Ideas for Google Summer of Code

From Apertium
Revision as of 14:16, 22 March 2009 by Jimregan (talk | contribs) (→‎Further reading: of course, for Trigram tagging, we must cite Brants)
Jump to navigation Jump to search

This is the ideas page for Google Summer of Code, here you can find ideas on interesting projects that would make Apertium more useful for people and improve or expand our functionality. If you have an idea please add it below, if you think you could mentor someone in a particular area — or just have interests or ideas for that, add your name to "Interested parties" using ~~~

The page is intended as an overview of the kind of projects we have in mind. If one of them particularly piques your interest, please come and discuss with us on #apertium on irc.freenode.net, mail the mailing list, or draw attention to yourself in some other way.

Maybe take a look at some open bugs, or browse some of the pages in the development category ?

Difficulty = 1 (Very Hard) ... 4 (Entry level)

See the list sorted by:

List

Task Difficulty Description Rationale Requirements Interested
parties
Improve interoperability 3. Medium Either to modify Apertium to accept different formats, or to modify the other tools to accept the Apertium format, or alternatively write some kind of generic "glue" code that converts between them. There is a lot of great free software that could be used with the Apertium engine. We already have support for SFST (for morphological analysis) and Constraint Grammar (CG, for disambiguation) and use CG in several language pairs. It would be cool to have support for other tools, such as being able to use grammars from Freeling or link grammar to do preprocessing. Unfortunately these, along with many other tools have incompatible input/output formats. C, C++, XML Francis Tyers, Jimregan
Constraint Grammar
dependencies
2. Hard Develop a modification to the Apertium stream format to allow information about dependencies to be passed efficiently from CG to the transfer stage. There are many Constraint Grammars out there in the wild which as well as disambiguation do dependency parsing. We support the disambiguation stage, but not the dependency information. It would be nice to come up a way to get CG dependencies to feed into Apertium, or into Matxin, a sister project. C++, knowledge of dependency parsing Francis Tyers, Jimregan
n-Stage transfer 2. Hard Adapt the Apertium transfer engine code to work with more than one stage of "interchunk" movement. This will involve making some minor changes to the transfer language. It would be worth looking at Steven Abney's work on partial parsing with finite-state cascades. Apertium currently has between one and three stages of transfer, sometimes, with distant languages this is not enough, it would be nice to make it possible to have more than one stage of interchunk transfer, which would also make it possible to "join" chunks. This would improve the treatment of less related languages and allow for more complex verb movement. C++ Francis Tyers, Sortiz, Jimregan
Accent and diacritic
restoration
3. Medium Create an optional module to restore diacritics and accents on input text, and integrate it into the Apertium pipeline. Many languages use diacritics and accents in normal writing, and Apertium is designed to use these, however in some places, especially for example. instant messaging, irc, searching in the web etc. these are often not used or untyped. This causes problems as for the engine, traduccion is not the same as traducción. C, C++, XML, familiarity with linguistic issues Francis Tyers
Porting 3. Medium Port Apertium to Windows complete with nice installers and all that jazz. Apertium currently compiles on Windows (see Apertium on Windows), but we'd like to see it compile with a free tool-chain (MingGW, etc.) While we all might use GNU/Linux, there are a lot of people out there who don't, some of them use Microsoft's Windows. It would be nice for these people to be able use Apertium too. C++, autotools, experience in programming on Windows. Francis Tyers, Jimregan, Xavi Ivars
Lexical selection 2. Hard Write a prototype lexical selection module for Apertium using a combination of rule-based and statistical approaches, or maybe only an statistical approach. Lexical selection is the task of choosing a sense (meaning) for a word out of a number of possible senses (related to word sense disambiguation), when languages are close, they often share semantic ambiguity, when they are further apart they do not, so for example Spanish "estación" can be either "station", "season" or "resort" in English. Lexical selection is the task of choosing the right one. See also: Category:Lexical selection C++, XML, good knowledge of statistics. Jimregan, Fsanchez
Interfaces 4. Entry level Create plugins or extensions for popular free software applications to include support for translation using Apertium. We'd expect at least Firefox and Evolution (or Thunderbird), but to start with something more easy we have half-finished plugins for Pidgin and XChat that could use some love. The more the better! Apertium currently runs as a stand alone translator. It would be great if it was integrated in other free software applications. For example so instead of copy/pasting text out of your email, you could just click a button and have it translated in place. Depends on the application chosen, but probably C, C++, Python or Perl. Francis Tyers, Jimregan
