Summary: How to distinguish events and non-events

Quick HowTo for annotating Czech sentences:

Events vs. non-events

All verbs (both action verbs and statives) in all their occurrences (in predication, modification, and reference) denote events.
Based on UMR lists, sheet "Abstract Rolesets"
- reifications ... events
- discourse relations/reifications ... events
  (!!! the only exception is "but-91", similarly as other discourse relations, sheet "Abstract concepts"!!!)
- non-prototypical pred rolesets (9 non-verbal predicates) ... events
- implicit rolesets ... some are events, some not :-((
  preliminary classification:
  - events: corelate-91, have-degree-91, include-91, infer-91, resemble-91, have-experience-91
  - non-events: everything else

See below for Julia's comments!

Consequences of being designated an 'event' in UMR

The consequences of being designated an 'event' in UMR are the following (as summarized by Julia Bonn):

roleset use:: if rolesets are being used, qualifying as an 'event' means that it is reasonable to represent the expression in question using a roleset.
(But note that it is also possible to use rolesets for expressions that do not qualify as 'events'. For example, in English, the roleset hunger-01 has been used to cover expressions 'hunger-verb', 'hunger-noun', and 'hungry-adjective'. According to the guidelines, 'hungry' in the sentence the hungry man is a state in modification, and therefore doesn't count as an 'event'. But the roleset for hunger is still semantically relevant -- both in terms of sense disambiguation and in terms of the allowed argument structure. What I've been proposing (and which seems to be agreed to by everyone now) is that we should be allowed to use the hunger-01 roleset here.)
:aspect annotation: 'events' receive aspect annotation.
:modal-strength annotation: 'events' receive modal annotation in the sentence and document graphs
:temporal dependency annotation: 'events' qualify for placement in the document-level temporal dependency.

1. Verbs and abstract predicates (in references and modification)

1.A Stative verbs ... in the same way as action verbs (as events)

We do not want to distinguish between action and state verbs (due to an unclear boundary in Czech.

If stative verbs such as milovat “love” occur as concepts, they will be treated as events regardless whether they are used in predication (on ji miluje), modification (muž, který ji miluje) or reference (milovat ji je jeho osud). (UMR guidelines: "Two-place statives, such as love in My cat loves wet food., are annotated in the same way as predicated processes, i.e. an event is identified and labelled with the predicate in the language.")

What about stative verbs in reference or modification? One would suppose that they should not be treated as events (they do not denote processes and are not used in predication). Am I right?

Feedback from Julia:
This question is one of my own great questions right now.

If a state is being used as modification of some referent, and it's not given as a predicate, then theoretically it wouldn't be treated as an event.
For me, with English, I would still use an applicable roleset if one exists-- I just wouldn't complete :aspect, :modal-strength, or :temporal dependency annotation.
For states in reference though, the guidelines specify that they should not be treated as events, but I'm not sure I agree. It seems to me like referring to the abstraction notion of 'kindness' is one thing-- certainly that doesn't seem like it needs :aspect, :modal-strength, or :temporal dependency annotation. But 'his kindness to his mother last week' is much more event-like.
I'm not sure where to draw the line. I am hoping to get some clarification on this soon.

Preliminary conclusion for Czech annotations:
We will annotate statives in the same way as action verbs, i.e., as eventive concepts (regardless the information packaging), unless it is a clear case of a non-eventive concept. To put it differently, we will:

represent it using its roleset (if exists)

annotate :aspect, :modal-strength (both in the sentence and document graphs);
further, it qualifies for placement in the document-level :temporal dependency annotation.

1.B "Abstract predicates" such as have-91, identity-91 ... in the same way as action verbs (as events)

These 9 predicates introduced in 3.1.1.3 are definitely considered events if structured as predication, as, e.g., in Petr je bohatý (with the have-mod-91 predicate). ~~We tentatively decided that the abstract predicates (reifications) will in general be treated as stative verbs, i.e., as events, independently of their packaging.~~

These predicates are used less frequently in modification because then we can use the relation they reify, e.g., :mod.

Feedback from Julia:
"Abstract predicates" (like have-91, ... identity-91) are events if they are structured as predication but non-events if structured as reference or modification (e.g., in relative clauses), right?

Preliminary conclusion for Czech annotations:
We will use "abstract predicates" only in cases of predication (using their rolesets and including the aspect, modal and temporal dependency annotation).
Otherwise we will stick to the relation they reify.

1.C Reification ... in the same way as action verbs (as events)

Are rolesets for reifications used in the same way as "abstract predicates" (above)?

Feedback from Julia:

Reification: I think it's much more unusual for a lot of the other implicit rolesets to come up in a predicated form, but some of them do occasionally, and that goes for the -91 rolesets used for reification as well. (Generally, this is a copular or existential construction in English).

Preliminary conclusion for Czech annotations:
Reification: The abstract rolesets for reification should be considered as events if they are seen as predicates (i.e., they will be treated in the same way as other -91 predicates (others than those for discourse relations)).

Rolesets have-quant-91 AND have-degree-91

Rolesets for quantities and rolesets for comparative/superlative constructions (as the special cases of reification should not be considered events? Also other quantity-related abstract predicates like rate-entity-91, range-91, etc.

Feedback from Julia:
I use these in the same way you mentioned above. For sentences like He was the tallest boy in the room or there are 10 more blue blocks than red blocks, I would use have-degree-91 or have-quant-91 as the top node of the graph and treat it as an event, with :aspect/:modstr/:temporal dependency.

But for a sentence like she ate the biggest banana of the bunch, I would use have-degree-91 so that I could capture the superset entity as a numbered argument (the bunch), but I wouldn't treat it as an event. Otherwise :quant and :degree relation/attribute are used.

Preliminary conclusion for Czech annotations:
Due to an unclear boundary, we will consider all cases annotated with have-quant-91 or have-degree-91 as predicates (unless it is a clear case of a non-eventive concept), i.e. using their rolesets and including :aspect/:modstr/:temporal dependency.

1.D "Discourse relation rolesets" ... as non-events

However, abstract predicates that the UMR spreadsheet lists under “discourse relation rolesets” (10 types (Jan 2024), see the list) will not be treated as events.
discourse.

Feedback from Julia:
ML: In particular, I suppose that discourse relations (10 types) should not be considered events even if the respective rolesets are referred to as abstract predicates from time to time in the Guidelines?
JB: Yes, this too!

Conclusion for Czech annotations:
Discourse relations, see the list, should NOT be considered as events (despite having rolesets)!

1.E Implicit rolesets ... it depends :-((

Implicit rolesets are used for for indicating such things as A. "metadata" (like publication-91, hyperlink-91, street-address-91, ...),
B. but also resemble-91, include-91 (e.g., for percentage).

