CyclicCheck < Main

Tags: view all tags
---+Grammar Engineering Frequently Asked Questions

---++The LKB says *"Cyclic check found cycle at ..."*. What does this mean and how do I debug it?

The feature structures are supposed to be directed *acyclic* graphs, that is, you can't follow
a path and get back to the same place.  The following constraint is cyclic, and therefore
illegal:

<pre>
foo := basic-verb-lex &
  [ SYNSEM #synsem & 
      [ LOCAL.CAT.VAL.COMPS < #synsem > ]].
</pre>

That would seem an unlikely thing for anyone to type.  A more frequent cause of cycles is
in diff-lists.  If you are getting this error, there's a reasonably good chance that you're
over-identifying the values of features that are involved in a diff-list append.  Alternatively,
a type and its supertype might be saying inconsistent things about the length of the
diff-list. Check features such as =RELS=, =HCONS=, =SLASH=, =QUE=, and =REL=.

---+++Related topics

   * [[DiffList][What's a difference list, and why do we use them?]]
   * [[MissingRels][Some of my relations/qeqs aren't showing up in the MRS for the whole parse. Why not?]]

---

[[GrammarEngineeringFAQ][Back to FAQs page]]

[[http://courses.washington.edu/ling567/][Back to main course page]]

-- Main.EmilyBender - 18 May 2005
Topic revision: r1 - 2005-05-18 - EmilyBender