Trade mark similarity assessment support system

Mohd Anuar, Fatahiyah 2014. Trade mark similarity assessment support system. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (10MB) | Preview
	PDF - Supplemental Material Restricted to Repository staff only Download (77kB)

Abstract

Trade marks are valuable intangible intellectual property (IP) assets with potentially high reputational value that can be protected. Similarity between trade marks may potentially lead to infringement. That similarity is normally assessed based on the visual, conceptual and phonetic aspects of the trade marks in question. Hence, this thesis addresses this issue by proposing a trade mark similarity assessment support system that uses the three main aspects of trade mark similarity as a mechanism to avoid future infringement. A conceptual model of the proposed trade mark similarity assessment support system is first proposed and developed based on the similarity assessment criteria outlined in a trade mark manual. The proposed model is the first contribution of this study, and it consists of visual, conceptual, phonetic and inference engine modules. The second contribution of this work is an algorithm that compares trade marks based on their visual similarity. The algorithm performs a similarity assessment using content-based image retrieval (CBIR) technology and an integrated visual descriptor derived using the low-level image feature, i.e. the shape feature. The performance of the algorithm is then assessed using information retrieval based measures. The obtained result demonstrates better retrieval performance in comparison to the state of the art algorithm. The conceptual aspect of trade mark similarity is then examined and analysed using a proposed algorithm that employs semantic technology in the conceptual module. This contribution enables the computation of the conceptual similarity between trade marks, with the utilisation of an external knowledge source in the form of a lexical ontology, together with natural language processing and set similarity theory. The proposed algorithm is evaluated using both information VI retrieval and human collective opinion measures. The retrieval result produced by the proposed algorithm outperforms the traditional string similarity comparison algorithm in both measures. The phonetic module examines the phonetic similarity of trade marks using another proposed algorithm that utilises phoneme analysis. This algorithm employs phonological features, which are extracted based on human speech articulation. In addition, the algorithm also provides a mechanism to compare the phonetic aspect of trade marks with typographic characters. The proposed algorithm is the fourth contribution of this study. It is evaluated using an information retrieval based measure. The result shows better retrieval performance in comparison to the traditional string similarity algorithm. The final contribution of this study is a methodology to aggregate the overall similarity score between trade marks. It is motivated by the understanding that trade mark similarity should be assessed holistically; that is, the visual, conceptual and phonetic aspects should be considered together. The proposed method is developed in the inference engine module; it utilises fuzzy logic for the inference process. A set of fuzzy rules, which consists of several membership functions, is also derived in this study based on the trade mark manual and a collection of trade mark disputed cases is analysed. The method is then evaluated using both information retrieval and human collective opinion. The proposed method improves the retrieval accuracy and the experiment also proves that the aggregated similarity score correlates well with the score produced from human collective opinion. The evaluations performed in the course of this study employ the following datasets: the MPEG-7 shape dataset, the MPEG-7 trade marks dataset, a collection of 1400 trade marks from real trade mark dispute cases, and a collection of 378,943 company names.

Item Type:	Thesis (PhD)
Status:	Unpublished
Schools:	Engineering
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > TA Engineering (General). Civil engineering (General)
Uncontrolled Keywords:	Trade mark retrieval; Trade mark similarity; Semantic similarity; Phonetic similarity; Visual similarity; Content - based retrieval
Date of First Compliant Deposit:	30 March 2016
Last Modified:	27 Oct 2023 14:14
URI:	https://orca.cardiff.ac.uk/id/eprint/73369

Actions (repository staff only)

Edit Item