For example, Cer, Daniel M., et al. "Parsing to Stanford Dependencies: Trade-offs between Speed and Accuracy." LREC. 2010. :

enter image description here

All 5 presented constituency parsers generate POS tags. Do constituency parsers always generate POS tags?

  • 2
    The question is kinda invalid. Constituency parsers read POS-tagged words, and then generate a structure that accommodates them.
    – prash
    Apr 1, 2014 at 22:58
  • 1
    @prash, I am also confused. Your comment suggests that dependency parsers do not first tag the words before generating the structure. What is the standard order of operation: first tag the words, then generate the structure, or generate the structure, then tag the words. I would think there is no major difference across dependency and constituency in this area. I would think that both tag the words first and then generate the structure. Apr 2, 2014 at 15:51
  • @TimOsborne I am not as confident of dependency parsers as I am of constituency parsers. That's why I limited the scope of my comment. I know for a fact that some dependency parsers need POS-tagged words, but I don't know if they all do.
    – prash
    Apr 2, 2014 at 16:29
  • My guess is that the table does not give times for those parsers because that phase of the process is not isolated like it is for the dependency parsers. I can't see the connection between the table and your question.
    – curiousdannii
    Jul 9, 2014 at 6:48
  • 2
    I don't think this question is invalid. Most, but not all, constituency parsers perform POS tagging while generated constituents. Most (but again, not all) dependency parsers require POS tags as input before they can produce dependency structures. So, I think it's a completely valid question as to whether constituent parsers always produce POS tags.
    – dmcc
    Jan 14, 2015 at 3:01


Browse other questions tagged or ask your own question.