On Objective Keywords Extraction: Tf-Idf based Forward Words Pruning Algorithm for Keywords Extraction on YouTube

Abstract

Discovery and subsequent effective retrieval of useful user generated content depends on proper meta-data annotation implemented on an object such as a title and Keywords. In this study, a simpler unsupervised non graph-based algorithm for extracting keywords is proposed. A novel key phrases chunking approach was adopted; this utilizes words sequences as they appear in the original document. The simple but effective Term frequency-inverse document frequency (tf-idf) weighting scheme was exploited to rank the novelty created keyphrases. Comparing to a similar algorithm that uses three metrics weighting scheme, the tfidf yielded a precision of 89%.Thus, the application of tf-idf algorithm on YouTube's metadata based keywords shows to be useful approach in its objectivity. 105 our proposed automatic keyword extraction approach and the evidence based discourse of the UGC creator"s subjectivity in tagging. Our keywords extraction method is an approach towards solving that bias.

Exploiting YouTube's Video ASR scripts to Extend Educational Videos Textual Representative Tags Based on Gibb's Sampling Technique

“…In addition, the video metadata as described in figure 1 were extracted together with the hidden official keywords using a Python based API. A novel keyphrasing algorithm [17] was used on the three textual sources to extend the number of textual representations for each video content to nine; namely the three originals, their three keyphrases representation and the official keyphrases from YouTube resource. The extra two sources were obtained by taking unions of the ASR keywords with metadata and the official keywords for comparison purposes.…”