AplikaceAplikace
Nastavení

This is an old revision of the document!

Content-word features

  • Used to annotate multiword (analytic, periphrastic) verb forms; a similar role is played by the verbtag attribute in more recent Czech corpora.
  • They are based on Universal Dependencies (UD) annotation, in which all function words depend on content words.
  • They are included in the InterCorp parallel corpus from version 17ud onwards for Slavic and Romance languages.
  • They are used with content (autosemantic) verbs and with the head of a nominal predicate.1)
  • They are represented as the value of the positional attribute cw_feats.
  • For single-word forms, the value of cw_feats contains the categories from the feats attribute, i.e. a sequence of “category name = category value” pairs separated by the “|” character. This makes it possible to formulate queries for cw_feats values regardless of whether the form is single-word or multiword. For example, content verbs expressing first-person singular in either single-word or multiword forms can be searched for in KonText using an advanced query such as: [cw_feats="Person=1" & cw_feats="Number=Sing"].2)
  • For multiword forms, cw_feats also contains categories expressed by auxiliary and copular verbs and short reflexives (in Czech se and si), dependent on the given content-word form, including negation.
  • In addition to the morphological categories defined in UD for the feats attribute, cw_feats also contains the following information:
    • references to words that are part of the multiword form; for example, Span=[8,9,11] refers to the corresponding word-order positions
    • the value Analytic=Yes distinguishes multiword forms of analytic morphology expressing verbal categories of tense and mood from multiword forms containing a copula or short reflexives, including passive forms (Analytic=No)3)
    • in the reflexive passive, the role of the reflexive is expressed by Expl=Pass; for other reflexives, Expl=Pv is used
  • The annotation of multiword verb forms is based on the work of Lenka Krippnerová (Identifikace složených gramatických tvarů, Periphrastic Verb Forms in Universal Dependencies) and was performed using her script.

In the examples below, the verb forms (underlined) are shown with their cw_feats values. Categories that are “additional” in cw_feats are shown in bold, i.e. categories that are either absent from the feats of the content verb or nominal predicate, or have a different value there.4)

  1. požádal jsem: Analytic=Yes | Span=[1,2] | Animacy=Anim | Form=Fin | Gender=Masc | Mood=Ind | Number=Sing | Person=1 | Tense=Past | Voice=Act
  2. je možné: Analytic=No | Span=[2,3] | Aspect=Imp | Form=Fin | Mood=Ind | Number=Sing | Person=3 | Tense=Pres | Voice=Act
  3. jste nebyl vyznamenán: Analytic=Yes | Span=[12,13,14] | Animacy=Anim | Form=Fin | Gender=Masc | Mood=Ind | Number=Sing | Person=2 | Polarity=Neg | Tense=Past | Voice=Pass
  4. ai demandé (fr): Analytic=Yes | Span=[7,8] | Aspect=Perf | Form=Fin | Mood=Ind | Number=Sing | Person=1 | Tense=Past | Voice=Act
  5. Prosiłem (pl):5) Analytic=Yes | Span=[1,2] | Animacy=Hum | Aspect=Imp | Form=Fin | Gender=Masc | Mood=Ind | Number=Sing | Person=1 | Tense=Past | Voice=Act
  6. Pokud1 je2 člen3 přítomen4 Analytic=No | Span=[2,4] | Aspect=Imp | Form=Fin | Mood=Ind | Number=Sing | Person=3 | Tense=Pres | Voice=Act a5 nehlasuje6 Analytic=No | Span=[6] | Form=Fin | Mood=Ind | Number=Sing | Person=3 | Polarity=Neg | Tense=Pres | Voice=Act ,7 bude8 se9 to10 posuzovat11 Analytic=Yes | Span=[8,9,11] | Expl=Pass | Form=Fin | Mood=Ind | Number=Sing | Person=3 | Tense=Fut | Voice=Pass tak12 ,13 jako14 by15 se16 zdržel17 Analytic=Yes | Span=[15,16,17] | Animacy=Anim | Expl=Pv | Form=Fin | Gender=Masc | Mood=Cnd | Number=Sing | Person=3 | Voice=Act hlasování18 .196)
1)
In Universal Dependencies, the copula is treated as an auxiliary verb.
2)
The query [cw_feats=".*Number=Sing|Person=1.*"] should return the same result.
3)
If, for example, the passive voice co-occurs with the past tense, the entire form is treated as analytic.
4)
The examples are taken from the corpus; the word-order positions given in Span have been preserved.
5)
Although this is written as a single word, it is analysed as two words: an l-participle and an auxiliary verb.
6)
In this example, which shows an entire sentence, the position of each word is indicated. If the cw-feats attribute is empty, the position number and the following slash are followed by a hyphen.

Trace: • content-word_features