish. The approaches it has created with this mission provide benefits both to many industrial studies and to researchers from different disciplines. Faculty members of the group give academic counseling for many companies and R&D centers in ITU ARI Teknokent. The first licensing in the USA was carried out by İTÜNOVA TTO in 2016 with the ITU Natural Language Processing Software chain. The team creates critical knowledge for both national and international companies with different projects in the field of natural language processing.

Can we learn about projects carried out by research groups and the projects that stand out?

Some of the research projects involving the members of the group, which conduct research at different levels on natural language processing in Turkish and concentrate on the application of these research outcomes to various sectors such as Telecom, banking, law, automotive;

• Separation of Web2.0 Sentences, TÜBİTAK, 2013-2015 (TÜBİTAK Project Performance Award)

• Automatic Translation System from Turkish to Turkish Sign Language and Avatar Animation, (TÜBİTAK, 2014-2016)

• Turkish Personal Assistant Application on Smart Phones and Tablets, (T.C. Ministry of Science, Industry and Technology, SANTEZ, 2013-2014)

• Commitment Analysis of Turkish, (ITU BAP, 2005-2007)

• Turkish-English (2011-2013) and Turkish-Turkmen (2006-2009) Machine Translation

Can you share some examples of industry-oriented projects within the scope of university-industry cooperation?

The group, which has developed strong industrial relations over the years, has been involved in and continues to take part in many industry-oriented projects within the framework of university-industry cooperation. Some of these projects can be listed as “Automotive Support Channels Dialogue System”, “Target Based Emotion Analysis in Telecom Field”, “Extraction of Information through Noisy Banking Texts”, “Banking Dialogue Systems” and “Text Analysis on Legal Documents”.

The group provides researchers working in the field of natural language processing with many Turkish natural language processing basic level tools (text normalization, word parser, sentence parser, etc.). Some of these also serve as a web service via tools.nlp.itu.edu.tr. Similarly, many important data sets produced by the group (Turkish Tree Structures, multi-word term and asset name sources, Turkish sign language sources, etc.) are shared with the researchers via the same address.

ITU DDI Group, actively participated and continues to participate in many EU projects such as;  UD (Universal Dependencies), CLARIN (Common Language Resources and Technology Infrastructure, EU 7th Framework), PARSEME (Parsing and Multiword Expressions, European Cost Action IC1207) and EnetCollect (European Network for Combining Language Learning with Crowdsourcing Techniques, European Cost Action CA16105). ITU DDI Laboratory was selected as the month's laboratory by the ENETCOLLECT action in January 2020.

Can you also talk about national and international collaborations?

Due to many European Union actions and international collaborations it has carried out from the past, ITU Natural Language Processing Group has signed many publications with international co-authors. Many of the researchers started working in different research institutions in Europe and the USA once they were graduated. Our group, which has adopted as a responsibility to support natural language research researchers especially in the Turkic Republics, strives to transfer its experience in Turkish to other Turkic languages. Many references edited through Turkish and Turkic Language resources continues to serve for the use of many researchers.

The team collaborates with many different disciplines, especially linguistics and robotics, and conducts joint projects. İTÜ DDİ has become a pioneering working group in the Turkish and Natural Language Processing in the world and in Turkey by opening many data sets and natural language processing tools developed over the years to the service of all researchers.

Which awards did you receive as ITU DDI Research Group?

-TUBITAK Project Performance Award, 2018

- ITU Appreciation Plaque - Academic Success “First licensing from ITUNOVA TTO to USA”, December 2016

- TÜBİTAK H2020 Project Application Above Threshold Award, November 2015

- ITU Thanksgiving Plaque - EU 7th Framework Program and Contribution to Research Infrastructures, January 2012

- 2 Siemens Excellence Awards, June 2007

- Winner of the “CoNLL-XI Shared Task: Multi-lingual Dependency Parsing” competition, June 2007

- “CoNLL-X Shared Task: Multi-lingual Dependency Parsing” competition winner, June 2006

itu-ddi-lab-1

About Assoc. Prof. Gülşen Eryiğit

A faculty member of Istanbul Technical University (ITU) Computer and Informatics Faculty, Dr. Gülşen Eryigit, is a founding member and the coordinator of the Natural Language Processing Group. Dr. Eryiğit received her master’s and doctorate degrees from ITU Computer Engineering Department in 2002 and 2007. She was a visitor researcher at Vaxjö University in Sweden in 2006 and took part in the studies selected first in the CoNLL (natural language learning conference series) 2006 and 2007 multilingual commitment workshops with her group. In 2007, she was granted for the Siemens Excellence Award due to her doctorate studies. She has worked as an author and referee in many prestigious magazines and conferences. As of 2020, Dr. Eryiğit has over 3000 citations in Google Scholar.

Dr. Eryiğit has been involved in the European Union (EU 7th Framework Program, Clarin - Common Language Resources and Technology Infrastructure) project as the representative of Turkey. In addition, she worked as an executive and researcher in the projects supported by the T.C. Ministry of Industry (Santez), EU (Cost) projects, TÜBİTAK (1001 and 1003) and worked as a consultant in 6 TÜBİTAK TEYDEB supported projects. At the same time, she served as a referee and observer in these funding institutions (EU H2020, Cost Action programs, TÜBİTAK, T.C. Ministry of Industry) in both research and industrial project proposals. Turkish Natural Language Processing Web Services (tools.nlp.itu.edu.tr) offered to national and international researchers serve over 200 researchers as of 2020.