PPLRE Nearest Neighbor Algorithm

From GM-RKB
Jump to navigation Jump to search

The PPLRE Nearest Neighbor Algorithm is a Binary Relation Recognition Algorithm developed for the PPLRE Project that is based on the Nearest Neighbor Algorithm in that it locates nearest pairs of Entities and identifies them as being in an OP() or PL() relation.

  • Context:
  • Example(s):
    • For the sentence "[PROTEINa virA] encodes a [LOCATIONa membrane-bound] sensor kinase protein and [PROTEINb virG] encodes a [LOCATIONb cytoplasmic] regulator protein." (ref: PPLRE Corpus 8611.a.2) the algorithm can clearly extract the two correct relations (PL(PROTEINa, LOCATIONa) and PL(PROTEINb, LOCATIONb) and with a distance containt also avoid the other two incorrect permutations.
    • For the sentence "In this study, [ORGANISM Escherichia coli] [PROTEIN TonB] was found to be distributed in sucrose density gradients approximately equally between the [LOCATIONa cytoplasmic membrane] and the [LOCATIONb outer membrane] fractions, while two proteins with which it is known to interact, [PROTEINb ExbB] and [PROTEINc ExbD], as well as the [PROTEINd NADH] oxidase activity characteristic of the [LOCATION cytoplasmic membrane], were localized in the [LOCATION cytoplasmic membrane] fraction." a word-based distance function would identify nine relations: OP(ORGANISM,PROTEINa), OP(ORGANISM,PROTEINb), OP(ORGANISM,PROTEINc), OP(ORGANISM,PROTEINd), PL(PROTEINa,LOCATIONa), PL(PROTEINb,LOCATIONa), PL(PROTEINc,LOCATIONa), PL(PROTEINd,LOCATIONa). Notice that one of the relations is missed (PL(PROTEINa,LOCATIONb)) bust also that many of the invalid permuations are not returned.
  • See: PPLRE Project, Nearest Neighbor Algorithm.