Notes from discussion held in Tomar on 2014-07-15.

Dan: the inventory of labels or elements that can be included in the SAS line is the syntactic architecture, do you have a sense how many such levels are envisioned in a successful multilingual resource? The easy ones are there, but how about tough-adjectives

It seems to me that as soon as interesting phenomena are included, particles, touch constructions, clitics, what the grammarian cares about, how to deal with that?

Why should this work?

Petya: my impressions (I might be wrong), I just picked very easy examples, there is more. They have additional annotations. I don't know about justifications, but there are… The idea is to start with specific languages and they will adapt it as more languages join. This is why I'm exited to include Bulgarian, even if the grammar is small.

Discriminants cannot unify everything. I had a long version and a short version: I only worked with the short version, because the long version was very specific and it was hard to map the types.

Emily: what is the value of the description if the valence itself has more information than will ever be available in the description.

Petya: it is for comparison and typological generalisations.

Tim: how to capture idiosyncrasies in the language itself?

Petya: the functional labels can be made very rich. Lars and Montse noticed the difficulty that is why there is a long and short list. The solution is to add functional types.

Tim: but how can you capture similar principles in opposite direction (one language has most optional , except for some arguments the other the other way around): you want to be able to discuss those kind of things.

Dan: impression goal of effort is extremely ambitious. A pan linguistic organisational scheme: it is worthwhile, but imposes an increasing burden on grammarian: on top of the difficulty of mastering their own language, they also have to link it to all languages (and understand that). It might fit in the Matrix ambitions, but it is a different granularity. Why do you need to know all possibilities in the field?

Petya: I remember these contra-arguments from last year. It should go more general at the cost of losing language specific information, this happened with the three languages with the long and short list and it will become more realistic. I do think it will become a good framework to compare languages. It won't be easy, but even a simplified version will provide a picture.