Support for users of CCG software closed 7-27-20

[Xuchen Yao <xuchen AT cs.jhu.edu>]

Hi,

Thanks for providing the Named Entity Tagger tool. I've been using both 
the server (NerServer) and batch (NerTagger) version on my local machine 
and am quite confused by the encoding. Suppose I need to tag the 
following sentence:

The studio  changed hands   several times   after   1953    and was 
home    to  such    early   television  series  as  ''  Superman ''  and 
''  The Red Skelton Show    .

When running NerServer, and setting taggingEncodingScheme to BILOU, I 
have (by printing out String[] predictions in tagData() in NETagPlain.java):

O   O   O   O   O   O   O   B-DATE  O   O   O   O   O   O   O   O O   
O   O   O   O   O   O   I-WORK_OF_ART   I-WORK_OF_ART I-WORK_OF_ART   O
The studio changed hands several times after [DATE 1953  ] and was home 
to such early television series as '' Superman '' and '' The 
[WORK_OF_ART Red Skelton Show  ] .

When running NerServer, and setting taggingEncodingScheme to BIO, I have 
instead:

O   O   O   O   O   O   O   U-DATE  O   O   O   O   O   O   O   O O   
O   O   O   O   O   O   I-WORK_OF_ART   I-WORK_OF_ART L-WORK_OF_ART   O
The studio changed hands several times after 1953 and was home to such 
early television series as '' Superman '' and '' The [WORK_OF_ART Red 
Skelton Show  ] .

Here are three things confusing me. May I have some clarification please:

1. The BIO setting is actually outputting the encoding scheme of BILOU, 
and vice versa. I doubled checked my setting and am pretty sure I didn't 
have the setting wrongly swapped.

2. In both output, I-WORK_OF_ART is not preceded by a B-WORK_OF_ART. 
Note that if you run the batch version (NerTagger) to tag this sentence, 
you will get B-WORK_OF_ART in front of I-WORK_OF_ART, but there are 
other sentences where the batch version would output an "I-*" tag right 
after an "O" tag.

3. Here's the NerTagger version of the same input, with BILOU 
taggingEncodingScheme:
O   O   O   O   O   O   O   B-DATE  O   O   O   O   O   O   O   O O   
O   O   O   O   O   B-WORK_OF_ART   I-WORK_OF_ART I-WORK_OF_ART   
I-WORK_OF_ART   O

Note that it correctly outputs "B" in front of "I". However, my run of 
NerTagger and NerServer used the same config file 
(IllinoisNerExtended-v2.1/DemoConfig/ner.ontonotes.config), why the 
difference? Is there any suggestion on the configuration in the batch 
vs. server settings?

Thanks.

Xuchen Yao