Fix output entities

[lfoppiano]

I've got the following problem, the current model:

grobid-material.zip

is used to parse string of materials names, more or less full of garbage and manage to get something useful out of it :-)

In the following example: (Mo -x 1 T x ) 3 Sb 7 with  x 0.1 it extracts three entities

• (Mo -x 1 T x ) 3 Sb 7 as <formula>
• x as <variable> and
• 0.1 as <value>

The current master implementation output 0. as the last entity's text. I've started digging and

a) reconstructed the entity text from the original text, and not from the tokens, the current approach concatenates with spaces in the middle which are not always needed.
b) fixed a problem when the entity is laying at the end of the sequence, somehow the last tag is ignored and the start/end offsets are lacking one element.
c) I've followed @oterrier suggestion in #103 and update the joblib import 😂
d) a858404 it's an old modification I've been committed but never pushed months ago, it removes another warning then the sequence is truncated

This might need some additional tests, to run my test case, here the modification in grobidTagger.py (in combination with the material model):

        elif model == 'material':
            someTexts.append("(Mo -x 1 T x ) 3 Sb 7 with  x 0.1")

[kermitt2]

Thanks Luca