Feedback from Julia:
Metadata info: I can't think of a case with publication-91 offhand-- maybe that one isn't one that occurs in English. My goal is to try to follow the same predicating/non-predicating guidelines with all of the -91 rolesets.

Preliminary conclusion for Czech annotations:
Metadata info: These rolesets should be typically treated as non-events (i.e., without the aspect, modal and temporal dependency annotation).
BUT: rolesets as resemble-91, include-91 probably more like stative verbs (thus events)?

2. Nouns (connected to verbs)

2.A Agentive nouns (and other cases of inverse participant roles)

Inverse participant roles (typically ARGx-of roles), see Part 3-2-1-3 and the notes on roles here, are typically used in nominal modifications (esp. relative clauses like sweater that ..., kindship relations like his father, etc.) and in references (esp. participant nominalizations like runner.).
Further, it allows us to annotate agentive nouns using the corresponding verb (with its frame = roleset): e.g., runner is a person who runs or volič is a person who takes part in the elections, i.e., is represented as ARG0-of the verb volit:

runner
 (p/ person 
     :ARG0-of r / run-02)

volič
 (p/ person 
     :ARG0-of v / volit-001)

Preliminary conclusion for Czech annotations:
In these cases, we not only use the verbs' rolesets to annotate agentive nouns but also include aspect annotation (and modstr/?temporal atributes as well).
Typically, the habitual value serves as default (esp. for imperfective verbs), like for učitel "teacher" (= ten, kdo učí) as a teacher teaches habitually, řidič "driver" (= ten, kdo řídí / umí řídit). Depending on semantics of verb, performance can be relevant, e.g., vrah "murderer" (it is enough to commit just one murder to be a murderer).

2.B Nouns denoting processes

Based on the Guidelines, nouns denoting processes like válka "war", příchod "arrival", bouře "storm" should be annotated as eventive concepts.

Following the guidelines, the ideal solution for Czech is to annotate them as events.

We will definitely interpret them as events if there is a frame for such nouns available in the PDT-Vallex.
However, not all nouns denoting processes are stored in PDT-Vallex (in particular, some similar words in PDT-Vallex are not assigned a valency frame). Thus, when annotating from scratch, we will have to create new entries and valency frames for such nouns when necessary.
For (semi)automatic transfer from PDT, this might not be achievable so nouns missing in PDT-Vallex will be probably left without eventive annotation

See Section 2. Deverbal nouns other than ní/-tí below for more details.

2.C Deverbal nouns other than ní/-tí:

JH: Be conservative (at least for the time being)!
Discussion on 31.7.2023, see the 31.7.2023 meeting minutes: “We do not want to add a big number of nouns / adjectives to the lexicon OR represent them as verbs, unless we have a clear case of a deverbal noun (ending with -ní/-tí) or deverbal adjective (see below) … YES – morphological criterion!!!”

ad food-related discussion:

[en] unhealthy food (3-3-3 (1a))

(t/ thing
	:ARG1-of (e/ eat-01)
	:mod (h/ healthy
		:polarity -))

[en] 3-2-1-1 (6a) He gave the cat some wet food.

He gave the cat some wet food.
(g/ give-01  
	:actor (p/ person
		:ref-person 3rd
		:ref-number Singular)  
	:theme (f/ food
		:mod (w/ wet)
		:quant (s/ some))  
	:recipient (c/ cat
		:ref-number Singular)
	:aspect Performance
	:modstr FullAff)

3. Deverbal adjectives other than active/passive participial adjectives:

??? Derivations of adjectives from the l-participle are less productive. Therefore, spadlý “fallen”, vzrostlý “full-grown”, bývalý “former” etc. are not events.

??? Other deverbal adjectives are also not events: kulhavý “limping”, učenlivý “quick to learn”, představitelný “imaginable”, obstojný “passable”.

According to the Guidelines, the abstract predicate have-mod-91 is used in case of predication and simple :mod if packed otherwise:

[en] The cat is black.

The cat is black.  
(h/ have-mod-91
	:ARG1 (c/ cat
		:ref-number Singular) 
	:ARG2 (b/ black)
	:aspect State
	:modstr FullAff)

[en] The black.mod cat ...

However, the Guidelines provide also the following example (the original example focused on "the more - the more" constructions, within Part 3-1-3-8 "Miscellaneous constructions").

[en] ... that you have a black heart. (= you have a heart and the heart is black)

(h / have-03
    :ARG0 (p/ person)
    :ARG1 (h/ heart  
       :ARG1-of (b / black-06)))

Eventive concepts - Comments to the Guidelines

Eventive concepts (or events) constitute important building blocks in UMR annotation – their identification is important for

annotation of participant roles, as well as for
annotation of aspect and modality.

According to the UMR Guidelines (Part 3-1-1.), identification of eventive concepts should be "based on a combination of semantic type and information packaging (Croft 2001)". Criteria for identification of eventive concepts are largely based on the criteria used in TimeML (Pustejovsky et al. 2005):

(Pustejovsky et al. 2005): [TimeML] Identifies all classes of event expressions;
(a) Tensed verbs; has left, was captured, will resign;
(b) stative adjectives and other modifiers; sunken, stalled, on board;
(c) event nominals; merger, Military Operation, Gulf War[.]

(See also the short description in terminologie.md.)

The UMR Guidelines are not clear enough about the consequences for the annotation when a concept is or is not treated as an event. However, example annotations suggest that the main difference is that for events, aspect and modality should be annotated (which is, in general, not true for entities; as for states, the UMR guidelines are not clear enough -- at least some cases of states (as, e.g., two-place stative verbs like love are annotated in the same way as processes; compare also (Pustejovsky et al. 2005), who classify stative adjectives as events).

Semantic type

Semantic type refers to the difference between:

entities (or, objects) ... prototypically nouns, but also nominalizations as deadjectival or deverbal nouns or infinitives ... TO READ: Paducheva (1995) – types of (nominal) reference,
states (or, properties) ... prototypically adjectives, but also other nominal modifiers as PPs, relative clauses, participles, and
processes ... prototypically (finite) lexical verbs, but also predicate nouns or adjectives, complements.

Information packaging

Information packaging, on the other hand, concerns the way how the semantic content is 'expressed', i.e., whether it is packed as

reference (what the speaker is talking about),
modification (additional information provided about the referent), or
predication (what the speaker is asserting about the referents in a particular utterance).

Comment:
Not entirely clear how to understand “information packaging”.

