search

jedgusse / project_lorenzo

0 stars 0 forks source link

"Unresolued link" #7

Open emanjavacas opened 7 years ago

emanjavacas commented 7 years ago

In the pl data, there are some more than 5500 instances of the bigram "unresolued link" (funnily enough with "u"), this has enough effect to leak into the generation.

Original text

data/pl/Alanus_de_Insulis.De_VI_alis_cherubim.xml-562-<s>
data/pl/Alanus_de_Insulis.De_VI_alis_cherubim.xml:563:unresolued    NAM <unknown>
data/pl/Alanus_de_Insulis.De_VI_alis_cherubim.xml-564-link  VBE <unknown>
data/pl/Alanus_de_Insulis.De_VI_alis_cherubim.xml-565-Ad    PRE ad
data/pl/Alanus_de_Insulis.De_VI_alis_cherubim.xml-566-explanationem SUB explanatio
data/pl/Alanus_de_Insulis.De_VI_alis_cherubim.xml-567-huius PRO hic3
data/pl/Alanus_de_Insulis.De_VI_alis_cherubim.xml-568-figure    SUB figura

Generated

corruerunt, sed cum erunt rex Edom( unresolued link): Sumebis ergo intelligere, nisi Messacebriam<eos>