What you are querying borders semantic search. That is you are not looking for the exact phrase but for something that semantically means the same thing. Unfortunately this is not something elasticsearch can do for you out of the box.

Traditional search engines work by matching the exact words of your query and looking for its occurrences in the document. They do this by creating an inverted index, which is nothing but a lookup table of all the words/tokens present in a document. They do this for all documents you want to index. Then when a query comes in they use some similarity algorithm to evaluate the contents of all the indexed documents against the words/tokens present in the query. They then return the documents in a ranked order from most similar to least based on the score. The semantics or the “meaning” of the text is not considered.

If your query has some overlapping words with the document you are looking for, then sure you will get some relevant documents back. But if there is NO words that are same then this will not work. For example you cant expect it to return a document with the phrase “the computer is not working “for the query “the pc is broken”

What you are looking for is semantic search. There are some pre-trained language models on HuggingFace whose embeddings can be used as search index. And there is an open-source FAISS library by Facebook that allows you to search it. But the specifics of this is highly dependant on your use case. Also the implementation is a bit more complex and will requires some coding expertise and understanding of ML. You will need professional help unless you already know the stuff or are willing to learn

I have experience doing this before. But to be honest this is a time consuming task and not something to be done for free ;)