According to Croft, “constructions always involve the information packaging of the semantic content of the sentence, that is, the function of constructions has to be defined in terms of both semantic content and information packaging. … information packaging functions are much more isomorphic to syntactic structures than semantic classes or semantic relations. Information packaging functions are less variable across languages than semantics, especially lexical semantics.” (Croft, invited talk, UDW17, http://universaldependencies.org/udw17/program.html)

TO READ: William Croft (2021): Eventive Complex Predicates and Related Constructions (draft from June 2021, sent by DZ)

Eventive concepts in UMR

RULE 1: (UMR rule)
The following should be annotated as an eventive concept:

whatever is a process (semantic type) or

whatever is expressed as predication.

Putting it differently:

RULE 1a ... Eventive concpets:

Processes packed as predication are events (esp. finite action verbs as predicates of main clauses)

Processes packed as referents or modifiers should also be identified as events, esp.:

event nominals (e.g., storm, conference, ...).

complements, infinitives, gerunds with action verbs (packed as reference, e.g., subject/object clauses, adverbial clauses, non-finite complement)

relative clauses with action verbs (packed as modification)

participles with action verbs (packed as modification)

secondary predication with action verbs (packed as modification)
(based on the UMR rule and on examples provided in the Guidelines)

Entities and states packed as predication are events, esp.

predicate nominals/adjectives, complements - What exactly are meant??;
In particular:

stative verbs as predicates (e.g., love),

so called non-verbal clauses (expressing possession, location, property/object predication, equational) (UMR rule)

It implies that states in modification, e.g., as an attribute or in a relative clause (The tall man...; The man, who is tall...), and states in reference (His happiness...) are not events.

Similarly, entities in modification (The man, who is a doctor...) and entities in reference (The doctor), are not identified as events.

Eventive concepts – rules applied in annotation of Czech

As there is no clear boundary between action and state verbs in Czech, we want to avoid distinguishing between these two classes – both types should be anchored in SynSemClass (PDT-Vallex), regardless of their information packaging (predication, modification, or reference); see below.
CONFLICT with UMR guidelines!!

RULE 2: (internal criterion for Czech)
Whenever a "verb-based" concept

denotes an activity or a state (in a broad sense) and

has a roleset,

then it should be represented as an event and anchored in the (valency) lexicon (with its roleset(s)); it should get aspectual annotation

... CONFLICT with UMR guidelines for stative verbs (not packed as predication)!!

JH: Be conservative (at least for the time being)!
Discussion on 31.7.2023, see the 31.7.2023 meeting minutes: “We do not want to add a big number of nouns / adjectives to the lexicon OR represent them as verbs, unless we have a clear case of a deverbal noun (ending with -ní/-tí) or deverbal adjective (see below) ... YES – morphological criterion!!!” ... (CONFLICT with UMR guidelines!!)

However, those deverbal nouns/adjectives that are already covered in PDT-Vallex and clearly describe a process should be considered as events.

Thus:

dělat “do” is an event because it is a verb, it denotes a process and it has a frame file.
mýt “wash” is an event because it is a verb, it denotes a process and it has a frame file.
milovat “love” is an event because it is a verb and it has a frame file; although it denotes a state, we will treat it always as an event, regardless information packaging

dělání “doing”, mytí “washing”, milování “loving, making love” are events if they denote an activity or state because they are deverbal nouns, derived using the most productive -ní/-tí suffixes (if describing a process);
- BUT: mytí (lit. washing) as "a small hygienic bag", šití (lit. sewing) as "sewing kit" are (deverbal) nouns denoting things (in a strict sense) and as such, they should be annotated as entities ?
prodej “sale” is a deverbal noun derived (from prodávat “sell”) in a less productive way; it has an entry and thus a roleset in PDT-Vallex - ideally, it should it be annotated as an event (if describing a process), similarly like the much less frequent prodávání “selling”;
válka “war” is a noun not derived from verb with the -ní/-tí suffixes, despite denoting processes; it has two entries (and thus rolesets) in PDT-Vallex, thus it should it be annotated as an event (if describing a process)
koncert “concert” is a noun not derived from verb with the -ní/-tí suffixes, despite it denotes a process. It is not covered by PDT-Vallex and thus it is not annotated as an event **.

myjící “washing” is an event because it is an active participial adjective derived from the present converb.
cestující "passenger" (lit. the travelling) - a conversion, it denotes a person so it is not an event; still, it should be annotated as ARG0-of of the verb cestovat (i.e., the same way es e.g., teacher, ex. 3-2-1-1-1 (1b))
??? mycí “to be used for washing” is not an event because it is a different type of derivation.
- ??? Analogously, dělající, milující, skládající, plovoucí, plnící are events while skládací, plovací, plnicí are not.
udělavší “having done” is an event because it is an active participial adjective derived from the past converb.
udělaný “done” is an event because it is a passive participial adjective. Analogously, mytý, umytý, milovaný are events.
??? Derivations of adjectives from the l-participle are less productive. Therefore, spadlý “fallen”, vzrostlý “full-grown”, bývalý “former” etc. are not events.
??? Other deverbal adjectives are also not events: kulhavý “limping”, učenlivý “quick to learn”, představitelný “imaginable”, obstojný “passable”.

Examples:

[cs] Muzeum zaplatí necelé 2 milióny korun za novou střechu.
[en] The museum will pay almost 2 million crowns for a new roof.
... process (zaplatit, pay) in predication --> 1 event (OK both rules)
[cs] Než šla do školy, opravila mi kolo.
[en] Before she went to school, she repaired my bike. (from the UMR Guidelines)
... both processes in predication --> 2 events (OK both rules)
[cs] Chtěla jít do školy.
[en] She wanted to go to school. (from the UMR Guidelines)
... state in predication (want, chtít) PLUS process (go, jít) --> 2 events (UMR rule), 1 event in PDT ... see One or two concepts?
--> BUT: annotated as a single event in Czech (allowed by UMR guidelines)
Compare also:
- [cs] Chtěla, abych šel do školy.
  [en] She wanted me to go to school.
  ... state in predication (want, chtít) PLUS process in reference/modification (go, jít) --> 2 events (OK both rules)

Annotation of (Lexical) Verbs -- action verbs vs. state verbs

As there is no clear boundary between action and state verbs in Czech, we want to avoid distinguishing between these two classes -- both types should be anchored in SynSemClass (PDT-Vallex), regardless of their information packaging (predication, modification, or reference).

RULE 2a: (internal criterion for Czech verbs)

Verbs denote processes (action verbs) and states (state verbs). Both types should be annotated as events (regardless of their information packaging) and thus anchored in the lexicon (with their rolesets).
... CONFLICT with UMR guidelines for state verbs!!
In particular:

verbonominal predicates (predicate nominals/adjectives = nouns/adjectives in predication) form a single unit with the copula verb (thus a single event) (see below Rules 2b for nouns and 2c for adjectives);

complex predicates (= light verb constructions) should be identified as events (a single predication, thus a single event);

secondary predication (= doplněk) should be annotated as an event;

complement clauses (e.g., subject/object clauses, adverbial clauses, non-finite complement) should be identified as events;

nominalizations, infinitives should be identified as events;

participles with action verbs or state verbs should be annotated as events;

??? relative clauses ???

