Annotation guidelines
corpora annotated for multiword expressions

---

Words and tokens

While the definition of an MWE inherently relies on the notion of a word, manual annotation and automatic identification of VMWEs in our task is performed on texts which are automatically tokenized. It is therefore important to understand the distinction between words and tokens in the context of VMWEs.

A word is a linguistically (notably semantically) motivated unit. The detection of words is, thus, language-dependent and annotation experts should have a clear idea of how to define it for their own language (even if this definition proves hard in general).

A token is a technical and pragmatic notion, defined according to more or less linguistically motivated clues and depending on the particular tokenization tool at hand. Note that the notion of a token is ambiguous in NLP. It can also mean an individual occurrence of a certain linguistic unit, as opposed to a type, i.e. the set of all surface realisations of a unit. In these guidelines, we refrain from using this seconf sense.

Tokens should ideally be as close as possible to words. However, in practice - due to the hardness of the (automatic) tokenization task - the relation between tokens and words is not always 1-to-1. The following cases occur:

• A token coincides with a word:
مدهشة surprising, نزهة walk, ب with, قام to do
вземам, решение, наяве, бял, на, се, д-р
mít, hlad, se, úžas
einen, Spaziergang, machen, Überraschung
κάνω, μία, βόλτα, έξυπνοςkano, mia, volta, exipnos make, a, walk, clever
take, a, walk, astonishment
dar, un, paseo, sorpresa, maldecir, bienvivir
ibilaldi, bat, egin, ezuste
faire, une, promenade, étonnement
tóg, siúl, ionadh
δίδωμιdidо̄mi give give
καλός kalos beautiful beautiful
περί peri about about
napraviti, jedan, šetnja, začuđenost
tesz, egy, séta, meglepetés
mengambil, sebuah, berjalan, heran
fare, una, passeggiata, sorpresa
取る, その, 歩く, 驚き
ferħ, libes, sabiħ
een, wandeling, maken, verrassing
robić to do, na on, dokładność precision
comer eat, uma a, guarda-chuva umbrella, antessala anteroom
face, o, plimbare
iti, na, en, sprehod, začudenost
седети sedeti to seat, скрштених skrštenih crossed, руку ruku hands
gå, på, promenad, förvåning
采取, 一个, 步行, 惊愕
• Several tokens build up one word, like in abbreviations, possessive markers, words with "accidental" separators, inflected or derived forms of foreign names, etc. In this case we speak of a multitoken word (MTW): The pipe symbol '|' indicates token separation in these examples
قرار|ها| decision-her her decision
т|.|н|. etc.
год|. year
z|.|B|. for instance
Wie geht|'|s How goes it How are you
κ. κύριος Mister
υπΔρ υποψήφιος διδάκτορας PhD candidate
M|. Mister
pp|. pages
Pandora|'|s
A|/|A|. a la atención de for the attention of
a|/|f|. a favor in favor
Rte|. remitente sender
etab|. eta abar and so on
می|-|روم، آیت|-|الله، کتاب|-|ها
aujourd|'|hui today
οἷον τ'εἰμίhoion t'eimi of.what.sort and be.1SG I am able to
danas today
időjárás|-|jelentés weather forecast
vice|-|presidente vice-president
libs|et she wore
a|.|u|.|b|.| please
Pandora|'|s Pandora's
Chomsky|'|ego of Chomsky
SMS|-|ować to write an SMS
vice|-|presidente vice-president
prim|-|ministru prime minister
d|-|voastră polite "you"
g|. Mister
str|. pages
le|-|to
FIFA|-|у FIFA|-|u FIFA.ACC
tweet|-|овање tweet|-|ovanje to write tweets
EU|:|s EU's
• One token can contain several words, like in contractions and compounds. In this case we speak of a multiword token (MWT). Identifying MWTs is important because they can be potential candidates for VMWEs. However, defining what is a word and a MWT is a hard and language-specific question and language-specific MWT tests are being designed to this end. Examples of MWTs include: See also the representation of MWTs in Universal Dependencies The precise word forms cannot always be straightforwardly deduced from the MWT containing them and vice versa, as in don't, della, du, etc.
وسيكتبوناها =و + س + يكتبون + هاand they are going to write it they are going to write it
вагон-ресторант train carriage+restaurant train buffet
Schulaufgabe = Schule+Aufgabe school+exercisehomework
Apfelbaum = Apfel+Baum apple treeapple tree
στου = σε+του stu = se + tu to-the
στον = σε+τον
don't = do+not
del = de+el of the from/of the
al = a+el to+the to the
compárese = compare+se compare SE_PARTICLE be it compared
suicidarse = suicididar+se suicide SELF to commit suicide
sudurluze = sudur+luze nose+long long-nosed
jarleku = jar(ri)+leku sit+place seat
کتابش=کتاب+ش
du = de+le from the
sa = i+an in the
b'fhearr = ba+fhearr be.COND better prefer
καίτοι = καί + τοιkaitoi = kai + toi and indeed and indeed
uzbrdo = uz+brdo uphill
della = di+la of the
huiswerk = huis+werk home+workhomework
appelboom = appel+boom apple treeapple tree
pannenkoek = pan + koek pancake
Białymstoku=Białym+stoku white+slope Białystok.INST (a city name)
robiłem=robi+łem do.3.SG.PRES+be.1.SG.PAST.AGLI did
żeśmy = że+śmy that+be.1.PL.AGL that-we
neles = em+eles on them
într-o = într-+o in a
nanj = na+njega on him
напоље = на + поље napolje = na + polje outside
новосадски = ново + садски novosadski = novo + sadski Novi Sad (an adjective from a city name)
arvsmassa = arv+massa genetic stock