Linguistically-driven filtering of the bilingual phrases used to infer shallow-transfer rules 3. Medium Re-working apertium-transfer-training-tools to filter the set of bilingual phrases automatically obtained from a word-aligned sentence pair by using linguistic criteria. Apertium-transfer-training-tools is a cool piece of software that generates shallow-transfer rules from aligned parallel corpora. It could greatly speed up the creation of new language pairs by generating rules that would otherwise have to be written by human linguists C++, general knowledge of GIZA++, Perl considered a plus. Fsanchez, Jimregan
Use of context-dependent lexicalized categories in the inference of shallow-transfer rules 2. Hard Re-working apertium-transfer-training-tools to use context-dependent lexicalized categories in the inference of shallow-transfer rules. Apertium-transfer-training-tools generates shallow-transfer rules from aligned parallel corpora. It uses an small set of lexicalized categories, categories that are usually involved in lexical changes, such as prepositions, pronouns or auxiliary verbs. Lexicalized categories differentiate from the rest of categories because their lemmas are taken into account in the generation of rules. C++, general knowledge of GIZA++, XML. Fsanchez, Jimregan
Automated lexical
extraction
2. Hard Writing a C++ wrapper around Markus Forsberg's Extract tool (version 2.0) as a library to allow it to be used with Apertium paradigms and TSX files as input into its paradigms and constraints. One of the things that takes a lot of time when creating a new language pair is constructing the monodices. The extract tool can greatly reduce the time this takes by matching lemmas to paradigms based on distribution in a corpus. Haskell, C++, XML Francis Tyers
Generating grammar
checkers
3. Medium The data that come with Apertium (morphological analysers -- and constraint grammars) could be used to create grammar checkers. This task would be to work on an automatic converter for Apertium formats (dictionaries, disambiguation rules) to other popular grammar checker formats, or alternatively work on a standalone grammar checker. Maybe using something like languagetool or An Gramadóir Grammar checkers can be useful, for languages other than English moreso. They are one of the "must have" items of language technology. If we can re-use Apertium data for this purpose it will help both the project (by making creating new language pairs more rewarding) and the language communities (by making more useful software). XML, whatever programming language and natural language are used for testing. Francis Tyers, Jimregan, Marcin Miłkowski
Support for agglutinative
languages
2. Hard Propose a new dictionary format that is suited to languages with agglutinative morphology and modify the morphological compiler/analyser. Our dictionary format isn't particularly suited to agglutinative languages, and those with complex morphologies. There are many of these types of languages in the world, so it would be good to support them better. See also: Agglutination C++, XML, knowledge of a language with these features (e.g. Finnish, Basque, Turkish, Estonian, Aymara, etc.) Sortiz
Complex multiwords 2. Hard Write a bidirectional module for specifying complex multiword units, for example dirección general and zračna luka. See Multiwords for more information. Although in the Romance languages it is not a big problem, as soon as you start to get to languages with cases (e.g. Serbo-Croatian, Slovenian, German, etc.) the problem comes that you can't define a multiword of adj nom because the adjective has a lot of inflection. C, C++, XML Francis Tyers, Jimregan
Adopt a
language pair
4. Entry level Take on an orphaned language pair, and bring it up to release quality results. What this quality will be will depend on the language pair adopted, and will need to be discussed with the prospective mentor. This will involve writing linguistic data (including morphological rules and transfer rules — which are specified in a declarative language) Apertium has a few pairs of languages (e.g. sv-da, sh-mk, en-af, ga-gd, etc...) that are orphaned, they don't have active maintainers. A lot of these pairs have a lot of work already put in, just need another few months to get them to release quality. XML, a scripting language (Python, Perl), good knowledge of the language pair adopted. Francis Tyers, Jimregan
Word compounder
and de-compounder
4. Entry level Write a de-compounder and compounder for Apertium. Many languages in the world have compound words, in Europe e.g. German, Dutch, Danish, etc. These are often very low frequency or completely novel, and as such do not exist in our dictionaries. If we had some software to split these into their constituent parts we might be able to translate them, and improve accuracy on our pairs with these languages. See also bug #13, and the page Compounds C, C++, XML Francis Tyers
Post-edition
tool
3. Medium Make a post-edition tool to speed up revision of Apertium translations. It would likely include at least support for spelling and grammar checking, along with defining user-specified rules for fixing translations (search and replace, etc.). This tool can reuse an existing grammar checker such as LanguageTool. After translating with Apertium revision work has to be done to consider a translation as an "adequate" translation. An intelligent post-edition environment will help doing this task. In this environment some typical mistakes in the translation process that can be automatically detected (for example unknown words and homographs) could be highlighted to be taken in consideration while doing post-edition. Some typical mistakes could also be defined to advise the post-editor to check them. XML, PHP, Python, C, C++, whichever programming language. Sortiz, Jimregan, Marcin Miłkowski
