It would be nice to have a multiword expression analysed by Manfred and Shuly, then by Dan in HPSG, then parsed by Joakim and then represented in a Treebank by Pawel and the Prague team. It would be a way of seeing all want we learnt together in a great summary.

—Carla Parra Escartin (Spanish)

Use at least one example per language (from the ones given by each participant) to analyse them according to the formalisms presented during the trainin sessions.

—Corina Forăscu (Romanian)

Disambiguation: how to distinguish between literal and figurative meanings of MWEs

—Uxoa Inurrieta (Basque)

It is clear that MWE lexicon is an usefull rresource, however designing such resource requires lots of consideration, especially, how to collect variants of single MWE and how to distinct it from other MWEs.

—Lauma Pretkalnina (Latvian)

The more frozen an expression gets, the more it tends to be treated as one "word" written either with hyphens or without spaces. What will happen to inflectable elements of such a MWE? For Hebrew, will particles then be added at the very end only? Can MWE be written as expression without spaces?

—Cerstin Mahlow (German)

If there have been experiments made, which annotation of MWE is better: as single units or as analysed string?

—Verginica Mititelu (Romanian)

What would be a good criterion for fine-grained scoring of MWEs along a range of semantic compositionality/non-compositionality (or transparency as well), and could that metric be correlated to the degree of syntactic flexibility they exhibit (e.g. maybe the more transparent/compositional idioms enter more freely into a greater number of grammatical constructions)?

—Alexander Popov (Bulgarian)

Why do MWEs only exist in some languages?

—Yan Shao (Chinese, Swedish )

How could negation affect the MWEs?

—Noa Cruz Diaz (Spanish)

Try to get translations for each example offered by participants (hence to have a small parallel collection of MWEs)

—Corina Forăscu (Romanian)

Representing structural preferences (e.g. word order preferences) of MWEs in lexicons/grammars

—Jakub Waszczuk (Polish)

conjunction modification of idioms (cf Ernst 1981)

—Sascha Bargmann (German)

verb valency lexicons for non-fixed idioms (e.g. kick [AG person, IDIOM "the bucket"]: how to build one, how to encode information about diffrent kinds of fixedness

—Zuzana Neverilova (Czech)

MWE that can be separated by other words, partially morphologically modified and modified by replacing some words with their synonyms

—Anisya Katinskaya (Russian)

We focused a lot in monolingual MWEs, what about going bilingual or multilingual? What would be a good practice to encode them?

—Carla Parra Escartin (Spanish)

What research projects are you involved inn (or plan to be involved) and how can the audience participate in these activities?

—Corina Forăscu (Romanian)

Handling structuraly idiosyncratic MWEs (e.g. atypical zero copula) in lexicon/grammar engineering

—Jakub Waszczuk (Polish)

The issue for a panel discussion is related to the inflectional and agglutinativity of Turkish. In Turkish, each of the prepositions and the ablative,locative,dative,accusative and genitive cases,and possessives are represented by a different suffix. And each of auxiliary verbs,tenses etc. is represented by a different suffix that is added to the verb in the sentence. When we want to decide if a group of words is a multi word expression or not, we have to look up the word group in our lexicon by matching the lemmas of the group because the surface forms would probably different from the form of the multi word expression in the lexicon. For example; " kafası bozulmak " (this is the form in the lexicon ) means that "get angry".(The lemma form is "kafa- boz-".) but in a sentence it can be :
"Kafam bozuldu." means ( I got angry ). The lemma form is "kafa- boz-".
"Kafamız bozulur." means ( We get angry ) The lemma form is "kafa- boz- ".
"Kafası bozulmuş." means ( He /She got angry) The lemma form is "kafa- boz- ".
By this way, we can find the MWEs that matches in the lexicon.But, in some cases, the lemma-matching comparison results in accepting word groups that are not really MWEs. For instance, (1) "Aradığımı(what I am searching for) buldum(I found)." means (I found what I am searching for.) The lemma form is "ara- bul-". ( it is not a MWE). (2) "arasını bulmak " means ( to make two persons agree or meet or come together) . The lemma form is "ara- bul-". (it is MWE that exists in the lexicon).
The lemma forms of the two word groups are the same, but the first word group is accepted as MWE although it is really not a MWE.The issue is how to prevent from this problem??

—Kübra Adali (Turkish)

The verbal part of (decompositional) verbal MWEs: head or "dependent"? I.e. in the lexical representation of "have one's head in the clouds", "let the cat out of the bag" etc.: is the meaning of the expression "licenced" by the verb, or by a non-verbal idiom "core" taking verbal collocate(s)?

—Gyri Smørdal Losnegaard (Norwegian)

Are "green tea" and "red carpet" multiword expression? If yes, are they non-compositional? If they are not (non-compositional), what (kind of idiosyncrasy) makes them MWE?

