simple NLP (natural language processing) Task

The task is to create simple chat bot based on the similar text search using one of the libraries like annoy ([login to view URL] ) or faiss or some other for the approximate nearest neighbours indexing. The steps to implement something that sort of works would be: 1. Collect some dialog dataset or some text document (e.x. project gutenberg would be a good source for the former as it is cleaner than wikipedia but feel free to choose whatever you wants.) 2. Index all or part of the text using vector representation of the sentences. Vector representation can be calculated with the spacy library ([login to view URL]) or any other technique like tf-idf or whatever one wants. 3. Use user prompt as the search vector (after representing it as vector using Spacy or any other library/model) and retrieve e.x. closest, not exactly the same, neighbor 4. When the data set that is being used is a dialog data set, use the return to the nearest neighbour (the prompt from the second dialog member). When the dataset is a book or a wikipedia, use the nearest neighbor as the answer.