Action and state verbs

Examples:

[cs] Moje kočka nesnáší granule.
[en] My cat loves wet food. (from the UMR Guidelines) ... state in predication (nesnášet "hate"), thus should be annotated as an event, see 3-1-1-3, where the guidelines explicitly acknowledge that there can be two-place states, and [en] love is an example. --> event (OK both rules)
[cs] Student hrající na housle má rád Bacha.
[en] The student playing the violin likes Bach. (from the UMR Guidelines)
... process in modification (hrát "hrát") PLUS state in predication, thus eventive, see below (mít_rád "like, ") --> 2 events (OK both rules)
[cs] Student, který hraje na housle, má rád Bacha.
[en] The student, who is playing the violin, likes Bach. (from the UMR Guidelines)
... process in modification (hrát "play") PLUS state in predication, thus eventive (mít_rád "like") --> 2 events (OK both rules)
[cs] Nikdo si nevšiml, že přišla.
[en] Nobody noticed that she arrived
... process/state in predication (všimnout_si "notice") PLUS another process (přijít "arrive"), no matter how packed --> 2 events (OK both rules)
- BUT: [cs] Jejího příchodu si nikdo nevšiml.
  [en] Nobody noticed her arrival.
  ... process/state in predication (všimnout si "notice") PLUS process packaged as reference (event nominal příchod "arrival") ... should be also annotated as an event - but see Rule 2b for nouns below???
  --> 1/2 event(s) in Czech data ???; 2 events (by UMR rules)
[cs] Zaskočilo mě, jak rychle se vrátila.
[en] It surprised me how quickly she returned.
... process/state in predication (zaskočit "surprise") PLUS another process (vrátit_se "return"), no matter how packed --> 2 events (OK both rules)
- BUT: [cs] Její rychlý návrat mě zaskočil.
  [en] Her quick return surprised me.
  ... process/state in predication (zaskočit "surprise") PLUS process in reference (event nominal návrat "return"; with a roleset in PDT-Vallex ... should be also annotated as an event - but see Rule 2b for nouns below???
  --> 1/2 event(s) in Czech data ???; 2 events (UMR rules)
[cs] Kdo neriskuje, nevyhraje.
[en] He who does not risk, does not win.
... this is a relative subject clause, although it lacks the governing nominal in the Czech version. It should be analyzed using the abstract concept person modified by the risking concept. Risking is a process, hence it is still treated as event, despite being packaged as modification. There is another process there (vyhrát "win") packed as predication (the main clause) --> 2 events by both rules OK
- BUT: [cs] Kdo není mazaný, nevyhraje.
  [en] He who is not cunning does not win.
  ... a variant of the previous example. Now we have a state (not process) in modification (hence not event according to the UMR Guidelines);
  BUT it is a subject clause with a verbonominal predicate (with have-mod-91), thus should be annotated using this predicate (non-verbal clause); see also Rule 2c for adjectives --> just 1 event (UMR rules) + a non-verbal clause
[cs] Moje dítě sedí na lavičce.
[en] My child is sitting on a bench.
... state in predication (sedět "sit") PLUS modification (můj "my") --> 1 event (OK both rules)
- BUT: [cs] Dítě sedící na lavičce je moje.
  [en] The child sitting on the bench is mine.
  ... here the sitting state is in modification, hence not an event based on original UMR rules (but anchored in the valency lexicon in Czech data in the same way as action verbs);
  ... the other state (můj "mine") is in predication, hence it is an event (if we consider being somebody's as a verbonominal predicate, probably with být-007 or its -91 analogy), see also Rule 2c for adjectives --> 2 events in Czech annotation
[cs] Pokud teplota klesne pod 7 stupňů, nasadíme zimní pneumatiky.
[en] If the temperature drops below 7 degrees, we will put on winter tires.
... process in adverbial clause (klesnout "drop") is an event. --> OK both rules
[cs] Když budeš hodný, koupíme ti zmrzlinu.
[en] If you are nice, we will buy you ice cream.
... state in adverbial clause (hodný "nice") is also an event because it is in predication. --> OK both rules
[cs] Děti jedly zmrzlinu sedíce v autě.
[en] Sitting in the car, the children ate ice cream.
... the converb is a non-finite adverbial clause, hence a state in predication (sedět "sit"), hence an event. --> OK both rules

Verbonominal predicates

see below (Annotation of Nouns, Annotation of Adjectives)

Complex predicates = light verb constructions

see below (Annotation of Nouns, Annotation of Adjectives)

Secondary predication

Examples:

[cs] Koupala se v jezeře nahá. [en] She swam in the lake naked. ... state in secondary predication (nahý, naked) is an event. --> OK 2 events by both rules

Complement clauses

[en] Before she went to school, she repaired my bike. (3-1-1-1 (1))
... both verb in the main (independent) clause (repair) and verb in the dependent clause (go) denote events

Nominalizations, Infinitives

TODO

[en] She wanted to go to school. 3-1-1-2 (1b)
... both verb in the main (independent) clause (want) and infinite verb (go) denote events .... BUT another solution for Czech (see modal verbs)

The second training was cancelled yesterday.

Participles

TODO

TODO relative clauses ???

TODO

Note on causal relationships

Causal relationship should be annotated as events if packaged as predication (and as non-events otherwise).

Examples:

[cs] Exploze způsobila zhroucení domu. [en] The explosion caused the house to collapse. (from the UMR Guidelines) ... three eventive concepts: 1 causal verb in predication (cause, způsobit), 1 complement/nominalization (collapse, zhroucení), 1 event nominal in reference (explosion, exploze) Alternative rule 2 (deverbal noun but not ending with -ní,-tí): explosion probably not as an eventive concept?
[cs] Dům se zhroutil kvůli explozi. [en] The house collapsed because of the explosion. (from the UMR Guidelines) ... two eventive concepts: 1 process in predication (collapse, zhroutit se) and 1 event nominal in reference (explosion, exploze - but see above); causal relationship not expressed in predication
[cs] Dům se zhroutil, protože v něm něco explodovalo. [en] The house collapsed because something exploded in it. ... main clause and adverbial clause, each with one event (process in predication); the causal relationship (protože, because) is not an additional event.

