Default Configurations - quhfus/DoSeR-Disambiguation GitHub Wiki

###Default Parameters for DoSeR (SIGIR Paper)

  1. PageRank
  • Iterations=75
  • Jump Probability alpha=0.09 Step1
  • Jump Probability alpha=0.2 Step2
  1. Disambiguation Algorithm
  • margin_1 = 0.5
  • margin_2 = 0.3
  • lambda = 1.57
  1. Word2Vec
  • Architecture=Skip-Gram
  • Vector-dimensions=400
  • Window-Size=8
  • Min-Occurrences=1
  • No sampling
  1. Doc2Vec
  • Architecture=PV-DM
  • Vector-dimensions=1000
  • Window-Size=8
  • Min-Occurrences=3
  • Sampling in default configuration

###Default Parameters for DoSeR (ESWC Paper)

  1. PageRank
  • Iterations=50
  • Graph reduction=25%
  • Jump Probability alpha=0.1
  1. RandomWalk Corpus Creation RDF-KB
  • Jump Probability alpha=0.1
  • Number of Walks=50000000
  1. Word2Vec
  • Architecture=Skip-Gram
  • Vector-dimensions=400 (as suggested by Mikolov et al.)
  • Window-Size=8
  • Min-Occurrences=1
  • No sampling

###Some Links to Gerbil Result Sheets (ESWC Paper)

  1. CoreKB: DBpedia (without categories) - Baseline Gerbil Sheet
  2. CoreKB: DBpedia (With categories), Optional KB: Wikipedia - Gerbil Sheet (MSNBC-OLD)

###Links to other entity disambiguation systems:

  1. Wikifier
  2. AIDA
  3. AGDISTIS