الفهرس | Only 14 pages are availabe for public view |
Abstract Data have grown into enormous size in the last few years where its influence has increased on everyone{u2019}s life. Data resources like YouTube videos, images or news are being posted to the web daily with huge amounts. With the existence of such massive amount of resources, there is a continuous need to find intelligent methods to process and benefit from it in an unsupervised manner. We define intelligent methods in this context as methods that do not need any human intervention for processing web resources. Human manual effort would be cumbersome in dealing with such massive amount of resources. Hence, intelligent methods should be dependent on data only to learn and provide computational intelligence to the user. In this thesis, we propose clustering as our intelligent method to provide the required computational intelligence. Clustering is the task of finding different phenomena hidden in the data. We intend to use phenomena information to benefit from the data in data mining tasks with the best way possible. Towards this goal, we propose two different paths to benefit from the data in unsupervised manner via clustering. The first path is to give identification to anonymous data gathered from the web. Converting anonymous data into identifiable one would represent a new source of data to learn from. We propose the idea of identifying relations between entities in the text from randomly gathered news from the web at their publication time. Gathering a news dataset and automatically extracting entities and relations between them allowed us to build a dataset of relations examples and learning from |