avinashvarna
commented
7 years ago I haven't looked at the PCRW algorithm mentioned in Paper 4 in your doc (Amrith Krishna, et al.), but is that simple enough to implement in this context?
Otherwise, can we split this into two stages of enforcing local constraints vs global constraints?
- Enforce local constraints such as samasa constituents, upasarga, etc.:
Start with lexical graph G to create morphological graph MG with local constraints
for node in G: for tag of node: for child in node.children: for child_tag of child: if tag, child_tag satisfies local constraints: MG.add_edge(((node, tag), (child, child_tag))) - Check if global constraints such as visheShaNa/visheShya, etc. are satisfied
for path in MG.shortest_simple_paths(start, end): if path satisfies global constraints: yield path
It will involve creating a larger graph and thereby potentially increase the no. of paths. However, we can then use a simpler checker to filter out invalid paths. Caveat: The computational overhead of finding paths in the larger graph might be significant, and we may end up throwing away many of the paths in the filtering step
I will think about this and see if there are better solutions
kmadathil
gasyoun
Some thoughts - not aiming to handle all possibilities right away
Input - Top n paths from Lexical Graph output from L2.
Step 1 - Karaka assignments
Step 2 - Use a constraint solver to pick paths (with tags) that satisfy all constraints
If this is possible to do using the Lexical Graph output of L2 directly, that might help. That would be sort of a constrained path search. I fear, though that that would make things worse in comparison to working on paths. Perhaps this can be decomposed into a set of graph problems that will simplify things?