home edit page issue tracker

This page still pertains to UD version 1.

Specific constructions

Clausal structures

Reflexive pronouns

Czech has a reflexive personal pronoun that takes different forms in different cases and these forms differ from the normal, irreflexive pronouns:


The clitic forms se, si are very frequent and serve various purposes. Their default function is to represent object that is identical to the subject of the same verb. The test is that they could be substituted by a normal personal pronoun. Such instances are attached to the verb as cs-dep/dobj or cs-dep/iobj.

Jan se bude bránit . \n Jan himself will defend .
dobj(bránit, se)
dobj(defend, himself)
Barbora si přidělí strážce . \n Barbora herself will-assign bodyguard .
iobj(přidělí, si)
iobj(will-assign, herself)

The Czech reflexive pronoun is also used in reciprocal actions where other languages use a special reciprocal pronoun. These instances are still attached as dobj or iobj, respectively:

Jan a Marie se políbili . \n Jan and Marie each-other kissed .
dobj(políbili, se)
dobj(kissed, each-other)

If the reflexive pronoun can be substituted by another nominal but it is not a core argument (object) of the verb, it will be attached as cs-dep/nmod.

Zuzana si opřela kolo o zeď . \n Zuzana for-herself propped bike against wall .
nmod(opřela, si)
nmod(propped, for-herself)

The reflexive pronoun can be used to form a passive construction. This is called reflexive passive; there is also the “normal” passive built with the passive participle and the auxiliary verb být “to be”. Reflexive pronoun that forms a reflexive passive is attached as cs-dep/auxpass:reflex.

To se řekne snadno . \n It is said easily .
auxpass:reflex(řekne, se)
auxpass:reflex(said, is)

There are inherently reflexive verbs, i.e. the verb always occurs with a reflexive prounoun, and the pronoun cannot be replaced by a non-reflexive pronoun or any other nominal.

With these verbs, the reflexive pronoun is attached as cs-dep/expl.

Martin se bojí zvířat . \n Martin REFLEX fears animals .
expl(bojí, se)
expl(fears, REFLEX)

If a reflexive verb (inherently or not) has been turned to a verbal noun, the reflexive pronoun is attached to the noun as cs-dep/nmod:

Jediným cílem je utvrzení se v pocitu , že … \n Only goal is strengthening oneself in feeling , that …
nmod(utvrzení, se)
nmod(strengthening, oneself)

Finally, the dative reflexive si is sometimes used in situations where it is redundant. Such instances are attached as cs-dep/discourse:

Klaus si odsloužil 154 dnů . \n Klaus himself served-out 154 days .
discourse(odsloužil, si)
discourse(served-out, himself)

Adjectival and adverbial constructions

Comparatives (degree)

Unlike in English, most Czech adjectives and adverbs have morphological comparative and superlative forms (see the cs-feat/Degree feature): chytrý “smart”, chytřejší “smarter”, nejchytřejší “smartest”. Periphrastic constructions such as English more intelligent cannot be completely excluded but they are infrequent and often deemed poor style: inteligentnější is preferred over více inteligentní. The exception is when the adjective or adverb applies less to the entity being compared than to the entity being compared to: méně inteligentní “less intelligent” is the only way of reversing the comparison. Equality comparisons are also periphrastic.

Martin je inteligentnější než Vojta .
nsubj(inteligentnější, Martin)
cop(inteligentnější, je)
nmod(inteligentnější, Vojta)
case(Vojta, než)
punct(inteligentnější, .)
Martin je nejinteligentnější ze všech .
nsubj(nejinteligentnější, Martin)
cop(nejinteligentnější, je)
nmod(nejinteligentnější, všech)
case(všech, ze)
punct(nejinteligentnější, .)

To keep the analyses of the morphological and the periphrastic cases parallel (and also to keep the analyses parallel cross-linguistically), in the periphrastic examples the entity comapared to modifies still the adjective and not the adverb:

Vojta je méně inteligentní než Martin .
nsubj(inteligentní, Vojta)
cop(inteligentní, je)
advmod(inteligentní, méně)
nmod(inteligentní, Martin)
case(Martin, než)
punct(inteligentní, .)

