The Apertium Project

Mentors: Student: Uliana Sentsova

<p>The project goal is to create a machine translation package for Sicilian-Spanish language pair on the base of Apertium’s machine translation system. This project is using the existing Apertium’s Spanish dictionary in order to build a package with two monolingual dictionaries for Spanish and Sicilian languages and a bilingual package for Spanish-Sicilian equivalents and translation rules.</p> <p>So, the project flow is the following:</p> <ul> <li>building a monolingual package for Sicilian,</li> <li>creating a bilingual Spanish-Sicilian equivalents dictionary,</li> <li>writing lexical decision rules and transfer rules for these languages,</li> <li>and collecting a parallel corpus for final evaluation. </li> </ul> <p>As a result, the final coverage of monolingual Sicilian package must be more than 90% of Sicilian Wikipedia corpus, and the evaluation on the parallel corpus must be error-free. First of all, this Sicilian dictionary has good prospects to become a base for new Apertium language pairs. Second, automatic translation based on the developed package could be useful for Wikipedia users: Apertium's machine translation for Sicilian could help to translate more articles for the growing Sicilian Wikipedia.</p>

Mentors: Student: Kira

<p>New features provide benefits both to Apertium users and Apertium team. Apertium website users will get the improved tool which provides a new dictionary lookup mode which is the second important feature after translation itself. The feedback feature is important to Apertium team. Apertium team will be able to know more about Apertium from users and the tool obtains more testing from people who don't have technical background. Both the feedback page and reliability visualization make the site more user-friendly thus it will grow to one of the coolest online translation tools.</p>

Mentors: Student: Jaipal

<p>My draft proposal for the Lint for Apertium project.</p>

Mentors: Student: tiefling-cat

<p>The idea of the task is to implement a mechanism of resolving transfer rule conflicts using previously obtained rule weights. The tool for obtaining the weights is to be implemented as well.</p> <p>Currently, transfer rules have to be devised with no possibility of conflicts, and that limits the system's capabilities, because ambiguity is an essential part of any natural language, and that fact should be reflected by the rules. Introducing of weighted rules would have a positive impact on both how the transfer rules represent the language pair and the quality of the resulting translation.</p> <p>The deliverables will be as follows:</p> <ol> <li>Standalone training script to calculate the weights given the corpus and the rules.</li> <li>The code in C++ integrated into Apertium to choose the rules given the weights.</li> </ol>

Mentors: Student: Maria Sheyanova

<p>Currently the pol-rus language pair is in the beginning state (in the incubator). There are very few words in the bilingual dictionary and no rules. The pol dictionary also does not have enough words. My goal is to fill the dictionaries, make as many rules as I'll be able to and bring it near to the release quality.</p>