The py_qualityestimator's intro from mikeboydbrowne

py_qualityestimator's Introduction

There are three Python programs here (-h for usage):

./default performs clustering on the training data
./baseline performs k-nearest-neighbor
./grade compares the predicted score with the actual score and returns the accuracy

The first two commands are designed to work in a pipeline with the last. For instance, this is a valid invocation:

./baseline | ./grade

The data-train/ directory contains

+ `es-en_score.train`, which contains the score assigned to each sentence (one score per line)
+ `es-en_source.train`, which contains 1050 Spanish sentences
+ `es-en_target.train`, which contains 1050 English translations
+ `train_features`, which contains the values of each of the 17 features for each sentence

as well as the feature-list, which contains the list of features being used.

The data-dev/ directory contains:

+ `es-en_score.dev`, which contains the score assigned to every alternate sentence in the test data, i.e. 225 scores. This is what is used to develop locally.

The data-test/ directory contains:

+ `es-en_source.test`, which contains 450 Spanish sentences
+ `es-en_target.test`, which contains 450 English translations
+ `test_features`, which contains the values of each of the 17 features for each test sentence

Recommend Projects