Cognitive Science Colloquia


Cognitive Science Colloquia are held* throughout the year and consist of seminars/talks given by invited speakers from various universities, as well as seminars by the METU staff or our own PhD students. The Colloquia aim to share information on cognitive science research, as well as promote academic research and collaboration. The seminars are open to anyone interested and provide an opportunity for our students to keep up with the recent research conducted in the field, allowing them to observe how research topics develop through academic exchange.

* Unless it is announced otherwise, seminars take place every Wednesday at 12:40, in the Informatics Institute (room II-03), METU.

For past years' colloquia, please visit here.

NEXT IN THE SERIES     >    

Modeling Hereditary and Behavioral Patterns of Facial Expressions
by Hamdi Dibeklioglu (Bilkent)
on 26th of February

Abstract Over the last two decades, computational analysis of facial expressions has been a very active area of research. Today, following the recent dramatic advances in the fields of machine learning and computer vision, we are able to model subtle and complex patterns of facial responses in a reliable manner. Our findings indicate that facial expressions do not only reveal instant emotional state of individuals but also display hereditary and behavioral characteristics. This talk will focus on deep architectures to model such patterns for various tasks such as recognizing preferences, assessing psychopathology, and predicting how your future child's expressions look like.

This Year (2019-2020)

Spring Semester

Algorithmic Identification of Observers in Arbitrary Dynamical Systems
by Gülce Kardeş (Physics, METU)
on 19th of February

Abstract Since Shannon’s introduction to information theory, we have tended to fix our attention to a syntactic analysis of information. Yet a profound disclosure of meaning in information is reported by a group of systems (e.g. biological organisms) that use information from their environment which causally contributes to their ability of maintaining existence. In this study, we consider the problem of identifying such systems, called observers, that render a priori companion of semantic information (Kolchin- sky and Wolpert, 2018) possible. We introduce spin models as a platform in which we construct an identification procedure. Our findings present the physical properties of spin systems that have determinant role in specifying observers. We also investigate further characteristics of semantic information, and list several implications to discuss the applicability of our work to any physical system.


Fall Semester

Quasi-supervision for morphology learning without morphological analysis
by Prof. Cem Bozşahin (Cognitive Science, METU)
on 18th of December

Abstract Learning the meaning of sublexical elements is a main concern in language understanding. Turkish children do that quite early. We report some experiments that we do at ODTU, Hacettepe and Groningen which treats morphology as learning from a latent variable, using phonology, semantics and lexicon, and without prior morphological analysis or information.


Counter-expectational sense in Turkish and in general
by Ceyhan Temürcü (Cognitive Science, METU)
on 4th of December

Abstract In an attempt to inaugurate a line of research we started with Deniz Zeyrek two years ago, I will outline our proposal for analyzing the functioning of counter-expectational (CE) markers.  Apparently, all human languages have strategies for challenging possible inferences from a preceding clause or discourse segment. This sense, which take part in the semantic ranges in Turkish ama or fakat, and of English but, has often been dubbed 'counter-expectational' and subsumed under the more general family of 'contrastive' senses. I this talk I will (i) provide critical data that reveals some semantic and pragmatic ingredients of the CE sense, (ii) propose a descriptive specification for this sense and (iii) provide prospects for a formal treatment of CE in a multi-layered version of DRT, which distinguishes temporal, epistemic, volitional, and illocutionary layers. 


Visual perception of human and robot actions: Interdisciplinary studies between cognitive neuroscience and social robotics
by Dr. Burcu Ayşen Ürgen (Cognitive Science, METU)
on 27th of November

Abstract In the first part of talk, I will talk about an fMRI study in which we investigate the representational properties of the human action observation network. More specifically, I will present data that show how the human brain represents evolutionarily new, human-specific actions such as writing and drawing. In the second part, I will talk about an EEG study that shows how we can utilize our knowledge of the human action observation network for designing social robots that can interact with humans successfully. 


Annotation as a science: Lessons learnt from discourse annotation
by Dr. Deniz Zeyrek (Cognitive Science, METU)
on 20th of November

Abstract In this talk, I will talk about aspects of transferring the theoretical formulation of discourse phenomena, based on empirical observations, into a model that can be used for the specification of linguistic annotation.  I will outline the procedure in general terms, concentrating on my work in Turkish Discourse Bank and TED Multilingual Discourse Bank. 


TÜRKÇE İÇİN GÖZETİMSİZ SÖZDİZİMSEL BELİRSİZLİK GİDERME
by Özkan Aslan (Cognitive Science, METU)
on 13th of November

Abstract Doğal dillerde bir tümce, her biri farklı yapısal yorumlara karşılık gelen birden çok sözdizim ağacı ile gösterilebilir. Bu durum sözdizimsel belirsizlik olarak adlandırılır. Sözdizimsel belirsizlik giderme, basitçe, tümceden elde edilen sözdizim ağaçlarının bağlama göre en uygun olandan en az uygun olana doğru sıralanmasıdır. Bu tezde, sözdizimsel belirsizlik giderme problemi Türkçe için ele alınmış ve gözetimsiz yönteme dayanan bir çözüm önerilmiştir. Yöntemin gözetimsiz olarak adlandırılmasının nedeni sözdizim ağaçlarının sıralanmasında kullanılan olasılık modellerinin imlenmemiş bir metin koleksiyonundan elde edilmiş olmasıdır. Tez kapsamında, sözdizimsel belirsizlik giderme işini gerçekleştirmek amacıyla, sözdizimsel çözümleyici, Morfolog adlı biçimbilimsel çözümleyici ve TrLex adlı sözlükçe gibi özgün altyapı ögeleri tasarlanmış ve bunları eşgüdümlü biçimde yöneten TMoST adlı bir dizge oluşturulmuştur. Ayrıca öbek yapı dilbilgisine dayanan yeni bir tümce çözümleme gösterimi önerilmiş ve bu gösterimde biçimbilimsel ve sözdizimsel yapıları birlikte işleyebilmeyi sağlayan ve dizimbirim adı verilen yeni bir kavram tanıtılmıştır. Çalışmada, bazıları özgün olan 24 olasılık modeli kullanılmıştır. Modellerin problem üzerindeki başarımını ölçmeye imkân veren AUT adlı bir ağaç yapılı derlem üretilmiştir. Alanyazında sözdizimsel belirsizlik giderme için başarım, en uygun ağacın sıralamada bulunduğu konum ile veya birinci sıradaki ağacın en uygun ağaca olan benzerliği ile ölçülmektedir. Tezde iki yeni başarım ölçüsü daha önerilmiş ve bağıntı adı verilen ölçünün daha kararlı olduğu değerlendirilmiştir. Olasılık modelleri tek başına kullanıldığında en iyi başarım, üçlü biçimbirim dil modeliyle elde edilmiştir. Modeller birleştirildiğinde ulaşılan en iyi bağıntı değeri ise yaklaşık 0,41 olmuştur.


Some Introduction to Peircean Semiotics and Pragmatism
by Tunç Güven Kaya (Cognitive Science, METU)
on 16th of October

Abstract Among many things, Charles Sanders Peirce (1839-1914) was a logician, scientist and philosopher. He is mostly known for his Semiotics and Pragmatism. According to him, "all thinking is conducted in signs" and Semiotics is the scientific study of the signs and their relations. Pragmatism, on the other hand, is a certain subject under the general theory of Semiotics. In this talk, we will very briefly introduce these concepts and try to connect them to the contemporary debates of Cognitive Science.

Fırat Öter


Modeling semantics: Vector representation, classical composition and beyond
by Fırat Öter (Cognitive Science, METU)
on 16th of October

Abstract Language and vision are two complementary modalities that employ deixis in communication. Verbal deictic expressions are necessarily accompanied by other modalities, such as pointing by gestures or pointing by gaze. The presence of multiple modalities facilitates disambiguating spatial reference to objects in the environment. We report an experimental study that investigated the deictic function of gaze in avatar agents. For this, we designed a virtual reality (VR) environment, which provided a joint attention setting between human participants and avatars. The participants responded to a set of questions that involved explicit or implicit referring expressions about the spatial locations of objects on a table, while their eye movements were recorded. The explicit statement of referring expressions, such as “what is the object at the left bottom?” revealed different patterns in terms of accuracy, response times and gaze allocation of the participants, compared to the implicit statements of referring expressions, i.e. “what is the object there?” that was accompanied by deictic gaze. The findings also showed that the avatar morphology might have a significant influence on the results. Overall, our findings show that the study of deictic gaze in virtual reality environments has the potential to expand our knowledge about mechanisms that underlie human spatial cognitive abilities as well as our knowledge about the interaction between humans and agents in VR.


Human Agent Interaction in Virtual Reality: An Experimental Investigation of Deictic Gaze in a Joint Attention Setting
by Dr. Cengiz Acartürk (Cognitive Science, METU)
on 9th of October

Abstract Language and vision are two complementary modalities that employ deixis in communication. Verbal deictic expressions are necessarily accompanied by other modalities, such as pointing by gestures or pointing by gaze. The presence of multiple modalities facilitates disambiguating spatial reference to objects in the environment. We report an experimental study that investigated the deictic function of gaze in avatar agents. For this, we designed a virtual reality (VR) environment, which provided a joint attention setting between human participants and avatars. The participants responded to a set of questions that involved explicit or implicit referring expressions about the spatial locations of objects on a table, while their eye movements were recorded. The explicit statement of referring expressions, such as “what is the object at the left bottom?” revealed different patterns in terms of accuracy, response times and gaze allocation of the participants, compared to the implicit statements of referring expressions, i.e. “what is the object there?” that was accompanied by deictic gaze. The findings also showed that the avatar morphology might have a significant influence on the results. Overall, our findings show that the study of deictic gaze in virtual reality environments has the potential to expand our knowledge about mechanisms that underlie human spatial cognitive abilities as well as our knowledge about the interaction between humans and agents in VR.


The Relation between Time and Meaning
by Giray Songül (Cognitive Science, METU)
on 2nd of October

Abstract There are at least two strong opinions in computational theories of mind arguing whether the syntax suffices to produce semantics. In this talk, using intuitions from recently developed artificial neural network models, the problem of exactly where semantics are formed in a syntactic system such as a brain will be discussed. The main argument of the talk, that will start about computation and go through philosophy, is that the semantic value can only be formed on the presence of time using nonlinear activation functions and its main purpose is to create time efficiency for its possessor. Perception of time greatly affects the perception of semantic value. If there is no time, there is no meaning at all.