Ambiguity packing is a well-known technique for enhancing efficiency
of context-free parsers. For constraint-based parsing (i.e. HPSG
parsing with typed feature structures), it has also been shown that
local ambiguity packing with efficient bidirectional subsumption test
improves the parser efficiency greatly. For deep linguistic processing
with large grammars (both parsing and generation), ambiguity packing
is almost mandatory nowadays.

In this talk, I will give a brief overview of the ambiguity packing
based on my recent work with a HPSG parser. Both theoretical and
empirical study of ambiguity packing and selective unpacking
algorithms will be shown. Furthermore, I will talk about the partial
parsing which directly benefits from the efficiency gain with ambiguity
packing and unpacking.