- Spontaneous vs. prepared speeches -> how to decipher one from another
- Would be helpful to have this done on a inter/intra-speech level
- Build argument structure for entire speech and then correlate other speeches
- Perhaps some kind of common-element extension to form complete graph
- Build up graph structure for speeches and then check how adversarial they are
- First step could be to check if there are any spontaneous speeches at all, or perhaps spontaneous segments of speeches
- Perhaps could build up from ontologies of existing atoms and meanings in political domains
-
Use microtext corpora to create argumentation structure (https://github.com/peldszus/arg-microtexts & https://github.com/peldszus/arg-microtexts-multilayer)
- This would utilize theoretical argumentation structures; work would have to put in to extrapolate structures from local to global ones
-
Use US election debate corpus to create claims/premises of speeches (https://github.com/ElecDeb60To16/Dataset)