Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.486043
Title: The role of terminology and local grammar in video annotation
Author: Al-Athel, Mohammed S.
Awarding Body: University of Surrey
Current Institution: University of Surrey
Date of Award: 2008
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
The linguistic annotation' of video sequences is an intellectually challenging task involving the investigation of how images and words are linked .together, a task that is ultimately financially rewarding in that the eventual automatic retrieval of video (sequences) can be much less time consuming, subjective and expensive than when retrieved manually. Much effort has been focused on automatic or semi-automatic annotation. Computational linguistic methods of video annotation rely on collections of collateral text in the form of keywords and proper nouns. Keywords are often used in a particular order indicating an identifiable pattern which is often limited and can subsequently be used to annotate the portion of a video where such a pattern occurred. Once' the relevant keywords and patterns have been stored, they can then be used to annotate the remainder of the video, excluding all collateral text which does not match the keywords or patterns. A new method of video annotation is presented in this thesis. The method facilitates a) annotation extraction of specialist terms within a corpus of collateral text; b) annotation identification of frequently used linguistic patterns to use in repeating key events within the data-set. The use of the method has led to the development of a system that can automatically assign key words and key patterns to a number of frames that are found in the commentary text approximately contemporaneous to the selected number of frames. The system does not perform video analysis; it only analyses the collateral text. The method is based on corpus linguistics and is mainly frequency based - frequency of occurrence of a key word or key pattern is taken as the basis of its representation. No assumptions are made about the grammatical structure of the language used in the collateral text, neither is a lexica of key words refined. Our system has been designed to annotate videos of football matches in English a!ld Arabic, and also cricket videos in English. The system has also been designed to retrieve annotated clips. The system not only provides a simple search method for annotated clips retrieval, it also provides complex, more advanced search methods.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.486043  DOI: Not available
Share: