Máté Pataki

Comparison of Overlap Detection Techniques

Easy access to the World Wide Web has raised concerns about copyright issues and plagiarism. It is easy to copy someone elses work and submit it as someones own. This problem has been targeted by many systems, which use very similar approaches. These approaches are compared in this paper and suggestions are made when different strategies are more applicable than others. Some alternative approaches are proposed that perform better than previously presented methods. These previous methods share two common stages: chunking of documents and selection of representative chunks. We study both stages and also propose alternatives that are better in terms of accuracy and space requirement. The applications of these methods are not limited to plagiarism detection but may target other copy-detection problems. We also propose a third stage to be applied in the comparison that uses suffix trees and suffix vectors to identify the overlapping chunks.

StreamOnTheFly: a peer-to-peer network for radio stations and podCasters

The StreamOnTheFly network demonstrates new ways of management and personalisation technologies for audio. The architecture is based on a decentralized network of software components using automatic metadata replication in a peer-to-peer manner. The network also promotes a new common metadata schema and content exchange format. Content reuse and content exchange is made possible by StreamOnTheFly in several use cases.

An assistive interpreter tool using glove-based hand gesture recognition

An assistive tool (InterpreterGlove) for hearing- and speech impaired people is created, enabling them to easily communicate with the non-disabled using hand gestures and sign language. An integrated hardware and software solution is built to improve their standard of living, consisting of sensor network based motion-capture gloves, a low-level signal processing unit and a mobile application for high-level natural language processing. This paper introduces the overall system architecture and describes our automatic sign language interpreter software solution that processes the gesture descriptor stream of the motion-capture gloves, produces understandable text and reads it out as audible speech. The main logic of our automatic sign language interpreter consists of two algorithms: sign descriptor stream segmentation and text auto-correction. The software architecture of this time-sensitive complex application and the semantics of the developed hand gesture descriptor are described. We also present how the beta testers feedback from the deaf community influenced our work and achievements.

Search in WikiImages using mobile phone

Demonstration will focus on the content based retrieval of Wikipedia images (Hungarian version). A mobile application for iOS will be used to gather images and send directly to the crossmodal processing framework. Searching is implemented in a high performance hybrid index tree with total ~500k entries. The hit list is converted to wikipages and ordered by the content based score.

StreamOnTheFly: a network for radio content dissemination

A distributed digital library has been designed and implemented for the support of community radios. This framework, developed by the StreamOnTheFly IST project of the EU, provides a common background for preparation, archival, exchange and reuse of radio programs and supports radio personalization. The architecture is based on a decentralized network of software components using automatic metadata replication in a peer-to-peer manner. This approach combines the principles and practice of OAI (Open Archives Initiative) with the peer-to-peer networking paradigm, and extends usual content dissemination with the aggregation of use statistics and user feedback. The network also promotes social self-organization of the community and a new common metadata schema and content exchange format.