Sentiment Inference: Pro and Contra relation dataset

500 German sentences annotated for pro/con relations and polar roles of entities (negative/positive actors/effects): see References for a conceptual introduction.

files: annotator1.conll .. annotator3.conll

format: conll (parzu parser) with annotations

- annotations at the end of the conll parse tree
  - c = con
  - p = pro
  - neff,peff = negative, positive effect
  - nac, pac = negative, positive actor
- the head indices are used for annotation (see below)
  - c1,6 = Hofstetter con Gewerkschaften
  - neff6 = negative Effekt on Gewerkschaften

Note: in these annotations, pro/con is not an intentional relation

- in "Snow blocks the driveway" it holds: con(snow,driveway)
- "snow" is a negative element wrt. to driveway
- use our animacy classifier to identify those case with an actor (see References lrec, available via IGGSA download)

Example:
1    Hofstetter    Hofstetter    N    NE    _|Nom|Sg    2    subj    _    _    
2    wirft    werfen    V    VVFIN    3|Sg|Pres|Ind    0    root    _    _    
3    im    in    PREP    APPRART    Dat    2    pp    _    _    
4    Interview    Interview    N    NN    Neut|Dat|Sg    3    pn    _    _    
5    den    die    ART    ART    Def|Fem|Dat|Pl    6    det    _    _    
6    Gewerkschaften    Gewerkschaft    N    NN    Fem|Dat|Pl    2    objd    _    _    
7    vor    vor    PTKVZ    PTKVZ    _    2    avz    _    _    
8    ,    ,    $,    $,    _    0    root    _    _    
9    sie    sie    PRO    PPER    3|Pl|_|Nom    10    subj    _    _    
10    wollen    wollen    V    VMFIN    3|Pl|Pres|_    2    s    _    _    
11    die    die    ART    ART    Def|Fem|_|Sg    12    det    _    _    
12    Branche    Branche    N    NN    Fem|_|Sg    13    obja    _    _    
13    anschwärzen    anschwärzen    V    VVINF    _    10    aux    _    _    
14    .    .    $.    $.    _    0    root    _    _    
c1,6
p1,12
neff6

References:

@inproceedings{stance,
       booktitle = {LSDSem 2017/LSD-Sem Linking Models of Lexical, Sentential and Discourse-level Semantics},
           month = {April},
           title = {Stance Detection in Facebook Posts of a German Right-wing Party},
          author = {Manfred Klenner and Don Tuggener and Simon Clematide},
       publisher = {ResearchBib},
            year = {2017},
        language = {english},
             url = {https://doi.org/10.5167/uzh-136567}
}
@inproceedings{perspectives,
       booktitle = {18th International Conference on Computational Linguistics and Intelligent Text Processing},
           month = {April},
           title = {Verb-mediated Composition of Attitude Relations Comprising Reader and Writer Perspective},
          author = {Manfred Klenner and Simon Clematide and Don Tuggener},
       publisher = {ResearchBib},
            year = {2017},
        language = {english},
             url = {https://doi.org/10.5167/uzh-136569},
             doi = {10.1007/978-3-319-77116-8\_11}
}
@inproceedings{harmonization,
       booktitle = {Proceedings of the 5th Swiss Text Analytics Conference (SwissText) \& 16th Conference on Natural Language Processing (KONVENS)},
          editor = {Sarah Ebling and Don Tuggener and Manuela H{\"u}rlimann and Martin Volk},
           month = {Juni 2020},
           title = {Harmonization Sometimes Harms},
          author = {Manfred Klenner and Anne G{\"o}hring and Michael Amsler},
          publisher = {Virtual Event}
            year = {2020},
        language = {english},
             url = {https://doi.org/10.5167/uzh-197961}
}
@inproceedings{lrec,
           month = {Juni},
          author = {Manfred Klenner and Anne G{\"o}hring},
       booktitle = {Proceedings of the Language Resources and Evaluation Conference},
         address = {Marseille, France},
           title = {Animacy Denoting {G}erman Nouns: Annotation and Classification},
       publisher = {European Language Resources Association},
           pages = {1360--1364},
            year = {2022},
        language = {english},
             url = {https://doi.org/10.5167/uzh-219148},
        abstract = {In this paper, we introduce a gold standard for animacy detection comprising almost 14,500 German nouns that might be used to denote either animate entities or non-animate entities. We present inter-annotator agreement of our crowd-sourced seed annotations (9,000 nouns) and discuss the results of machine learning models applied to this data.}
}