CiteSeer: Citation Matching


“CiteSeer” is a relational dataset of publication citations for Alchemy, the original dataset is available on their website. This version has modifications to work with BoostSRL; including the associated background, train/test folders, and the positives/negatives/facts.

Three targets are possible:

  • infield_fauthor
  • infield_ftitle
  • infield_fvenue


Download: (1.62 MB)

  • md5sum:


  • sha256sum:




  1. After downloading, unzip


  1. If you’re using a jar file, move it into the CiteSeer directory:

mv (jar file) CiteSeer/

  1. Learning:
  • Learning may take a long amount of time on all three targets.
  • java -jar BoostSRL.jar -l -train train/ -target infield_fauthor,infield_ftitle,infield_fvenue -trees 10
  1. Inference:
  • java -jar BoostSRL.jar -i -test test/ -model train/models/ -target infield_fauthor,infield_ftitle,infield_fvenue -trees 10


// Parameters
usePrologVariables: true.
setParam: treeDepth=4.
setParam: nodeSize=2.
setParam: numOfClauses=8.
setParam: numOfCycles=8.
// Modes & Bridgers
mode: center(+bib, +pos).
mode: center(+bib, -pos).
mode: firstin(+bib, +pos).
mode: firstin(+bib, -pos).
mode: firstnonauthortitletkn(+bib, +pos).
mode: firstnonauthortitletkn(+bib, -pos).
mode: followby(+bib, +pos, #token).
mode: hascomma(+bib, +pos).
mode: hascomma(+bib, -pos).
mode: haspunc(+bib, +pos).
mode: haspunc(+bib, -pos).
mode: infield_ftitle(+bibpos).
mode: infield_fauthor(+bibpos).
mode: infield_fvenue(+bibpos).
mode: isalphachar(+token).
mode: isdate(+token).
mode: isdigit(+token).
mode: lastinitial(+bib, +pos).
mode: lastinitial(+bib, -pos).
mode: lessthan(+pos, -pos).
mode: lessthan(-pos, +pos).
mode: next(+pos, -pos).
mode: next(-pos, +pos).
bridger: next/2.
mode: nextbibpos(+bibpos, -bibpos).
mode: nextbibpos(-bibpos, +bibpos).
nextbibpos(BP1,BP2) :- isbibpos(BP1, B,P1), isbibpos(BP2,B,P2), next(P1,P2).
mode: isbibpos(+bibpos, -bib, -pos).
mode: isbibpos(+bibpos, +bib, -pos).
mode: isbibpos(+bibpos, -bib, +pos).
mode: isbibpos(-bibpos, +bib, +pos).
bridger: isbibpos/3.
mode: token(+token, +pos, +bib).
mode: token(+token, -pos, +bib).
mode: token(-token, +pos, +bib).