I need a help with optimizing a transliteration model in Python.
Optimization has the following steps:
1. Use the Viterbi function based on the function
implement forward_scores by adding the sum through a
replace maximization(Viterbi) and remember which of the three values is greatest
was. (Edit Distance) Then you have to implement the extraction of the best aligning.
2. Modify the function get_transliterations so that they are
additionally calculated with the Viterbi algorithm the best alignment
and returns this. This is then also output.
3. The transliteration system that must implement later can
do not handle insertions. Therefore, you should next in because
Alignments eliminated eliminate all inserts
Insertion: z merged with the previous pair x: y to x: yz. If
an insertion is at the beginning, she will instead use the
subsequent couple merged. For example, the sequence becomes: x a: a: b
: f c: c: d replaced by a: xabf c: cd. (N-grammmodell-training for transliterations)
4. Implement the transliteration system. It uses
a NGramm model and thus takes into account context information in the
Contrary to the mining model. The parameters of the NGramm model become off
estimated previously calculated Viterbi alignments. For that
use the program train-ngram-model.py as a template.
5. Endstep is a program that defines the parameters of the
NGramm model reads in and the list of
Transliterative word pair candidate filters: For each word pair becomes
the probability of the best transliteration and the
Probability of the second element of the word pair calculated.
There are pseudocode for almost every step and even model for optimization.
The entire project needs to be completed within two weeks, there should be deliveries of work in progress at least every two to three days (the steps as displayed)
Бюджет: 6000 руб.
Вид предложения: Удаленная работа (разовый заказ)
Добавлено: 29.10.2018 в 22:18