Closed namednil closed 5 years ago
Unclear why this occured. Below are the results on the dev set.
From Mario's e-mail: With lowering:
{"n": 1049,
"exact": 2,
"tops": {"g": 1049, "s": 1325, "c": 1045, "p": 0.7886792452830189, "r": 0.996186844613918, "f": 0.8803706823925863},
"labels": {"g": 0, "s": 0, "c": 0, "p": 0.0, "r": 0.0, "f": 0.0},
"properties": {"g": 0, "s": 0, "c": 0, "p": 0.0, "r": 0.0, "f": 0.0},
"anchors": {"g": 17188, "s": 16990, "c": 15617, "p": 0.9191877575044144, "r": 0.9085990225738888, "f": 0.9138627187079408},
"edges": {"g": 25262, "s": 23724, "c": 14474, "p": 0.6100994773225425, "r": 0.5729554271237431, "f": 0.5909443514473522},
"attributes": {"g": 1155, "s": 0, "c": 0, "p": 0.0, "r": 0.0, "f": 0.0},
"all": {"g": 44654, "s": 42039, "c": 31136, "p": 0.7406455909988344, "r": 0.6972723608187397, "f": 0.7183048227653905},
Without lowering:
{"n": 1049,
"exact": 2,
"tops": {"g": 1049, "s": 1327, "c": 1045, "p": 0.787490580256217, "r": 0.996186844613918, "f": 0.8796296296296297},
"labels": {"g": 0, "s": 0, "c": 0, "p": 0.0, "r": 0.0, "f": 0.0},
"properties": {"g": 0, "s": 0, "c": 0, "p": 0.0, "r": 0.0, "f": 0.0},
"anchors": {"g": 17188, "s": 16990, "c": 15725, "p": 0.9255444379046498, "r": 0.9148824761461485, "f": 0.9201825735853474},
"edges": {"g": 25262, "s": 23082, "c": 12904, "p": 0.5590503422580365, "r": 0.51080674530916, "f": 0.5338408075459209},
"attributes": {"g": 1155, "s": 0, "c": 0, "p": 0.0, "r": 0.0, "f": 0.0},
"all": {"g": 44654, "s": 41399, "c": 29674, "p": 0.7167805985651827, "r": 0.6645317328794733, "f": 0.6896679953052189},
"time": 24.02891492843628,
"cpu": 23.902485000000002}
Row 18 in our spreadsheet contains a new UCCA model from last night. I accidentally didn't use edge lowering in post-processing. When I did, I got this instead: