Browsing by Author "Sezerer, Erhan"

Now showing 1 - 2 of 2

Discovering specific semantic relations among words using neural network methods
(Izmir Institute of Technology, 2021-10) Sezerer, Erhan; Tekir, Selma; Izmir Institute of Technology
Human-level language understanding is one of the oldest challenges in computer science. Many scientific work has been dedicated to finding good representations for semantic units (words, morphemes, characters) in languages. Recently, contextual language models, such as BERT and its variants, showed great success in downstream natural language processing tasks with the use of masked language modelling and transformer structures. Although these methods solve many problems in this domain and are proved to be useful, they still lack one crucial aspect of the language acquisition in humans: Experiential (visual) information. Over the last few years, there has been an increase in the studies that consider experiential information by building multi-modal language models and representations. It is shown by several studies that language acquisition in humans start with learning concrete concepts through images and then continue with learning abstract ideas through text. In this work, the curriculum learning method is used to teach the model concrete/abstract concepts through the use of images and corresponding captions to accomplish the task of multi-modal language modeling/representation. BERT and Resnet-152 model is used on each modality with attentive pooling mechanism on the newly constructed dataset, collected from the Wikimedia Commons. To show the performance of the proposed model, downstream tasks and ablation studies are performed. Contribution of this work is two-fold: a new dataset is constructed from Wikimedia Commons and a new multi-modal pre-training approach that is based on curriculum learning is proposed. Results show that the proposed multi-modal pre-training approach increases the success of the model.
News story analysis with credibility assessment by opinion mining
(Izmir Institute of Technology, 2015-07) Sezerer, Erhan; Tekir, Selma
With the growing influence of media and the popularity and widespread use of social networks, credibility of the news sources became an important subject that needs more attention. The biggest problem of finding credible sources is, instead of giving every aspect of the incident, news sources tend to accept one of the parties’ idea as a whole while rejecting every other ideas, or even worse, they focus on only one side of the incident and ignoring the rest. Credibility is defined as “The quality of believable and trustworthy”. The notion of trustworthiness can further be decomposed into components like bias, fairness, factual/ opinionated, etc. In this thesis, credibility is measured using the fact/opinion ratio of the articles. Two methods, which are the traditional Naive Bayes method and the Relativistic method, are proposed. The intuition of relativistic method comes from the theory of relativity where the sentiment of the articles is determined relatively to the ordinary context used by people in daily speech. We have tested our methods on four different types of data, hand-written articles, editorials, New York Times articles and Reuters articles, and aimed to show that our proposed models are able to differentiate the sentiments in the articles. In the experimental work, we provided a detailed evaluation of the results.