allow user to define unknown token symbol

[congchan]

Description

Add new feature for issue #10068. It allows unknown token to be added to vocab if user provides a vocabulary and specifies a symbol(e.g. 'UNK'). Along with new default behaviour as ignoring the unknown token, instead of the present way which throwing an error.

Checklist

Essentials

Please feel free to remove inapplicable items for your PR.

• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage:
• Unit tests are added for small changes to verify correctness (e.g. adding a new operator)
• Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore)
• Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL)
• Code is well-documented:
• For user-facing API changes, API doc string has been updated.
• For new C++ functions in header files, their functionalities and arguments are documented.
• For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable
• Check the API doc at http:https://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$PR_ID/$BUILD_ID/index.html
• To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change

Changes

• Feature1, allow unknown token to be added to vocab if user provides a vocabulary and specifies a symbol(e.g. 'UNK');
• Feature 2, new default behaviour is ignoring the unknown token, instead of throwing an error.

Comments

• If this change is a backward incompatible change, why must this change be made.
• Interesting edge cases to note here

[piiswrong]

why?

[congchan]

Hi, there is situation where users has their own dictionary, say dict = {'a':1, 'b':2, 'c':3} 'abc' are frequent tokens the user care about. All the rest rare tokens are considered as unknown token (say the user define it as 'UNK'), that return a encoded list [[1,2,3],[2,3,0]], a key-value pair 'UNK': 0 is added into the dictionary.

But the previous version will raise error for this case, which by default assuming that user will provdie a thorough vocaburary.

[piiswrong]

you should change the assertion to ignore cases where unkown_token is give instead of changing new_vocab to true

[szha]

Ping

[congchan]

Sorry for late reply (in exam). Just fix according to your suggestion.

[ThomasDelteil]

@ShootingSpace thanks for adding a test