Some languages (e.g., Basque) have morphological causative:

[eu] (a) Zopa izugarri gustatzen zaio mutilari. lit. "soup greatly pleasing it.is.it to.boy" [en] "The boy likes the soup." (normal active voice)
[eu] (b) Goseak zopa izugarri gustatuerazi zion mutilari. lit. "hunger soup greatly made.pleasing it.has.it.it to.boy" [en] "Hunger made the boy like the soup." (causative voice) ... (DZ) I suppose that both in (a) and in (b) there is just one event (gustatzen "like"). It is a state but it is packaged as predication. Or should we decompose (b) to two events (causing and liking)?

Annotation of nouns -- event nominals vs. primary nouns

RULE 2b: (internal criterion for Czech nouns)

(Primary) nouns typically denote entities (in reference), thus without a roleset.

Event nominals ending with -ní/-tí should be represented as coresponding verbs (e.g., přijíždění "arrival") if they denote a process (and thus they should get aspect annotation).

Agentive nouns (e.g., učitel"teacher", řidič "driver"), even when represented as ARG0-of the respective verb, are entities (persons).

Other event nominals (e.g., příjezd "arrival", jídlo "food") represent as entities (unless they are already covered by the valency lexicon). ... CONFLICT with UMR guidelines!!

Complex predicates (= light verb constructions) - predicative nouns should be identified as events (a single predication, thus a single event);

Verbonominal predicates (predicate nominals) - nouns in predication; the whole predicate as a single predication (thus a single event).

JH: Discussion on 31.7.2023, see the 31.7.2023 meeting minutes: "We do not want to add a big number of nouns / adjectives to the lexicon OR represent them as verbs, unless we have a clear case of a deverbal noun (ending with -ní/-tí) / deverbal adjective (??). ... YES - morphological criterion!!!"

(Primary) nouns

(Primary) nouns typically denote entities (in reference), thus without a roleset. Examples:

Examples:

[cs] Muzeum zaplatí necelé 2 milióny korun za novou střechu.
[en] The museum will pay almost 2 million crowns for a new roof. ... the muzeum "museum" as an entity and it probably refers (depending on the context) to a particular NE (no rolesets)

Event nominals ending with -ní/-tí

Event nominals ending with -ní/-tí should be represented as coresponding verbs (e.g., přijíždění "arrival") if they denote a process (and thus they should get aspect annotation).

Examples: see above

Agentive nouns

Agentive nouns like teacher, driver, payer denote entities (persons) that conduct the respective process like teaching, driving, paying , definitely not processes. They are annotated using relation using the respective verb frameset (teach-01, drive-01, pay-01 - even in contexts different than teaching, driving, paying (an intention of the speaker to use just these nouns)). This they do not get aspect annotation.

[en] The bus driver turned the corner too sharply. (3-1-1-2 (2b))
... driver as ARG0-of the verb drive, no aspectual annotation (as it is a person=entity, not process)
[en] ... information that Obama uses tax payer money to do interviews and bus trips? 3-1-3-5 (1)
... tax payer is a person paying taxes, annotated as :ARG0-ofof the predicate pay-01; as such, no aspect annotation is expected ... BUT it appears in the guidelines example 3-1-3-5 (1) CONFLICT???

Entities as non-agentive participants

According to the guidelines, driver treated as an entity (ex.). Similarly, teacher is an entity; in addition, it is annotated using ARG0-of teach-01 (ex.) even in mentions where there is nothing about teaching :-((

DZ: See also issue 16 in UMR guidelines.

Examples:

[cs] jídlo "food"
... might be represented as ARG1-of of the predicate jíst-001 "eat"; however, it can be related also to other eventive concepts as, e.g., vařit "cook". Still, it is an entity (not a process) so no aspectual annotation!
[la] praefectus (en. prefect). Originally praefectus is a past participle of praeficio meaning 'to put in charge', but it also occurs as a substantive. This alternation is reflected in the grammatical case its dependents occur in. Examples:
- praefectus praetorio (en. praetorian prefect): praetorio in dative, because the verb requires dative case.
- praefectus annonae (en. prefect of the provisions): annonae unclear, could be either dative or genitive.
- praefectus Alexandreae et Aegypti, praefectus classis (en. prefect of navy), praefectus vigilum (en. prefect of the watchmen): genitive case is used, as in any nominal modification. --> Praefectus behaves like a substanstive, like driver/teacher (main difference: it has passive meaning).

Compare the annotation of food in 3-3-3 (1a) and in 3-2-1-1(6a)

Part 3-3-3

Unhealthy food.
(t/ thing
	:ARG1-of (e/ eat-01)
	:mod (h/ healthy
		:polarity -))

3-2-1-1(6a)

He gave the cat some wet food.
(g/ give-01
	:actor (p/ person
		:ref-person 3rd
		:ref-number Singular)
	:theme (f/ food
		:mod (w/ wet)
		:quant (s/ some))
	:recipient (c/ cat
		:ref-number Singular)
	:aspect Performance
	:modstr FullAff)

Other event nominals

Other event nominals (e.g., příjezd "arrival", jídlo "food") represent as entities (unless they are already covered by the valency lexicon). ... CONFLICT with UMR guidelines!!

Examples:

[cs] Bouře poničila střechu.
[en] The storm damaged the roads. (3-1-1-2 (1a))
... a process in predication (poničit "damage"),
... the :ARG0 of which is an event nominal (bouře "storm", in reference) indicating another process --> 2 events (the UMR rule) Czech: The eventive noun bouře "storm" does not have a frame/roleset in PDT-Vallex so it shall be annotate as non-eventive concept --> 1 event in Czech annotation _
Compare also to 3-2-1-1 (7g):
[cs] Bouře poničila elektrické vedení. [en] The storm damaged the power lines. (3-2-1-1 (7g))
... analysed as a single event (damage), with an non-eventive concept (storm) as a :force role (??? not in PropBank for damage-01) --> 1 event (both UMR for Eng and Czech)

ŠZ: other examples of possible processes: Christmas, wind, (financial) inflation

?? states in reference ... what to do with them? (sharpness https://www.wikidata.org/wiki/Q55433472 )

Complex predicates = light verb constructions

Complex predicates (= light verb constructions) should be identified as events (a single predication, thus a single event with single aspect annotation).

Example:

[en] _Where do you get this information from that Obama uses tax payer money to do interviews and bus trips?? (3-1-3-5 (1))
... Guidelines: "In this case the UMR concept interview-01 is created based on the eventive noun, while the verb is simply dropped. Note however the verb may contribute to the aspectual attribute of the concept." The same for the concept trip anchored as trip-03 -- :aspect Habitual in both cases.

Nouns in verbonominal predicates

Verbonominal predicates (predicate nominals) - nouns in predication; the whole predicate as a single predication (thus a single event).

Examples:

Entities (nouns) in predication are annotated as eventive concepts, these constructions are treated as an instance of the so-called non-verbal clauses with the abstract predicates (see alsa Rule 2b below).

