Classifies Coherent Propositional Contents of locution Pairs as either RA (Support), CA (Attack), or N/A Argument Relationship. N/A consists of instances where there are other and no argument relationships present in the propositional content pairs.
Takes two coherent propositions and classifies as either Support, Attack, or N/A.
Results:
| precision | recall | f1-score | support | |
|---|---|---|---|---|
| CA | 0.38 | 0.27 | 0.32 | 22 |
| N/A | 0.79 | 0.85 | 0.82 | 284 |
| RA | 0.58 | 0.50 | 0.54 | 115 |
| accuracy | 0.72 | 421 | ||
| macro avg | 0.58 | 0.54 | 0.56 | 421 |
| weighted avg | 0.71 | 0.72 | 0.72 | 421 |
Dataset: https://corpora.aifdb.org/mm123 The data preprocessing and fine-tuning technique can be found here: https://discovery.dundee.ac.uk/en/studentTheses/exploiting-illocutionary-forces-in-dialogue-structures-for-enhanc/
- Downloads last month
- 2