Lexically-based Relation Mention Recognition Structure

A Lexically-based Relation Mention Recognition Structure is a linguistic relation recognition structure that is a textual pattern (which uses only lexical information (e.g. words, POS and/or concept type) to detect and classify a relation mention).

• AKA: Textual Relation Recognizer.
• Context:
  • It can be expressed as a Shallow Parse Pattern, such as “NP is a part of NP”.
  • It is a type of Mention Pattern.
  • It can range from being a Textual Relation Pattern to being a Spoken Relation Pattern.
• Example(s):
  • a Wrapper Pattern.
  • a Five-Tuple Lexically-based Relation Recognition Classifier.
  • a Hearst Pattern, such as:
    • [<NAMED ENTITY>-based <NAMED ENTITY>] can find Organization_Headquarter_Location() relation mentions; but can also misclassify sentences such as "Florida Orange-based Heinz Marmalade".
  • [Ø | <PROTEIN> | activity, found, in |<LOCATION> | Ø] can find the PL() relation in PPLRE Corpus 357.a.1.
  • “NP is a part of NP” for the PartOf Relation.
  • “NP is a kind of NP” for the IsA Relation.
  • “NP is made of NP” for the MadeOf Relation.
  • “NP is used for VP” for the UsedFor Relation.
  • “NP can VP” for the CapableOf Relation.
  • “NP wants to VP” for the DesireOf Relation.
  • "You make NP by VP” for the CreatedBy Relation.
  • "An example of NP is NP” for the InstanceOf Relation.
  • “NP is part of NP” for the PartOf Relation.
  • “NP is AP” for the PropertyOf Relation.
  • "The effect of VP is NP|VP” for the EffectOf Relation.
    • [<Class Noun>(plural) such as <Noun>, and <Noun>] can find Is_A() relations. (Hearst, 1992).
  • …
• Counter-Example(s):
  • • an Embedding-based Relation Mention Recognition Structure.
    • a Dependency Relation Recognition Classifier.
    • a Text Graph Relation Recognition Classifier.
• See: Snowball Algorithm, DIPRE, ConceptNet Semantic Relation Ontology.

---

---

References

1992

• (Hearst, 1992).

---