Scopus İndeksli Yayınlar Koleksiyonu / Scopus Indexed Publications Collection
Permanent URI for this collectionhttps://hdl.handle.net/11147/7148
Browse
3 results
Search Results
Conference Object Citation - Scopus: 11An Analysis of Large Language Models and Langchain in Mathematics Education(Association for Computing Machinery, 2023) Soygazi,F.; Oğuz, DamlaThe development of large language models (LLMs) has led to the consideration of new approaches, particularly in education. Word problems, especially in subjects like mathematics, and the need to solve these problems by collectively addressing specific stages of reasoning, have raised the question of whether LLMs can be successful in this area as well. In our study, we conducted analyses by asking mathematics questions especially related to word problems using ChatGPT, which is based on the latest language models like Generative Pretrained Transformer (GPT). Additionally, we compared the correct and incorrect answers by posing the same questions to LLMMathChain, a mathematics-specific LLM based on the latest language models like LangChain. It was observed that the answers obtained were more successful with ChatGPT (GPT 3.5), particularly in the field of mathematics. However, both language models were found to be below expectations, particularly in word problems, and suggestions for improvement were provided. © 2023 ACM.Article Citation - Scopus: 1An Interestingness Measure for Knowledge Bases(Elsevier, 2023) Oğuz, Damla; Soygazi, FatihAssociation rule mining and logical rule mining both aim to discover interesting relationships in data or knowledge. In association rule mining, relationships are identified based on the occurrence of items in a dataset, while in logical rule mining, relationships are determined based on logical relationships between atoms in a knowledge base. Association rule mining has been widely studied in transactional databases, mainly for market basket analysis. Confidence has become the most widely used interesting measure to assess the strength of a rule. Many other interestingness measures have been proposed since confidence can be insufficient to filter negatively associated relationships. Recently, logical rule mining has become an important area of research, as new facts can be inferred by applying discovered logical rules. They can be used for reasoning, identifying potential errors in knowledge bases, and to better understand data. However, there are currently only a few measures for logical rule mining. Furthermore, current measures do not consider relations that can have several objects, called quasi-functions, which can dramatically alter the interestingness of the rule. In this paper, we focus on effectively assessing the strength of logical rules. We propose a new interestingness measure that takes into account two categories of relations, functions and quasi-functions, to assess the degree of certainty of logical rules. We compare our proposed measure with a widely used measure on both synthetic test data and real knowledge bases. We show that it is more effective in indicating rule quality, making it an appropriate interestingness measure for logical rule evaluation. & COPY; 2023 Karabuk University. Publishing services by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).Article Citation - WoS: 4Citation - Scopus: 5Dma: Matrix Based Dynamic Itemset Mining Algorithm(IGI Global Publishing, 2013) Oğuz, Damla; Yıldız, Baroş; Ergenç, BelginUpdates on an operational database bring forth the challenge of keeping the frequent itemsets up-to-date without re-running the itemset mining algorithms. Studies on dynamic itemset mining, which is the solution to such an update problem, have to address some challenges as handling i) updates without re-running the base algorithm, ii) changes in the support threshold, iii) new items and iv) additions/deletions in updates. The study in this paper is the extension of the Incremental Matrix Apriori Algorithm which proposes solutions to the first three challenges besides inheriting the advantages of the base algorithm which works without candidate generation. In the authors' current work, the authors have improved a former algorithm as to handle updates that are composed of additions and deletions. The authors have also carried out a detailed performance evaluation study on a real and two benchmark datasets.