[Feature Request] Exception List

[MG2033]

Hello,

I have got an idea. If an exception list or whitelist is implemented such that the word sequence is not split, it'll be very good. For example, "tensorflow" is split into "tensor" and "flow". So, we can provide entities that we don't want them to be split and after the algorithm runs, they remain the same.

[sumba101]

this would be a great feature addition

[timminator]

For people stumbling across this in the future - I made a fork where this can be achieved now. When initializing the LanguageModel you can now optionally specify a list of words you wan to add to the dictionary. If it is being split currently, this almost always means, that it is not currently in the dictionary. You can find my fork here:
https://github.com/timminator/wordninja-enhanced

Edit:
You can do it like this:

import wordninja_enhanced as wordninja
joined_text = wordninja.rejoin("Palaeoloxodonisanextinctgenusofelephant.")
print(f"Before adding the word: {joined_text}")

custom_lm = wordninja.LanguageModel(language="en",
    add_words=['Palaeoloxodon'],
    blacklist=[],
)

joined_text = custom_lm.rejoin("Palaeoloxodonisanextinctgenusofelephant.")
print(f"After adding the word: {joined_text}")

Output:

Before adding the word: Palaeo lox odon is an extinct genus of elephant.
After adding the word: Palaeoloxodon is an extinct genus of elephant.