If a property is compared to a clause, the clause is attached as cs-dep/advcl instead of cs-dep/nmod and the conjunction (než, jako) is attached to the subordinate clause as cs-dep/mark.

Martin je inteligentnější , než jsme mysleli . \n Martin is more-intelligent , than we-have thought .
nsubj(inteligentnější, Martin-1)
cop(inteligentnější, je)
advcl(inteligentnější, mysleli)
punct(mysleli, ,-4)
mark(mysleli, než)
aux(mysleli, jsme)
punct(inteligentnější, .-8)
nsubj(more-intelligent, Martin-10)
cop(more-intelligent, is)
advcl(more-intelligent, thought)
punct(thought, ,-13)
mark(thought, than)
aux(thought, we-have)
punct(more-intelligent, .-17)

Very commonly the complement clause in a comparative undergoes various amounts of partial reduction or ellipsis, sometimes to a quite extreme extent. In general, we treat whatever remnant that remains as still an cs-dep/advcl, as above.

On hraje opilý lépe než střízlivý . \n He plays drunk better than sober .
nsubj(hraje, On)
advcl(hraje, opilý)
advmod(hraje, lépe)
advcl(lépe, střízlivý)
mark(střízlivý, než)
punct(hraje, .-7)
nsubj(plays, He)
advcl(plays, drunk)
advmod(plays, better)
advcl(better, sober)
mark(sober, than)
punct(plays, .-15)

The limiting case is that only a nominal is present; then we analyze it as an cs-dep/nmod, although one could see Martin is more intelligent than Vojta as a reduced expression of Martin is more intelligent than how Vojta is intelligent. We lean towards minimizing the postulation of unobserved structure and opt to treat these cases as just an oblique nominal complement.

více závisí na stavu linky než na rychlosti přístroje \n more depends on state of-line than on speed of-device
advmod(závisí, více)
advmod(depends, more)
dobj(závisí, stavu)
dobj(depends, state)
case(stavu, na-3)
case(state, on-13)
nmod(stavu, linky)
nmod(state, of-line)
nmod(více, rychlosti)
nmod(more, speed)
case(rychlosti, na-7)
case(speed, on-17)
case(rychlosti, než)
case(speed, than)
nmod(rychlosti, přístroje)
nmod(speed, of-device)

Comparatives (quantity)

In the periphrastic comparatives in the previous section, the words více “more” and méně “less” are comparative forms of the adverbs hodně/mnoho “much/many” and málo “little”, respectively. However, in other situations they combine directly with nouns and act as quantifiers (termed indefinite numerals in the Czech grammar but labeled cs-pos/DET in accord with our definition). They behave syntactically like high-value numerals (see cs-dep/nummod for details) and we attach them as cs-dep/det:numgov or cs-dep/det:nummod.

Dnes přišlo více chlapců , než jsme čekali . \n Today came more boys , than we-have expected .
advmod(přišlo, Dnes)
nsubj(přišlo, chlapců)
det:numgov(chlapců, více)
advcl(více, čekali)
mark(čekali, než)
aux(čekali, jsme)
punct(čekali, ,-5)
punct(přišlo, .-9)
advmod(came, Today)
nsubj(came, boys)
det:numgov(boys, more)
advcl(more, expected)
mark(expected, than)
aux(expected, we-have)
punct(expected, ,-15)
punct(came, .-19)

As with qualitative comparisons, we use nmod instead of advcl and case instead of mark when the comparative complement is reduced to just a nominal:

Petr má více peněz než Pavel . \n Petr has more money than Pavel .
nsubj(má, Petr-1)
nsubj(has, Petr-9)
dobj(má, peněz)
dobj(has, money)
det:numgov(peněz, více)
det:numgov(money, more)
nmod(více, Pavel-6)
nmod(more, Pavel-14)
case(Pavel-6, než)
case(Pavel-14, than)
punct(má, .-7)
punct(has, .-15)

In certain contexts the comparative complement combines both the action or adjective that is being compared and the quantity it is compared to:

In these cases we consider více než to be a multi-word expression because the two words are inseparable. One cannot say *více procent než 90 (the word procent can be pulled to the front but then it will skip the whole MWE, as in těch procent nebylo více než 90 lit. the percent were-not more than 90.)

