LRLM
Jump to navigation
Jump to search
LRLM is short for Left-to-Right, Longest-Match, the parsing strategy used by lt-proc of lttoolbox in analysis and bilingual modes, as well as hfst-proc. Basically, it means: read tokens from left to right, matching the longest sequence that is in the dictionary (like "greedy" matching of regular expressions).
LRLM is also used for structural transfer, so if input is a determiner followed by a noun, and there are rules for "det", "n", and "det n", the "det n" rule will match.
Another term for longest-match is Maximal Munch.
