spaCy Not Recognizing Biological Terms with Special Symbols

[valerivankov]

Hi everyone,

I am using spaCy to process texts about human medical and biology data. However, I notice that it is processing symbols and punctuation marks that are part of biomedical term as partial separate entities instead of as a whole term. For instance, 'poly(A)' is separated as 'poly(A' and ')', 'H(2)O(2)' is separated as 'H(2)O(2' and ')', and in 'pre mRNA (U)GCAUG consensus sequences' the '(U)GCAUG' becomes 'U)GCAUG.'

Here are additional examples of this issue:

1. 'CA2+' -> ['Ca2', '+']
2. 'CA(2+)' -> ['Ca(2', '+', ')']
3. 'C18:3(n-3) acyl-CoAs' -> ['C18:3(n-3', ')', 'acyl', '-', 'CoAs']
4. 'tRNA(Thr)' -> ['tRNA(Thr', ')']
5. '7-(a-amino-a-carboxypropyl)wyosine' -> ['7-(a', '-', 'amino', '-', 'a', '-', 'carboxypropyl)wyosine']
6. 'OHyW*' -> ['OHyW', '*']
7. 'CDP-ribitol' -> ['CDP', '-', 'ribitol']
8. 'O-mannosyl trisaccharide' -> ['O', '-', 'mannosyl', 'trisaccharide']
9. 'N-acetylgalactosamine-beta-3-N-acetylglucosamine-beta-4-(phosphate-6-)mannose' -> ['N', '-', 'acetylgalactosamine', '-', 'beta-3', '-', 'N', '-', 'acetylglucosamine', '-', 'beta-4-(phosphate-6-)mannose']

Interestingly, some are with () while others are with - or +.

I understand that spaCy treats the parentheses at the start and end of words as prefixes or suffixes, respectively, which makes sense if these are standalone entities, such as ‘111)’ or ‘[Photosynthesis’. How do we modify spaCy such that it recognizes words that contain symbols and punctuation marks as “whole words” such as the ones above?

Thanks!
cc: @k-blenman