How to find information in a text by a pattern based on neural networks?

Question

Machine learning

Greetings! :)
Let's say we have a text of not very large volume, about 10-15 sentences. It needs to find information about something according to the specified pattern. For example, there is a template "PLACE AND TIME" and a message like this:

"Hey. I won't be able to come now. arrested at work. Then you will need to go home. Only then will I be free. Let's meet at the subway at 15:00. Can you get to that time? »

The answer may be:
"metro at 15:00" .
Having tried to find articles, I did not find a lot of clear information, and you need to do all this on Android. Can you advise me: How to implement this idea? What should be used for this? What libraries?

Answer 1

https://tech.yandex.ru/tomita/

Answer 2

Compile dictionaries = gazetteers of values ​​of certain types (place, time, etc.), tokenize the text using them, then regular expressions. It is convenient to replace the numbers with one character, for example 0. So that 12:34 and 15:56 look the same as 00:00.
If the spelling options are quite diverse or vague, then it is possible not with regular expressions, but with the help of neural networks for NER tasks (the standard is BiLSTM + CRF)