Shashank's MediaBot log

Title: Semi-Automatic, Data-driven construction of Multimedia Ontologies

§ Videos are processed on a centralized server.

§ At the pre-processing stage, three tasks are performed: Scene cut detection, automatic speech recognition, and metadata extraction. In our approach, we can use scene cut detection (manually, using IBM?s VideoAnnex) and metadata extraction (automatically).

§ Textual data (manual annotations, metadata, data from ASR) is processed using KAON, a tool for semi-automatically building ontologies, which has functionality for stop word elimination, word stemming, etc. For each term from text, its frequency, TFIDF score and entropy is computed. Next part is Manual. User manually selects terms to as concepts in ontology and KAON discovers relations within these concepts.

§ Experiments are performed on 24 hours of video from the NIST TREC 2002 Video Benchmark collection. Each video is annotated manually using VideoAnnex.

Title: Crawling the Web

§ Mainly about crawlers and not the mobile agents, but many ideas about indexing and retrieval can be used in our project.

§ Parsing: Section 2.4: HTMLParser implements parsing the HTML, extracting text only part of the HTML and conversion into XML format. It has provision for stop word elimination and word stemming.

§ Page Importance: Section 4.1: Relevance scores of the page.

Title: Converting Web pages into Well-formed XML Documents

§ Purpose is to automatically extract info from web pages and format it correctly.

§ They convert HTML into well-formed XML format.

o Syntactic mapping of HTML into XML.

o Resolving ambiguity introduced by HTML tagging rules

o Handling errors that may occur due to improper coding of HTML.

:: [MediaBot] Summaries of some relevant approaches - Indexer ::
HOME

[MediaBot] Summaries of some relevant approaches - Indexer