I wasn't sure where else to ask so I'm asking here. Feel free to delete the issue and refer me to proper forum for these kind of queries.
Are there any annotation tools for AMR that formalise resources like the AMR dictionary and guidelines into a single algorithm? It would reduce annotation variability if there was a standard tool that suggested annotations based on the current AMR guidelines. E.g. if you have bead :mod glass it could auto-correct to :consists-of.
I'm interested in semi-automated AMR parsing more than fully automated. Are there any parsers that could underline parts of the text that it's unsure about, allowing you to manually correct them? Like a 'sense-checker'. Then it would learn the correction so that next time a similar sentence appears, it would parse it correctly without input.
Nouns: are they connected to any kind of knowledge base? If I use 'person' in an AMR, it seems that there's no way of uniquely linking it to a concept in an ontology, apart from string search.
Semantic depth: if for example I use the throw.01 frame, are there any resources that state in detail what throwing actually involves? In that, to throw something implies that at some point the object being thrown is not touched by anyone. Like FrameNet, but deeper. Or 'give' implies that the object given is now in the possession of the beneficiary. I'm imagining linking semantic frames to some kind of logical operator that modifies the original state or context. The 'give' action would be something like: possess(personA, item) -> possess(personB, item). They could also have prerequisites, e.g. give(personA, personB, item) would require that possess(personA, item).
Any idea when tense and plurality might be included officially? Also multi-sentence.
Any insight into these questions would be hugely appreciated. I'm not a linguist, I'm just a hobbyist that's fascinated by semantic representations.
I wasn't sure where else to ask so I'm asking here. Feel free to delete the issue and refer me to proper forum for these kind of queries.
Any insight into these questions would be hugely appreciated. I'm not a linguist, I'm just a hobbyist that's fascinated by semantic representations.