Video text retrieval involves extracting relevant videos to a given textual query and vice-versa. Usually this is done as a ranking task, where all the entries are ranked in relevance to the query. The retrieval can also be done across languages due to multi-lingual nature of dataset

Textual summarisation involves outlining the key actions happening in the video in the form of meaningful sentences. These textual descriptions are then evaluated using BLEU, Meteor metrics. The same task can also be extended to cross lingual summarisation.

Video recommendation usually works by mining users' profiles and history and after that by ranking videos according to their preferences and viewing history. In absence of such data, one can recommend similar videos just based on content of video with no-textual metadata of any form. In the same way as earlier, this can also be enhanced to include multi-lingual video recommendation.