KonradHoeffner / cubeqa

CubeQA—Question Answering on Statistical Linked Data
https://aksw.org/Projects/CubeQA.html
GNU General Public License v3.0
20 stars 5 forks source link

sector is not found in AlgorithmTest #37

Closed KonradHoeffner closed 8 years ago

KonradHoeffner commented 8 years ago

find out why the sector is not found in AlgorithmTest even when boostString is set to 0.1 (in ObjectPropertyScorerTest it works)

KonradHoeffner commented 8 years ago

The problem is that for "How much money was given to strengthen civil society in Yemen?", the phrase "strengthen civil society" is never considered, only "strengthen civil society in Yemen", which barely goes over the maximum phrase length of 30 and contains "in Yemen" as well.

There are multiple ways to fix this sorted ascending by time investment and effect:

  1. Set the phrase length to 40
  2. Create a place detector to catch "in X" phrases for places (how to know when the phrase ends, parsing?)
  3. Replace the recursive parse tree traversal algorithm with a different one

I am looking through the benchmark and favor option 2 right now.

KonradHoeffner commented 8 years ago

Option 2 it is. AlgorithmTest#testAnswer() now runs through successfully.