parvvaresh / evaluator-targoman

We at the Telecommunication Research Center decided to test and evaluate the Tergoman machine translation system. This evaluation is done by 6 algorithms

evaluate Targoman

We at the Telecommunication Research Center decided to test and evaluate the Tergoman machine translation system. This evaluation is done by 6 algorithms:

• WER
• BLEU
• WER
• GLEU
• METEOR
• CHRF(3)

special thanks to my dear teammate 🙏 : Elham Ghasemi Faeezeh Gholamrezaee Mahsa asghari

---

about Targoman

Targoman, a bilingual machine translation system, was developed in 2011 by a group of researchers at the Natural Language Processing and Machine Learning Laboratory at Amirkabir University of Technology. The project was initiated under the supervision of Dr. Shahram Khadivi and received support from the Telecommunication Research Center of Iran.

Initially, Targoman was based on open-source tools for statistical machine translation and involved the development of algorithms and language processing tools specifically tailored for the Persian language. It started as a research and development project. From 2013 to 2015, the system underwent revision and localization, with the translation engine and all its components being redesigned and implemented from scratch.

The native translation engine developed within Targoman achieved remarkable stability and substantial improvements in speed, offering a 270% increase in translation speed and a 16-fold reduction in resource consumption while maintaining precision, even at a detailed level.

In addition to the translation engine, various language processing tools were developed, including preprocessing, post-processing, word root extraction, named entity recognition, transliteration, and language modeling. These tools were designed to be language-independent for versatile use in language processing projects.

Starting from late 2016 to early 2018, Targoman's maintenance and enhancement were taken up as part of a native development project. This project was carried out through a technical and executive partnership between two companies, "Pardazesh Hoshmand Targoman" and "Vira Afzar Adan."

Before August 2022, Targoman offered English-to-Persian and Persian-to-English translation services using a neural translation engine developed jointly by "Pardazesh Hoshmand Targoman" and "Vira Afzar Adan." Furthermore, through collaboration with a European company, Targoman enabled translations between nine of the world's most spoken languages.

As the collaboration with the European company ended, Targoman embarked on developing its proprietary translation engines for various languages. Bidirectional translations will be supported for the following language pairs:

• English to Persian
• English to Russian
• Persian to Russian
• English to German
• Persian to German
• English to French
• Persian to French
• English to Spanish
• Persian to Spanish
• English to Turkish
• Persian to Turkish
• English to Japanese
• Persian to Japanese
• English to Arabic
• Persian to Arabic

Targoman aims to provide efficient and accurate machine translation services across these language pairs.