Latent Semantic Indexing is derived from Latent Semantic Analysis. It goes through the content to find related words. Like if the content is about dogs, then there should be other related words in it like puppies, kennel, bark, tail, etc. LSI is supposed to make stuffing keywords randomly meaningless.
The mathematical technique used for indexing or retrieval is latent semantic indexing. In which, content are crawled by the search engine and it looks for the title of the page.
Latent Semantic Indexing is search engine algorithm to determine what a page is about outside of specifically matching search query text. LSI return relevant results that don't contain keyword.
Search engine expect that your page conent must related with your title. How much you can define your title in your page content, Search engine identified that factor and display your webpage in SERP for relevent query. If your page title is about Ice Bucket Challenge then you should define the challenge in brief and write about its process and its purpose etc...
LSI explicitly represents terms and documents in a rich, high-dimensional space, allowing the underlying ("latent"), semantic relationships between terms and documents to be exploited during searching.