To je více než pravděpodobné . \n That is more than likely .
nsubj(pravděpodobné, To)
nsubj(likely, That)
cop(pravděpodobné, je)
cop(likely, is)
advmod(pravděpodobné, více)
advmod(likely, more)
mwe(více, než)
mwe(more, than)
punct(pravděpodobné, .-6)
punct(likely, .-13)


Ellipsis means that there is something missing in the sentence. Something that has been omitted from the surface form, although it is understood by both the speaker and the listener. Various phenomena can be classified as ellipsis; the most important and difficult are those where the missing word has dependents. Where do we attach these orphans to?

Several different solutions can be found in treebanks. One of them is to include an empty node (labeled NULL, #Fantom etc.) that represents the missing word. Orphans are then attached to the empty node with their real dependency relation labels. Such analysis would be linguistically adequate but it would violate our principle that dependencies exist between real syntactic words. (It would also make parsing more difficult.) We do not insert empty nodes.

If empty nodes are not an option, some treebanks attach all orphans to the grandparent, i.e. to the parent of the missing parent node. Then they may

Another possibility is that one of the orphans gets promoted to the place of the missing parent and the other orphans are attached to it.

We use a combination of approaches in the Czech UD. The only limitation is that we do not reconstruct nodes that are not present in the surface sentence form.

If the head noun is missing from a noun phrase, i.e. there is just an adjective, possibly also a numeral or a determiner, then one orphan is selected as the main dependent and it gets promoted:

Zatímco mně zbylo pět malých zelených jablíček , Petra měla tři velká červená . \n While to-me remained five small green apples , Petra had three big red .
dobj(měla, červená)
dobj(had, red)
nummod(červená, tři)
nummod(red, three)
amod(červená, velká)
amod(red, big)

Note that Czech does not have promotion of auxiliaries like in English I did not come but he did. Occasionally yes/no is used to construct similar sentences, as in Já jsem nepřišel, ale on ano. lit. I have not-come, but he yes.

We do not use promotion when a verb is missing and two or more arguments of the verb are present. A frequent special case of this is coordination of clauses that share the same verb but only the first occurrence of the verb is retained on the surface, while the other copies have been deleted and only their dependents remain: Pavel si objednal hovězí a Markéta [si objednala] vepřové. “Pavel ordered beef and Markéta [ordered] pork.” Universal Dependencies annnotate such cases using the cs-dep/remnant relation, which enables reconstruction of the functions of the arguments, without inserting an empty node for the missing verb:

Pavel si objednal hovězí a Markéta vepřové . \n Pavel himself ordered beef and Markéta pork .
nsubj(objednal, Pavel-1)
nsubj(ordered, Pavel-10)
dobj(objednal, hovězí)
dobj(ordered, beef)
remnant(Pavel-1, Markéta-6)
remnant(Pavel-10, Markéta-15)
remnant(hovězí, vepřové)
remnant(beef, pork)

Sometimes a verb is missing but there is no coordination and no overt copy of the verb, hence we cannot use the remnant analysis. In particular, there are sentence-like segments that lack the main verb: A co na to [říká] MF? “And what [does] MF [say] to it?”

Since release 1.2 of the Czech UD treebank, there is just one node with the root dependency relation in every tree; when there are multiple orphaned dependents at the top level of the tree, the leftmost dependent is promoted to the head (root) position and the other orphans are attached to it.

ROOT A co na to MF ? \n ROOT And what to it MF ?
root(ROOT-1, A)
root(ROOT-9, And)
dobj(A, co)
dobj(And, what)
nmod(A, to-5)
nmod(And, it)
case(to-5, na)
case(it, to-12)
nsubj(A, MF-6)
nsubj(And, MF-14)
punct(A, ?-7)
punct(And, ?-15)