—Meghdad Farahmand (Persian, but my question is about English MEWs)

feedback from trainees

—Veronika Vincze (Hungarian)

This is related to Shuly's lecture: I agree that many MWEs are translated as only one word into another language (this is in fact a helpful feature when extracting them using parallel, word-aligned text cf. Villada Moiron and Tiedemann 2006), However, I have no clue how many MWEs are present across languages, i.e. use similar (or even the same) words across languages to express (the same or maybe a different) idiomatic meaning. An example is Shuly's "fell on his head" —> lost his mind example which exists in German but only negated and then has a different meaning ("jemand ist NICHT auf den Kopf gefallen" - so. did not fall on his head —> so being clever). I wonder whether anyone within PARSEME is working or interested on/in such cross-lingual phenomena

—Fabienne Cap (German)

Derivative prepositions in Russian, how to distinguish them from prepositions with nouns.

—Anisya Katinskaya (Russian)

the use of genitive-dative in Mwe's

—Alexandra Fiotaki (Greek)

The use of genitive in Greek MWEs

—Katerina Tzortzi (Greek)

classification of MWEs

—Aleksandar Petrovski (macedonian)

The Sailer-Wintner def of MWEs was: 1< words which are somehow idiosyncratic. Let's extend this to single words as well. Thus we could investigate idiosyncratic behaviour of morphemes. Hungarian example: kényelem = convenience, kényelem+tlAn = inconvenient, szabály = rule, szabály+tlAn = irregular, BUT: egy = one, egy+tlAn = (only) one. We could even call such thing a multi-morpheme expression (MME).

—Bálint Sass (Hungarian)

Some MWEs might also occur in their literal meaning. To take up the "jmd ist NICHT auf den Kopf gefallen" example where the idiomatic meaning is "s.o. being clever" and the literal meaning is "someone did NOT fall on his head". Depending on the context, this construction might (at least in German) be used in its literal reading, Imagine children playing and hurting themselves, an event where s.o. later might report "luckily my son did not fall on his head (but only on his bottom). In many cases the actual context might help to distinguish between a literal and an idiomatic usage of these words. I wonder if there are strategies to deal with such different usages (either in lexica or in the context of translation)

—Fabienne Cap (German)

idioms and the N-P-N construction (cf. Jackendoff 2008)

—Sascha Bargmann (German)

ideas for similar future events

—Veronika Vincze (Hungarian)

Could we (and should we) ever come up with a consistent, unambiguous terminology? MWE? Idiom? Collocation? Phrasal verb? LVC? VPC? In a recent CfP idioms are defined as non compositional at the beginning, but two paragraphs later they mention compositional idioms. Can we agree once and for all that e.g. MWEs are idiosyncratic (not ordinary), idioms are non compositional, and collocations are words which just tend to occur together?

—Bálint Sass (Hungarian)

Is there a class of lemmas which can be inserted into the multi-word string? The case of "fucking", "proverbial", "literal" (as metalinguistic commentaries - they show the attitude of the speaker, e.g. anger, irony).

—Monika Czerepowicka & Sebastian Przybyszewski (Polish)

Can be phrase structure and dependency structure considered isomorphic? If not, what are the aspects which clearly never can have their correspondent in the other formalism?

—Bálint Sass (Hungarian)

MWE formal description in a lexical database

—Panagiotis Minos (Greek)

UAS is used regularly. Do an unlabeled tree have any possible uses? (and one additional issue:) It's interesting to see that different languages often have (exactly) the same MWEs.

—Bálint Sass (Hungarian)

Try to sketch a follow-up for PARSEME, maybe in Horizon 2020.

—Corina Forăscu (Romanian)

feedback from trainers

—Veronika Vincze (Hungarian)

I'm new to HPSG, what to read first?

—Bálint Sass (Hungarian)

opinion mining (because most idioms express opinion on the act/state or the subject)

—Zuzana Neverilova (Czech)

If it is true that "morphology is frozen syntax" (from Shuly's Hebrew examples, it became quite clear), are fixed MWEs with no syntactical variance and no possibility to modify elements a morphologic phenomenon?

—Cerstin Mahlow (German)

The relation between language perceived as a system and texts as a realization of the system. Where is the border of the MWE's modifications which we want to implement in the parser?

—Monika Czerepowicka & Sebastian Przybyszewski (Polish)

automatic extraction of multi-word expressions (some examples, why it is difficult, what are the main problems)

—Aedmaa Eleri (Estonian)

MWEs are very diifferent in terms of their properties. Could modified approaches for different types of MWEs provide more successful results than universal one?

—Justina Mandravickaite (Lithuanian)

All of the current work on MWEs focuses on either human-annotated syntactic patterns or on specific patterns (Noun Compounds, Verb-Particle constructions...). How can we move toward a statistical approach of MWE learning?

—Silvio Ricardo Cordeiro (Portuguese)