I'm pursuing AI training in Chennai and want to list the words that match the abbreviation. For example, if I'm searching for AI or learn artificial intelligence then it should list for any record that contains 'Artificial Intelligence' as a part of the sentence. Similarly when I search for MCA or Master of Computer Application then it would list all the records which contain Master of Computer Application as a part of a sentence.

Assuming that my MySQL table is 'topics' with columns TopicID, TopicName, TopicDesc. Sample data: http://sqlfiddle.com/#!9/77f8df

Note: I was asked to achieve it without an additional table or column.

Please let me know if more information is required.

Consider the following example "10% of on all Artificial Intelligence courses online." In this example, I have to extract two predefined classes like Artificial Intelligence and courses. Even the program has to classify words like ANN, CNN, RNN, AI, etc. into the Artificial Intelligence category. I have used spacy to train but I am not impressed with the results as it is not labeling correctly. Is there any alternative to extract entities from a sentence in Python?

I'm a trainer from AI Training institute in Bangalore and have data to analyse

Input data - is an ad from a channel or group of a social network. For example, this: "A room for rent in a two-room apartment (preferably a married couple), the Verkhniye Likhobory metro station is a 3-minute walk. You can move in from September 21 The room is large, with a balcony, the apartment has everything. A family lives in another room. Details by p. 8922five682461."

There is no clear address in this ad, but the address itself is present - the Verkhniye Likhobory metro station.
There were also mistakes in the ad:
"tqo-room apartment" instead "two-room apartment",
"with a balcony" instead of "with a balcony".
Also, the ad used to bypass the prohibition of some groups to publish phone numbers: 8922five682461. And I think this ad is useful in terms of content because it can lead to conversion, that is, monetization can be done with such data. This ad may be in demand, although a lot of data is lacking.
And, if a human parses the data from the ad, then the human will extract the following structure from the ad:

The system must be able to extract data from the text and distribute it among the fields of the structure - even when some of the data does not correspond to the standard.

What should be the architecture and configuration (weights, and so on) of a neural network to recognize:

phone numbers,
E-Mail addresses,
geographic addresses,
images,
the costs from the texts of ads in social networks, taking into account the errors of manual text input and different formats of the required data?
And also - which platform from (TensorFlow, Keras, PyTorch, DeepCognition, something else) would you recommend to use to solve this problem? Where is the more competent and flexible architecture? How much will the choice of programming language for such a task influence? For example, in C++ with CUDA there is a possibility of acceleration on video processors. Are there analogs in Python? Or Python has rich libraries for working with NLP. Are there any analogs in C++?