i have developed a small dependency treebank and now i want to run maltparser on it.i am working on urdu which is written in right-to-left (RTL) form. when i try to train parser it gives following error "The input file 'train_data.conll' contains a column where the value is an empty string. Please check your input file."

what i understant is " the parser reads from left to right. when it compares the format in xml file and conll file it doesn't matches. that's why it gives this error. am i getting it right?

also i tried to change xml file and then run the command to train the parser. but xml file automatically changes back to it's orignal form. how to remove this error?

  • The way the question asked, it may be difficult to reproduce your problem. Consider adding some data (maybe, linked to some file hosting if they are large) so that someone who is familiar with the tool you're using could reproduce your case. Don't forget to specify versions of the software and libraries you're using as the error may be version-specific. Dec 26, 2015 at 12:32
  • Other than that, usually data corpus contains Unicode-encoded strings. This means that the characters have specific order which is not linked to RTL or LTR writing system. E.g., first is first, regardless of where it is written. See this answer for more details. Dec 26, 2015 at 12:32


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.