Create a search engine based on chat bot with machine learning

Hello, I need to create a search engine internal to my website (under construction) based on natural language processing NLP with python preferably and able to interface with PHP and MYSQL for the database. create a HTML (blank with nothing don't mind) with an input for the search -> create 1 or several specific table in a database (note that it will be in a SaaS environment so with 1 database for each tenant) : -> one for Q/A -> one for documents Search page will be divided in several parts separated by a line in results : -> Answer -> Training -> Document -> People -> Users can ask for question or ask information : -> Question shall be treated with NLP (token, synonym, cleaning, stem etc...) -> if question (marked by a ? at the end of the sentence, otherwise don't display the part answer ), script should look within a Q/A table (to be created) and establish a confidence rate to decide which one to take and display or not. obviously must understand the question and be based on NLP for synonyms and narrow questions/answers -> if not the direct answer from the table or confidence rate too low, script should be interfaced with BING API or GOOGLE API for question based on snippet. when there is very accurate question and script return the answer (exact answer not document or link) If nothing, just return a link with a modal opening to enter the answer into the Q/A table the part answer = see above for the returning of answer the part for training = it go searching in a table (create one for the testing) with in row name of training and in column keywords and sum up. Must be able to look in with NLP and find the most relevant training with the question based on keywords and their weight compare to the question. Must propose max 5 trainings sort by most relevant Samething for document with a specific table Samething with people with a specific table. the return must be in each part, the name - sum up - confidence rate in % with link to open table must be optimized and indexed and organized in a smart way, particularly to avoid useless search Machine learning process must be integrated to get more accurate results based on question posed and propose better answers/training/documents. must stay fast and light as possible. Shall work for English and French