PERFORMANCE TEST OF STATISTICAL MACHINE TRANSLATION IN TRANSLATING ENGLISH TO INDONESIA

• Main Author: RAMDHANI (NIM 23204068), MOHAMAD
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/10917
• Institution: Institut Teknologi Bandung

<p align="justify">One of the natural language processing applications is translation machine, which can translate sentences from one natural language to another natural language. Statistical machine translation is a technique of translation based on statistics by using learning model from a group of sentences (corpus) as data training.<p align="justify"><p>Statistical machine translation works by finding biggest probability or possibility of a sentence to be translated in target language when the source language sentence is given. This translation machine is composed using modeling of statistic from Bayes theorem, that consist of language model, translation model and decoder.<p align="justify"><p>There are three studies conducted in this research : (1) to comprehend the working of statistical machine translation, (2) to understand how far translator bases on this statistic is used to conduct translation from English language to Indonesian language, and (3) to conduct software integration open source from component of statistical machine translation. Testing is conducted with software integrated from statistical machine translation based on examined aspects, testing techniques and system configuration, and analysis of testing result.<p align="justify"><p>Based on testing, there are a few results. First, that statistical machine translation can remember all sentences which have been trained previously, second, it can be trained to repeat in innovating its own knowledge, third, it can translate sentences which have not yet been introduced by using sample sentence examples which have been studied , and four, as the data are more intensively trained, the machine translation will progressively give even better performance.