[cs] Je vítěz/vítězem!
[en] She is the winner!
[en] This young man is a shaman. ? ... the predicative nouns vítěz "winner", shaman with the have-role-91 abstract predicate form a single event - so the aspect (value: state) should be annotated in such sentences.

BUT: Event nominals that occur in predicate nominal constructions (as in the following example) are not annotated as events (= they do not get aspect annotation in this phase, based on UMR example 3-3-1-1 (4))

[cs] To bylo zemětřesení.
[en] It was an earthquake. (UMR Guidelines, 3-3-1-1 (4))
[cs] Petr je šťastným otcem dvojčat.
[en] Peter is the happy father of twins!

BUT: Entities in other packaging than predication are not considered as events (thus do not get aspectual annotation).

[en] I met the happy father of twins. (based on UMR example 3-2-1-3 (3a))
... meet as a single event with aspect value 'Performance'; father is ARG1-of the predicate have-rel-role-92 (no aspect), similarly as in 3-2-1-3 (3a):

I met my father.
(m/ meet-03
	:ARG0 (p/ person
		:ref-person 1st
		:ref-number Singular)
	:ARG1 (p2/ person
		:ARG1-of (h/ have-rel-role-92
			:ARG2 p
			:ARG3 (f/ father)))
	:aspect Performance
	:modstr FullAff)

TODO Annotation of adjectives -- event adjectives vs. primary adjectives

RULE 2c: (internal criterion for Czech adjectives)

(Primary) adjectives typically denote state (in modification), thus without a roleset (chytrý "clever", vysoký "tall").

Deverbal adjectives (participles) should be mapped onto coresponding verbs as ARGx-of (e.g., přijíždějící "arriving"))

Other deverbal adjectives ... ??? should be mapped onto coresponding verbs as ARGx-of ??? )

Verbonominal predicates (predicate adjectives) - adjectives in predication; the whole predicate as a single predication (thus a single event)

JH: Be conservative with reification (reification = converting a role into a concept)!

JH: Discussion on 31.7.2023, see the 31.7.2023 meeting minutes: "We do not want to add a big number of nouns / adjectives to the lexicon OR represent them as verbs, unless we have a clear case of a deverbal noun (ending with -ní/-tí) / deverbal adjective (??). ... YES - morphological criterion!!!"

TODO (Primary) adjectives

(Primary) adjectives typically denote state (in modification), thus without a roleset (chytrý "clever", vysoký "tall").

DZ: I think that primary anchoring of states should be in SynSemClass. It should provide anchors for all processes and states, regardless whether they are packaged as predication, modification, or reference.

ML: We do not want to add primary adjectives (like chytrý, vysoký, clever, high) to the SynSemClass lexicon, thus they will be annotated as non-eventive concepts (contrary to proposal by Dan).

Compare the UMR examples:

[en] 3-1-1-3 (2a) The tall man ... (the UMR graph by ML, guidlines only says that this is not an event),
[en] 3-1-1-3 (2b) The man, who is tall ...
[en] 3-3-1-3 (1b) The doctor is tall.

 3-1-1-3 (2a) The tall
 (m/ man
	:mod(t/ tall))

The doctor is tall.
(h/ have-mod-91
	:ARG1 (d/ doctor
		:ref-number Singular)
	:ARG2 (t/ tall)
	:aspect State
	:modstr FullAff)

Deverbal adjectives (participles)

Deverbal adjectives (participles) should be mapped onto coresponding verbs as ARGx-of (e.g., přijíždějící "arriving"))

Examples:

[cs] Náměstí bylo plné tančících lidí.
[en] The square was full of dancing people.
... here we have a process (tančit "dance") in modification; it will be annotated as an event --> OK both rules
Similarly: [cs] válčící strany (lit. warring/fighting parties, enemies) ... the parties are in a real fight as a process, thus the word válčit is an event. --> OK both rules

TODO Other deverbal adjectives

Other deverbal adjectives ... ??? should be mapped onto coresponding verbs as ARGx-of ??? )

??? Derivations of adjectives from the l-participle are less productive. Therefore, spadlý “fallen”, vzrostlý “full-grown”, bývalý “former” etc. are not events.

??? Other deverbal adjectives are also not events: kulhavý “limping”, učenlivý “quick to learn”, představitelný “imaginable”, obstojný “passable”.

????? (e.g., podobný "resembing" --> podobat se "resembe" (based on the derivative morphology!)

Compare also the AMR guidelines, sect. Adjectives that invoke predicates

[en] the attractive man = the man who is ARG0-of attract-01 the same solution for Eng. adjectives ending with -ed (acquainted) the same solution for other types of endings, like -able (edible), -ful, ... OK for CZ (morphology!)
[en] to be+adjectives … often exist natural corresponding verbal predicates The soldier was aware of the battle. ... be aware (of X) --> realize-01 ... NOT for CZ ?
[en] adjectives without natural verbal predicates … create predicate!! be responsible (for X) --> responsible-01 (cause), responsible-02 (trustworthy), responsible-03 (duty) be nervous (about X) --> nervous-01 be serious (about X) --> serious-01 (no kidding), serious-02 (grave) ... NOT for CZ ??
[en] adjectives like sad, white, and free … as predicates if there is an implied event or proces ... NOT for Czech ??

Adjectives in verbonominal predicates

Verbonominal predicates (predicate adjectives) - adjectives in predication; the whole predicate as a single predication (thus a single event)

States/properties (esp. adjectives) in predication are annotated as eventive concepts, such constructions are treated as an instance of the so-called non-verbal clauses with one of the abstract predicates (see alsa Rule 2c below).

[cs] Ta kočka je moje.
[en] Tha cat is mine.
... state in predication (moje "my") as a non-verbal clause --> 1 event (OK both rules)
[cs] Moje dítě sedí na lavičce.
[en] My child is sitting on a bench.
... modification (můj "my") PLUS state in predication (sedět "sit")
--> 1 event (OK both rules)
[cs] Jirka je moc mazaný!
[en] Jirka is very crafty!
... (být) mazaný "(be) crafty" is a state (not process) in predication (hence event according to the UMR Guidelines), should be annotated using the verbonominal predicate have-mod-91, see also Rule 2c for adjectives --> just 1 event (UMR rules)
[cs] Kdo není mazaný, nevyhraje.
[en] He who is not cunning does not win.
... (být) mazaný "(be) cunning" is a state (not process) in modification (hence not event according to the UMR Guidelines) in the subject clause (with a verbonominal predicate have-mod-91), see also Rule 2c for adjectives, PLUS event predicate vyhrát "win" --> just 1 event (UMR rules) + a non-verbal clause
[cs] Jirka je moc mazaný!
[en] Jirka is very crafty!
... (být) mazaný "(be) crafty" is a state (not process) in predication (hence event according to the UMR Guidelines), should be annotated using the verbonominal predicate have-mod-91, see also Rule 2c for adjectives --> just 1 event (UMR rules)

