we only interpret the very last modifier word; that is, things like
"sehr weit" are handled just like "weit"
"ganz weit nach" should really be interpreted like "ganz nach", not like "weit nach"
at the same time, "ganz bisschen" *should* be interpreted like "bisschen"
in other words: max followed by strong should remain strong.
by default, an IU
- is complete if the last grounding unit is complete,
- is upcoming if the first grounding unit is upcoming,
- is ongoing if inbetween those two states
- is null if there are no grounding units.
return a list of word and/or unit tokens extracted from a linked token list
if both words and units are to be returned, then the word token
will always precede the unit tokens belonging to this word
the algorithm iterates through the tokens *against* temporal order
(as this is the natural order of the tokens)
and finally reverses the list before it is output.
a first, unfinished implementation to extract unscoped predicate logic statements
of this rmrs, i.e. to get rid of the predicate-argument factorisation
calculate the WER between two lists of words based on the levenshtein distance
code adapted from http://www.merriampark.com/ldjava.htm
list a is taken as the reference for word error rate normalization
only if recognition is final, output the results of the
underlying deltifier; otherwise return an empty list;
which will then contain a list of all ADD edits for the
words in the final result.
only if recognition is final, output the results of the
underlying deltifier; otherwise return an empty list
which will then contain the words in the final result.
this prototype will allow the user to generate utterances like
"<Action> bitte das <Color> <Piece>" (for now, later maybe as below)
first, something like:
╱ nimm ╲ ╱ rote ╲ ╱ Kreuz ╲
(Jetzt?)
this prototype will allow the user to generate utterances like
"<Action> bitte das <Color> <Piece>" (for now, later maybe as below)
first, something like:
╱ nimm ╲ ╱ rote ╲ ╱ Kreuz ╲
(Jetzt?)
GreifarmActor is the NLU and action management component for the Greifarm prototype
(as the Greifarm prototype is not really a "dialogue system", but rather
a speech-controlled application, we avoid the term "dialog manager")
This class consists of the following components, which all live in their
own subclasses:
GreifarmActor: the incremental processor (PushBuffer) which handles incoming
IUs and initiates the corresponding processing.