Skip to main content

Actions

Actions (or more fully, Action types) represent individual semantically disambiguated verbs.

Three valencies: Entity type valency, grammatical valency, and semantic valency

Actions acquire three kinds of valencies per any actant slot (subject, object 1, object 2; the data model is potentially extensible further, beyond trivalent verbs):

  1. entity type valency, which defines which entity type is allowed in the given actant slot;
  2. morphosyntactic valency,  which is a free text field defining the prepositions and grammatical cases, but uses a formalized notation (grammatical cases are noted with numbers 1-6, prepositions are in quote marks "", alternative is marked with a pipe "|"); and
  3. semantic valency, i.e. what kind of role the entity occupying the given actant slot has by implication (e.g., the subject of the Action “to travel” would have the semantic valency C “traveller”).

The main benefits from valencies are that they:

  1. guide coders in their choice of the correct Action (or towards creating a new one if none among the existing yet fits the syntactic and semantic definition);
  2. allow us to implement data validation features in a data collection interface;
  3. facilitate machine understanding of text, allowing semantic disambiguation of verbs based on their morphosyntactic valency (recognized by dependency parsing), and optionally, entity type valency (recognized e.g. through through named entity recognition).

Before assigning an Action the approved status, it should meet the following standards:

  • Its meaning is described in the “detail” field. (You will benefit from the use of printed or online dictionaries.)

  • It has the Action/Event Equivalent relation filled in with a Concept which has its meaning defined in its own “detail” field.

  • It has full information on the three valencies for each actant slot (including the explicit declaration of “empty” in the entity type valency, if no entity is allowed in that slot).

  • It has a reference to an external lemma collection ID (in DISSINET, the LiLa Lemma Collection).

  • If you have found a corresponding meaning among WordNet synsets:

    • It has the definition from WordNet copy-pasted in the “detail” field.

    • It has a Reference to this WordNet synset.

  • If you haven’t found a corresponding meaning among WordNet synsets:

    • You have defined the meaning yourself or based on dictionaries.

    • If there is any synset in WordNet which is a superclass of this (more specific) meaning, then an Action corresponding to the WordNet meaning is created (if Latin WordNet has it, then in Latin; if not, then in English), described, has a Reference to the WordNet synset, this and it forms the Superclass of this more specific Action you are working on.

  • All of this has been checked, i.e. it is not just a first draft of the Action that you still plan to come back to.

For something to be aligned with a synset definition in WordNet, it is not required that you accept its hypernyms or synonyms, just the definition needs to match.

Back to top