Is there a Special character that represents a word like a city name?

TensorFlow General Discussion
, , ,
1

Hi, i am working on a small AI for sentence recognition. I am curious if there is a special character that will not be trained but is seen as a thing that can be anything.

For instance: In what country lays new york.
new york can be anything so is there a way to say with a word of a special character that means:
new york can be anything
I use yml as the file for all the data. it is structured below

  - command:
    input: In what country lays ...
    entity: place
    action: getCountry

I did do a small test where I trained the model with this sentence:

  - command:
    input: where lays {}?
    entity: place
    action: getPlace

when I tested it it did come out right:

where lays Amsterdam
Text: “where lays amsterdam” is classified as “place\getPlace”

But IDK if this Is because of the first two words of not.

I work in python version 3.8.10
I work with tensorflow 2.7.0

I hope to hear from you.

3

Hello @anonymous023

Thank you for using TensorFlow

In this input input: where lays * , * can be used as the placeholder, so that it could represent any place name and more generalize the training.
If the first two words are influencing the learning process, we can add augmentations in the data to make it more generalized way of learning.