Database search and neural network

Question

MongoDB

Hey!
I am writing my diploma.
A couple of questions have come up:
Is it possible to search the database using a neural network without unloading the entire database into it? That is, to do some kind of indexes, but for a neural network.
How is the network planned?
MongoDB stores documents: cars.
In the search bar, we enter the query "Car for the family"
And the neural, based on experience, will give out the necessary cars, and if there are new cars in the database, it will pick them up if they are suitable and add them to the index.
That is, in fact, we receive a request for a string and respond with a set of JSON objects.
How it is possible to implement similar and what to esteem?
I found a lot of literature and theory, but I can’t connect with practice ..
Maybe there are examples of source codes?

Answer 1

In your case, the neural network is ideal for clustering the set of all cars in the database into certain groups, which are ideal for "family", "youth", "retirees", etc.
Those. At the input of such a neural network, you can submit the parameters of the car, and at the output, a set of weights characterizing how well this car fits the categories listed above. In this way, you can train the NN to choose the best options for specific tasks.
The definition of the task that the user needs is either "the best car for the family", "the best car for young people", i.e. the definition of the category for which you want to get a sample can be done based on probability theory, for example, look at the naive bayes classifier. NN is not very suitable for linguistics (unless you convert the original query into a set of some numerical parameters and feed it to the NN).

Answer 2

The neural network will not help here and looks superfluous. You need to generate query parameters based on the request, that is, even before the sampling stage from mongo.