BUT:

[cs] Překvapilo mě, jak chytrý byl.
[en] It surprised me how smart he was.
... subject clause is reference to a state (chytrý, "smart"), hence it is not an event; another ?state in predication (překvapit, "surprise"); být chytrý as a predication (verbonominal predicate, probably with být-007 or its -91 analogy) --> 1 events in Czech data Do we want to distinguishe between state and action in a dependent clause??<
- [cs] Jeho chytrost mě překvapila.
  [en] His smartness surprised me.
  ... the subject is reference to a state in reference (deadjectival nouns chytrý, "smart"), hence it is not an event.; another ?state in predication (překvapit, "surprise"), thus event --> 1 event by both rules, OK
[it] ambasciatore itinerante (en. ambassador-at-large? check wikipedia) itinerante as event or entity ??? ML: compound, thus entity (not process)?
[cs] Řekl jsem, že je chytrý.
[en] I said that he is smart.
... complement clause is reference to a state (chytrý "smart"), hence it is not an event according to the UMR rules
BUT there is a predication there (verbonominal predicate být chytrý "be smart"), thus will be annotated as a "non-verbal clause" with the have-mod-91 predicate (property) ... Do we want to distinguish between state and action verb
[cs] Moje dcera netouží být dospělá.
[en] My daughter does not long for being adult.
... complement clause is reference to a state (dospělý "adult"), hence it is not an event; another state in main predication (toužit "long"), thus event
BUT být dospělá is a verbonominal predicate, thus will be annotated as a "non-verbal clause" with the have-mod-91 predicate (property)???

Compounds with adjectives:

[en] firing squad (cs. "popravčí četa", from the UMR Guidelines) ... compound, thus entity (not process) --> OK both rules
BUT [en] floating hospital (from the UMR Guidelines) ... floating as an event ... BUT "The Floating Hospital" is a non-profit organization ??? https://en.wikipedia.org/wiki/Floating_Hospital --> OK both rules
[cs] Tančící dům (en. "Dancing House") ... a name of a building in Prague; the house is not really dancing, hence tančící does not denote a process in this context --> OK both rules
Similarly: [cs] létající pivovar ("gypsy brewery", lit. "flying brewery") ... a brewery company (society) working in different rental eqipped brewery houses - not an event --> OK both rules

TO READ

William Croft (2001) Radical Construction Grammar: Syntactic Theory in Typological Perspective
James Pustejovsky et al. (2005) The Specification Language TimeML.

About events expressed as MWE:

Julia Bonn et al. (2023) UMR Annotation of Multiword Expressions (sent by HH)
William Croft (2021): Eventive Complex Predicates and Related Constructions (draft from June 2021, sent by DZ)
Elena Paducheva, 1995

Previous discussions

Events as frame/bearing- concepts

DZ: The guidelines are not clear enough about the consequences for the annotation when a concept is or is not treated as an event. We can speculate that the intended difference is that events have :ARGX roles and non-events do not have them. At least it seems to indirectly follow from some parts of the guidelines. The guidelines also seem to suggest that the part of the lexicon that has frame files (i.e., the valency lexicon, defining the :ARGX participants for each predicate) contains concepts that are always events (eventive concepts), regardless the information packaging (see below); furthermore, I suspect that the UMR crowd assumes that eventive concepts are recognizable by a numeric suffix (e.g. -01), although no rule makes this assumption explicit, and one may need the suffix to disambiguate other concepts. In combination with the more clearly phrased parts of the guidelines this means that the valency lexicon contains only processes. States and entities, which may constitute events sometimes but not always, will be annotated with the help of abstract eventive concepts whenever they represent events. Even here the guidelines are not consistent because they say that two-place statives, such as English love, will be annotated without the abstract eventive concepts, that is, they have to be included in the valency lexicon. It makes the impression that after the nice typological introduction in Part 3-1-1 they slipped back to what is convenient for annotation of English, distinguishing verbs from adjectives (and not processes from states).

Action and state verbs

DZ: Note that the English state verbs are recognized by tests that are grammatical rather than semantic (you cannot use the progressive tense). Other languages may not have similar tests, or the tests may result in a different set of verbs. For example, in Czech, examples of state verbs (stavová slovesa) include stát "stand", mít "have", patřit "belong"; to stand would perhaps not be grammatically a state verb in English, but semantically it is a state both in Czech and in English.

There is a gray zone with verbs such as sleep or sit. They may not be "state verbs" according to the English grammar (it is grammatical to say I am sleeping or I am sitting on the couch) but semantically they seem to be states rather than processes. Sitting generally does not involve any activity or any changes; the only possible change is the end of this state by changing to another state (standing, lying, going...) Sleeping might be different, although also almost constant from the perspective of an observer. It may involve micro-sub-processes (snoring, turning, dreaming), and it may have a result at the end (compare it to recharging battery). So maybe sleeping is a process, even if not a frantic one.

ML: There is a simple Czech test for distinguishing state verbs (stavová slovesa) and action verbs (dějová / činnostní):

action verbs ... Co dělám? Běžím. "What am I doing? I'm running."
state/stative verbs ... Co se děje se mnou? Mám dost peněz. "What is happening to me? I have enough money."

The test illustrates also borderline cases like stát, as you can ask both questions Co dělám? Co se děje se mnou? Stojím. "What am I doing? What is happening to me? I'm standing."

válka_ "war"

DZ: See also issue 14 in UMR guidelines, as well as the document on entities. Named events are typically very complex processes that do not map easily to verbs with one actor and one patient. Nevertheless, they are processes, which means that the UMR guidelines want them annotated as events. At the same time, they have their class and types in Table 5 (the taxonomy of named entities), which means that the guidelines want them annotated as named entities. This is a conflict between different parts of the guidelines, which must be resolved somehow.

DZ: A possible compromise would be to annotate named events primarily as events, but with the :wiki attribute and :name relation used similarly to named entities. Hence [cs] válka "war" would be mapped to the same concept as the verb válčit "wage war" (possibly to SynSemClass class vec01002 (fight, bojovat)). The type war of class event in Table 5 of the guidelines would not be used as an abstract concept.

