Negex Negation Detection for German Clinical Text

German NegEx Trigger Set: Our trigger set is described in our BioTxtM 2016 paper (see paper below). It can be used in combination with NegEx in order to detect negations and speculations in German clinical data. The trigger set was created by a German native speaker, based on the trigger set multilingual_lexicon-en-de-fr-sv.csv which can be found on this webpage (see medinfo_2013_multilingual_negex_lexicon_v1_April30th2013.zip).

The following steps were performed:

• Source-original ConText terms have been selected.
• Translations and alternative translations have been reviewed and in some cases corrected.
• New alternative translations have been added.
• Regular expressions have been expanded.
• The scope (of negations) has been determined, taking into account GE Action, Category and our own thoughts. In many cases "bracketing" triggers, such as "lehnt ... ab" (reject) and "wies ... zurück" (declined) were shortened and only partially taken into account.

Download: negex_trigger_german_biotxtm_2016.txt

NOTE: In order to use the trigger set, the following two changes need to be applied to negex.python.zip (Jul 7, 2009). Replace line:

pattern = r'\b(' + trig + r')\b'

with:

if trig == "\?":
   pattern = r'(' + trig + r')'
else:
   pattern = r'\b(' + trig + r')\b'

and line:

sb3 = sb3 + ' ' + sentenceTokens[i]

with:

sb3.append(sentenceTokens[i])

We would appreciate if you cited our work :-)

Negation Detection in Clinical Reports Written in German, Viviana Cotik, Roland Roller, Feiyu Xu, Hans Uszkoreit, Klemens Budde and Danilo Schmidt, In Proceedings of the 5th Workshop on Building and Evaluating Resources for Biomedical Text Mining (BioTxtM), 2016, Osaka, Japan [PDF][BibTex]