Rare bug introduced in multi-word verb processing

[philip-schrodt]

In running the checks on the new multi-word verb processing code on three GigaWord files, I ran into a single instance (a report on a cricket match, of course) where the event code somehow got a tuple (rather than a string) that was used in multi-words: this caused a crash in PETRWriter around line 85. I trapped this and converted it to a '010' code, but obviously it is indicative of something odd going on somewhere. What is really weird is that neither of the verbs in the story were involved in multi-word combinations (suggesting some Pythonic equivalent to a pointer error, and we all know just how much fun those are).