Is there any resource that describes in detail a naive example-based machine translation algorithm?

Question

I'm looking to develop a machine translation tool for a constructed language. I think that the example-based approach is the most suitable because the said language is very regular and I can have a sufficient amount of parallel translations.

I already know the overall idea behind the example-based machine translation (EBMT) approach, but I can't find any resource that describes a naive EBMT algorithm (or model) that would allow me to easily implement it.

So, I'm looking for either:

• a detailed description,
• pseudocode or
• a sufficiently clear open-source project (maybe a GitHub one)

of a naive EBMT algorithm. So, I'm not looking for a software library that implements this, but I'm looking for a resource that explains/describes in detail a naive/simple EBMT algorithm, so that I am able to implement it.

Note that there are probably dozens of variations of EBMT algorithms. I'm only looking for the most naive/simple one.

I have already looked at the project Phrase-based Memory-based Machine Translator, but, unfortunately, it is not purely based on examples but also statistical, i.e. it needs an alignment file generated by, for example, Giza++ or Moses.

Answer 1

I have not found any simple implementation of a naive EBMT system, but I found some articles, papers and books that may be helpful (although I haven't read them, apart from the first and last one), so I will list them below.

The web article Example-based machine translation provides a decent high-level explanation of example-based machine translation.

The paper Example-Based Machine Translation: A New Paradigm (2002) by Chunyu Kit et al. also seems to provide a detailed description of the EBMT approach, so this paper should provide you with details you need to implement an EBMT system.

The paper A framework of a mechanical translation between Japanese and English by analogy principle (1984) by Makoto Nagao introduced the example-based machine translation approach, so it will be at least historically relevant.

Additionally, the paper Example-Based Machine Translation of the Basque Language and the book Recent Advances in Example-Based Machine Translation (2003), which is not apparently freely available online, could also be useful.

Finally, the article Machine Translation. From the Cold War to Deep Learning gives a nice high-level overview of the main machine translation approaches, so that you can understand the differences between EBMT and other approaches (especially, in case you are not able to distinguish between EBMT and other MT, e.g. those that use a parallel corpus, such as the supervised statistical machine translation approaches).