A Survey on Multimodal Summarization using Deep Learning

Abstract

In the present situation of fast-growing consumption of the Internet, searching for the required information becomes monotonous and sluggish task. Apart from finding relevant content from given results there is another very difficult task for a user, which is to manually summarize the large multimodal data. Now-a- days there is a rapid growth of multimodal data. Summarization is the process of truncating the enormous amount of data available in various formats into summarized documents, conserving the most important points and the overall meaning of the document. The process of summarization is divided into extractive summarization and abstractive summarization. Extraction is the process which concatenates sentences from the document depending on their importance, whereas abstraction involves generating novel sentences from information extracted from corpus. In this survey paper we investigate the popular and important work done in the field of multimodal summarization, various summarization methods that can create a compacted version of a set of documents and audios related to a specific topic without any human control.