Akerblad Sentence Aligner

A (very direct) Java port of LDC's Champollion Sentence Aligner (http://champollion.sourceforge.net). Thanks go to the LDC for making Champollion available under the GNU license so that this project could be possible. As of the current version, the structure of the algorithms in Akerblad is that of Champollion. We do not redistribute the CTK toolkit (i.e. the demo, evaluation, and other language-specific data) since this would be redundant. Currently we support all of the basic functionality of Champollion, plus we allow the processing of multiple document pairs per run. This avoids the high cost of reloading the dictionary and stopwords for each document.

All source code was written in Java. The code was ported from the perl language by Jonathan Clark (JonClark.Info)

Coming Soon

The next release of Akerblad will make it far more than just a port of Champollion. Planned features include:

• Plugin architecture to allow the user to configure almost every aspect of Akerblad
• Rule-based dictionaries (vocab is generated on-the-fly)
• Weighted dictionaries (e.g. Lex weights from GIZA++)
• Vocabulary-based tokens and caching of dictionary matches (speed optimizations)
• Parameters that can be optimized with machine learning
• Automatic dictionary learning

Downloads

All software is available from: http://sourceforge.net/projects/akerblad/.

All software is licensed under the OSI-approved GNU General Public License. Please contact us if you would like the software under another license.

Content of Distributions

• Source code for Akerblad
• Akerblad JAR file and supporting libraries

Why the funny name?

Who is Akerblad anyway?