Designing the SRL system,

it will be able to label nodes, and to label semantic relations,
but it will be not able to predict semantic relations

sematic role labeling

2h meeting

sematic role labeling:

how to predict the semantic role relation:

predict the name of the relation.

predict the relation it-selve

1.- tree distance knn

matching with the k neigbours the best fit.

2.- three kernels:

analizing how looks each posible subtree as a target

preprocessing samples before give them to the learning machines

There is a large work done on preprocessing samples before give them to the learning machines:

·         Remove noise

o   Algorithms for detecting noise samples based in knn algorithms.

·         Add noise

o   Small noise produces a better performance in neural networks (and maybe also in other algorithms).

·         Re-structure the dimensionality and distance metrix.

o   Nahanalobis distance.

o   Scaling the data: It give an improvement in SVM machines

o   Kernels: increase dimensions.

o   Genetic kernel (GK SVM)

o   Removing features:

§  removing dimensions (feature selection)

·         information gain (the best)

·         mutual information

·         x2 statistic chi (second best)

·         term strength

§  principal component analysis.

§  neighborhood component analysis

·         Re-sampling:

o   Under-sampling:

§  Randomly

§  Inconsistent data

§  Duplicate data

§  Removing noise (bis)

o   Over-sampling:

§  Randomly

§  SMOTE

§  Border SMOTE-1

§  Border SMOTE-2

§  Adding noise (bis)

§  Give more weight to hard samples.

·         Windowed data:

o   In some cases context information increases the accuracy.

Split features: in text categorization words can split using the morphology features: Morfesor. 

meeting with martin.

meeting with martin.

long explanation of kNN implementation algorithm,

dudani vote method  -> depend on the distance of the most fard point

the size of k (knn) can change if are several samples at the same distance.

our dataset is base on trees distance (there are not vectors), 

is about question answering, the original data are questions, and the labels are about what is looking for.

all is implemented in c++ 

QBank Manager class recover the original question.

knn with kernel tranformation --> changin the way to measure distance

kernelizing linear classificators