ML: How to add event annotation to event nominals? Probably should be represented as some relevant verb, e.g., válka should be the same concept as válčit (which is contained in the PDT-Vallex with ACT=ARG0, ADDR=ARG1, PAT=ARG1 (ARGs according to fight-01)), see below for two proposed annotations (válka mezi Indií a Pákistánem o Kašmír and Před 80 lety Německo přepadením Polska rozpoutalo druhou světovou válku.). But there is a problem with event nominals with no clear morphological verbal counterpart ... see the discusion on jídlo vs. food above and the UMR guidelines issue.

válka mezi Indií a Pákistánem o Kašmír "the war between India and Pakistan for Kashmir"
(v/ válčit-003
    :pdt-vallex "v-w7509f1"
    :wiki "Q198"
    :ARG0 (c/ country
        :wiki "Q668"
        :name (n/ :op1 "Indie"))
    :ARG1 (c/ country
        :wiki "Q843"
        :name (n/ :op1 "Pákistán"))
    :ARG2 (c/ local-region
        :wiki "Q43100"
        :name (n/ :op1 "Kašmír"))

Proposed solution (at least for the time being): We will stick to the PDT criteria and will not represent event nominals (others than those ending with -ní/-tí) as verbs (see the alternative rules and the discussion above)! Thus the example above is too advanced :-)

Examples:

[cs] druhá světová válka [en] WW II ... ?? should be anchored to https://www.wikidata.org/wiki/Q362
[cs] válka na Ukrajině ... ?? https://www.wikidata.org/wiki/Q110999040
[cs] Benátská noc (https://www.wikidata.org/wiki/Q11131287)

(w/ war
    :wiki "Q110999040"
    :name (n/ name
        :op1 "Ruská"
        :op2 "invaze"
        :op3 "na"
        :op4 "Ukrajině"))

Examples:

[cs] Před 80 lety Německo přepadením Polska rozpoutalo druhou světovou válku.
[en] 80 years ago, Germany started World War II by invading Poland.

(r/ rozpoutat-01
    :ARG0 (c/ country
        :synsemclass "???"
        :wiki "Q7318"
        :name (n/ :op1 "Německo"))
    :ARG1 (v/ válčit-01
        :synsemclass "vec01002"
        :wiki "Q362"
        :name (n2/ :op1 "druhá" :op2 "světová" :op3 "válka")))

DZ: Válčit is an event and it could have :ARGX relations but it does not have them because they are not expressed in the sentence. (One could deduce that Německo is one of the actors of válčit but the sentence does not say it explicitly.) Note that Wikidata Q7318 is the entry for Nazi Germany, not for the current country, which has Q183. ML: Thus the example above is too advanced :-)

CA: Some examples in Spanish/English of other named events with a link to Wikidata that could be also annotated as events:

[spa] España recuerda el papel de Adolfo Suárez en la Transición.
[en] Spain remembers the role of Adolfo Suarez in the Spanish transition to democracy
[spa] Esta situación no va a cambiar tras el anuncio del cese definitivo de la actividad armada de ETA.
[en] This situation is not going to change after the announcement of the definitive cessation of ETA's armed activity

Specific entities - Barack Obama

(Barack Obama, Barack Obama's cabinet)

Specific entities (instances) should be linked to Wikidata if they have an entry there. Regardless of information packaging. And even if the specific mention does not use the name (e.g., if the mention is president but it is clear that it refers to Barack Obama, it should be linked to Barack Obama). If the entity does not have an entry in Wikidata, we must create an entry for it in our local database, so that we can cross-reference that entity from various places even in different documents.

generic entities (sharp thorns, (any) bush’s thorns, (any) president)

We also want to anchor the generic type of the entity. This is relatively easy for generic mentions. If the sentence is One day I want to be president, we can link president to the corresponding Wikidata entry. Again, in the unlikely case that there is no such entry, we can create one in our local database. DZ: For specific mentions it is trickier. In theory, the specific Wikidata entry should contain information that gives us the type. But if we linked the word president to the entry of Barack Obama, we will not get a one-click way to the entry of president. According to Wikidata, Barack Obama is an instance of a human. In the list called position held, we learn that he was President of the United States, which is an instance of head of state (among other things of which it is an instance), and also a subclass of president (and also of politician). Given all this, maybe we want two anchors for a president in a specific context: one for the individual (Barack Obama) and another for the type that is closest to the word used in the mention (president). ML: If I mention president having in mind president Obama - is it still a generic entity? I would suppose this falls under specific entities? Or do you mean a context like Mr. Obama was a president.? But this will be annotated using the abstract predicate have-role-91 (based on 3-1-2 (1)), with Mr. Obama mapped to the Barack Obama and president to the respective Wikidata entry. Or do we want to distinguish that we mean US president in this case President of the United States?

Mr. Obama was a president.
(h/ have-role-91
      :ARG1 (p/ person :wiki "Q76"
            :name (n/ name :op1 "Mr." :op2 "Obama"))
      :ARG3 (p2/ president
      	    :wiki "Q30461")  ??? OR :wiki "Q11696"
      :aspect State
      :modstr FullAff)

Files

eventive-concepts.md

Latest commit

History

eventive-concepts.md

File metadata and controls

Summary: How to distinguish events and non-events

Quick HowTo for annotating Czech sentences:

Events vs. non-events

Consequences of being designated an 'event' in UMR

1. Verbs and abstract predicates (in references and modification)

1.A Stative verbs ... in the same way as action verbs (as events)

1.B "Abstract predicates" such as have-91, identity-91 ... in the same way as action verbs (as events)

1.C Reification ... in the same way as action verbs (as events)

Rolesets have-quant-91 AND have-degree-91

1.D "Discourse relation rolesets" ... as non-events

1.E Implicit rolesets ... it depends :-((

2. Nouns (connected to verbs)

2.A Agentive nouns (and other cases of inverse participant roles)

2.B Nouns denoting processes

2.C Deverbal nouns other than ní/-tí:

3. Deverbal adjectives other than active/passive participial adjectives:

Eventive concepts - Comments to the Guidelines

Semantic type

Information packaging

Eventive concepts in UMR

Eventive concepts – rules applied in annotation of Czech

Annotation of (Lexical) Verbs -- action verbs vs. state verbs

Action and state verbs

Verbonominal predicates

Complex predicates = light verb constructions

Secondary predication

Complement clauses

Nominalizations, Infinitives

Participles

TODO relative clauses ???

Note on causal relationships

Annotation of nouns -- event nominals vs. primary nouns

(Primary) nouns

Event nominals ending with -ní/-tí

Agentive nouns

Entities as non-agentive participants

Other event nominals

Complex predicates = light verb constructions

Nouns in verbonominal predicates

TODO Annotation of adjectives -- event adjectives vs. primary adjectives

TODO (Primary) adjectives

Deverbal adjectives (participles)

TODO Other deverbal adjectives

Adjectives in verbonominal predicates

TO READ

Previous discussions

Events as frame/bearing- concepts

Action and state verbs

válka_ "war"

Specific entities - Barack Obama