Phramer is a free and open-source decoder for statistical phrase-based machine translation, designed to analyze large amounts of multilingual data and provide accurate translations with speed and efficiency.
Phramer is an open source statistical phrase-based MT decoder that aims to deliver a great decoding experience for Statistical MT research. This project is built to be extensible, with source-code available for the creators, and written in a favorable programming language without segmentation faults and non-scripting languages. It is also designed to be compatible with Pharaoh.
As a software reviewer, you may ask why you should prefer Phramer instead of Pharaoh. Firstly, Phramer is an open-source program with minimal limitations compared to Pharaoh, which is currently available under much restrictive licensing. Secondly, Phramer provides the source code availability not present in Pharaoh. Thirdly, Phramer is cross-platform, written in Java, and tested in Linux and Windows 2000/XP. Fourthly, Phramer is faster; it is currently 1.5 - 3 times faster than Pharaoh 1.2.3 (not including data files loading time), and its performance is continually improved upon.
Phramer also features a modular architecture with hooks provided for enhanced extensibility. It is evident that Phramer has more features than Pharaoh 1.2.3 (not including undocumented ones). Additional parameters can also be seen in the list of features.
For you to use Phramer, you will need Java 5/1.5 or higher. A newer version of Phramer is now available, fixing some bugs and doubling the speed of the previous version under regular conditions. Multi-threading for both decoding and MER training is introduced, taking advantage of multi-core and multi-processor machines. New features are also added, including decoding, MERT, and the framework. Benchmarking testset is also introduced.
In conclusion, Phramer is designed to deliver a better statistical phrase-based MT decoder experience than Pharaoh, with its open-source availability, cross-platform functionality, speed, extensibility, and more features.
As a software reviewer, you may ask why you should prefer Phramer instead of Pharaoh. Firstly, Phramer is an open-source program with minimal limitations compared to Pharaoh, which is currently available under much restrictive licensing. Secondly, Phramer provides the source code availability not present in Pharaoh. Thirdly, Phramer is cross-platform, written in Java, and tested in Linux and Windows 2000/XP. Fourthly, Phramer is faster; it is currently 1.5 - 3 times faster than Pharaoh 1.2.3 (not including data files loading time), and its performance is continually improved upon.
Phramer also features a modular architecture with hooks provided for enhanced extensibility. It is evident that Phramer has more features than Pharaoh 1.2.3 (not including undocumented ones). Additional parameters can also be seen in the list of features.
For you to use Phramer, you will need Java 5/1.5 or higher. A newer version of Phramer is now available, fixing some bugs and doubling the speed of the previous version under regular conditions. Multi-threading for both decoding and MER training is introduced, taking advantage of multi-core and multi-processor machines. New features are also added, including decoding, MERT, and the framework. Benchmarking testset is also introduced.
In conclusion, Phramer is designed to deliver a better statistical phrase-based MT decoder experience than Pharaoh, with its open-source availability, cross-platform functionality, speed, extensibility, and more features.
What's New
Version 1.0.5 Pre: N/A