PHP library for fuzzy search?

Question

I am looking for a PHP library that will allow users to basically enter commands to run using real English. Basically I want the user to be able to do something like:

Please search for all users in Europe , which will be equal to a select * where users = 'Europe'

Or another example of my intentions:

Lookup (Find) the email address for John Smith

Note. It would be nice if you could say for John Smith, Jane Smith, and John Doe

Ideally, I need this library, if it exists, to be scalable, so I can program in what should happen when certain things appear, for example, find,email or search,users

Does anyone know of a PHP library that can do something like the above?

Answer 1

As far as I know, there are currently no libraries to search for queries in natural language, neither in PHP, nor in any other programming language (I suppose you cannot use IBM Watson :).

I think possible approaches are a parser and fuzzy search :

Using a parser generator, such as Jison , you can analyze and "understand" in the user's browser all the statements that correspond to the generating grammar, sending only the generated query or intermediate representation to the server.

This is better than the PHP parser because the user can get immediate feedback when entering text and is less disappointing than submitting a form and receiving an error message. Interpretation of the request in this case will be 99% correct, but in many cases the perfectly correct (from a human point of view) request will be rejected because it will not include a grammar.

In another case, you can do some pre-processing, for example, deleting stop words, creating lowercase lines, creating lines, etc., then searching using a full-text search engine (Lucene is probably the most powerful, but in Java). PostgreSQL supports it , and MySQL also has some full text search capabilities. It is also possible to build a primitive engine based on a basic DBMS using index and tokenizing text on spaces and punctuation.

Which method depends on how varied and noisy your data and various expected requests are. You can also try to implement a hybrid approach, that is, analyze the text using the grammar, and if you could not use the full text search.