Rebanking

Superlatives

Comparatives

The CCGbank analysis seems like an easy target to beat. Scanning through the instances of "more" in CCGbank, we get this by file 22:

Current|JJ|N/N PCs|NNS|N are|VBP|(S[dcl]\NP)/(S[adj]\NP) more|RBR|S[adj]\NP than|IN|(((S[adj]\NP)/(S[adj]\NP))/((S[adj]\NP)/(S[adj]\NP)))\(S[adj]\NP) 50|CD|((S[adj]\NP)/(S[adj]\NP))/((S[adj]\NP)/(S[adj]\NP)) times|NNS|(S[adj]\NP)/(S[adj]\NP) faster|JJR|S[adj]\NP and|CC|conj have|VBP|(S[dcl]\NP)/NP memory|NN|N/N capacity|NN|N 500|CD|((S[adj]\NP)/(S[adj]\NP))/((S[adj]\NP)/(S[adj]\NP)) times|NNS|(S[adj]\NP)/(S[adj]\NP) greater|JJR|S[adj]\NP than|IN|((S[adj]\NP)\(S[adj]\NP))/NP their|PRP$|NP[nb]/N 1977|CD|N/N counterparts|NNS|N .|.|.

Check out the categories required for "more" and "than". Attached are text files showing the 35 categories assigned to "more" in 02-21, and the 90 categories assigned to "than". Surely we can do better than this.

The XTAG grammar write up of comparatives is quite extensive: http://www.cis.upenn.edu/~xtag/tech-report/node197.html The Cambridge Grammar of English section on comparatives is 70 pages long...

MWE (Multi Word Expressions)

http://mwe.stanford.edu/resources/

Verb Phrase Ellipsis

Auxiliaries and modal verbs triggering VPE should get S\NP as lexical category (there are a couple of exceptions). Vice versa any aux or modal with S\NP should be an instance of VPE. As we have annotated all VPE in the WSJ, we can check/correct these cases.