While a VMWE always contains at least two words, the relation between VMWEs and tokens can be twofold:

• A VMWE contains several tokens, whether each of them coincides with a word or not:
نزهة ب قام make with walk make a walk (2 words , 2 tokens)
вземам решение make a decision (2 words, 2 tokens)
прочитам от корица до корица to read from cover to cover (5 words, 5 tokens)
eine Rede halten (2 words, 2 tokens) a speech hold to give a speech
wie geht's (2 words, 4 tokens) how goes it how are you
παίρνω μία απόφασηperno mia apofasi take a decision to decide (2 words, 2 tokens)
παίζω στα δάχτυλαpezo sta dachtyla play in-the fingers to know very well (3 words, 4 tokens)
to take a walk (2 words, 2 tokens)
to open Pandora's box (3 words, possibly 5 tokens)
dar un paseo 2 words, 2 tokens to give a walk to take a walk
dar por sentado 3 words, 3 tokens to give for seated to take for granted
irse de rositas 3 words, 4 tokens to go_self of little_roses to get off scot free
ibilaldia egin (2 words, 2 tokens)

دستور داد (2 words, 2 tokens)
b'fhearr liom (2 words, 4 tokens) I would prefer
τοὺς λόγους ποιέομαιtous logous poieomai the word do.1SG to speak
dignuti ruke to raise hands to give up (2 words, 2 tokens), otvoriti Pandorinu kutiju open Pandora's box to face with problems (3 words, 3 tokens)
sétát tesz to take a walk (2 words, 2 tokens)
tenere un discorso (2 words, 2 tokens) hold a speech to give a speech
cavalcare l'onda (3 words, 4 tokens) ride the wave ride the wave
kien idur fuq il-fatt turns on the fact
een wandeling maken (2 words, 2 tokens) a walk make to talk a walk
robi z igły widły make.3.SG a pitchfork out of a needle he makes a mountain out of a molehill (4 words, 4 tokens)
robił|em z igły widły made.3.SG.M1+be.1.SG.AGL a pitchfork out of a needle I made a mountain out of a molehill (4 words, 5 tokens)
dar uma caminhada to give a walk (2 words, 2 tokens)
cair de pára-quedas to fall with parachute to arrive unprepared in the middle of a situation (3 words, possibly 5 tokens) According to new orthography rules, this word would be written 'paraquedas'. Old spelling may still be found in annotated texts, though.
queixar-se-ia complain-self-would would complain (2 words, possibly 5 tokens)
a da ortul popii to die (3 words, 3 tokens)
klicati jelene to call cerfs to vomit (2 words, 2 tokens)
vreči puško v koruzo throw a rifle in the corn to give up (4 words, 4 tokens)
данути душом danuti dušom to breathe soul to feel relieved (2 words, 2 tokens)
причати на|памет pričati na|pamet to talk by heart to talk not relying on facts (3 words, 2 tokens)
hålla ett tal (2 words, 2 tokens) hold a speech to give a speech
做 一 个 决定 (2 words, 2 tokens) do one CL decision to make a decision
• A VMWE contains one (multiword) token:
no example found for Arabic
no example found for Bulgarian
vorbereiten to pre-arrange to prepare
anfangen at-catch to begin
έδωσα-πήρα gave-1SG took-1SG I tried hard
to pretty-print
suicidarse suicide_self to commit suicide
n.a.
court-circuiter to short circuit
προσ-άγωpros-agо̄ towards lead.1SG to lead towards
pripremiti to pre-arrange to prepare
kinyír out.cut to kill
corto-circuitare to short circuit suicidarsi suicide_self to commit suicide
voorbereiden to pre-arrange to prepare
aanvangen at-catch to begin
no example found for Polish
queixar-se-ia compain-SELF-would would complain
a se-ndura RCLI.ACC-have.the.heart to have the heart
pripraviti to pre-arrange to prepare
no example found for Serbian
klargöra clear-make clarify påpeka on-point point out

Note finally that multitoken words are not considered verbal MWEs since they contain one (multitoken) word only:

no example found for Bulgarian
??
αερολογώaerologo air+talk to talk aimlessly
n.a.
odolustu blood+empty to bleed
λογοποιέομαι logopoieomai word-do to speak
SMS-ati to write an SMS
anteporre to put + in front of
SMS-ować to write an SMS
pós-datar to post-date
a binedispunewell-disposeto cheer up
SMS-jati to write an SMS
SMS-овати SMS-ovati to write an SMS

Whenever the distinction between a word and a token is judged by a particular language team as hard to tackle, a possible option is to consider these two notions equivalent for the needs of this shared task.

---