Lexical insertion
tool
3. Medium Improving the current web based dictionary application to insert new word pairs into the Apertium dictionaries. This would involve both improving the functionality, and efficiency of the software. Currently people have to edit XML in order to add words to the dictionaries. We have a web application, written in Python that does a lot of this work, but it still lacks some functionality, for example multiwords, and complete support for the new dictionary format. What is more it is quite slow and memory intensive. Python Francis Tyers
Generating spell
checkers
4. Entry level The data that come with Apertium (morphological analysers) could be used to create spell checkers. This task would be to work on an automatic converter for Apertium formats to other popular spell checker formats. Maybe using something ispell, myspell, hunspell, youspell, etc. Spell checkers can be useful especially before translating. They are one of the basic items of language technology. If we can re-use Apertium data for this purpose it will help both the project (by making creating new language pairs more rewarding) and the language communities (by making more useful software). This will be particulary useful for minority languages having an Apertium translator but not having a free spell checker, and also to use spell checking tools as a controlled language tool. XML, whatever programming language and natural language are used for testing. Sortiz
Detect 'hidden' unknown words 3. Medium The part-of-speech tagger of Apertium can be modified to work out the likelihood of each 'tag' in a certain context, this can be used to detect missing entries in the dictionary. Apertium dictionaries may have incomplete entries, that is, surface forms (lexical units as they appear in running texts) for which the dictionary does not provided all the possible lexical forms (consisting of lemma, part-of-speech and morphological inflection information). As those surface form for which there is at least one lexical form cannot be considered unknown words, it is difficult to know whether all lexical forms for a given surface form have been included in the dictionaries or not. This feature will detect 'possible' missing lexical forms for those surface forms in the dictionaries. C++ if you plan to modify the part-of-speech tagger; whatever if rewriting it from scratch. Sortiz, Fsanchez
Format filters 4. Entry level Making apertium capable of dealing with more different formats, for the minimum: files marked up with LaTeX and Wiki. Apertium can currently deal with texts in plain-text, RTF, HTML and ODT formats by means of a format definition file. It should be easy to use the same language to define filters for other formats. Apertium format definition language and/or scripting languages. Mlforcada Francis Tyers, Jimregan
Multi-engine translation synthesiser 3. Medium Write synthesiser to make a "better" translation out of several translations. There are other open-source machine translation systems in existence (for example Moses and OpenLogos), the point of this project would be to write a "synthesiser" which can, given several translations, produce a better translation. The program will probably take each of the output sentences from each system, decompose them into chunks or phrases and then score them against a language model to come to the final synthesised translation. Care should be taken to not overly bias fluent translations over adequate translations. C++ or Python (for prototyping) Francis Tyers
Trigram tagger 3. Medium Modify the tagger to use trigrams instead of bigrams We use bigrams -- for speed -- in the tagger, but now computers have improved and we have 3-stage transfer which will dominate CPU usage anyway. See also these questions C++. Jacob Nordfalk, Jimregan
Language independent
target-language tagger training
2. Hard Modify apertium-tagger-training-tools so that it does not need the segmentation provided by the transfer rules in order to score translations, and to prune the k-best list of possible translations at run-time for efficiency apertium-tagger-training-tools is a program for doing target language tagger training. This means that it tunes the parameters of an HMM model based on the quality of the translations through the whole system. At the moment this relies on the segmentation of the input by the transfer rules, and also runs all the possible translations (undisambiguated). It would be good to be able to run it without having this segmentation. C++ Fsanchez Francis Tyers Mlforcada
Daemon mode 3. Medium Write a program to allow Apertium to be run in "server" or "daemon" mode. When a lot of documents are translated, we launch the translator (the whole pipeline) multiple times, this is results in it being slower than ideal because of the time taken in process creation, loading transducers, grammars etc. and OS overheads. If it were to be run as a server or daemon, then these operations would only need to be performed once. C++ Xerakko, Jimregan
Geriaoueg
vocabulary assistant
4. Entry level Extend Geriaoueg so that it works more reliably with broken HTML and with any given language pair. Geriaoueg is a program that provides "popup" vocabulary assistance, something like BBC Vocab or Lingro. Currently it only works with Breton--French, Welsh--English and Spanish--Breton. This task would be to develop it to work with any language in our SVN and fix problems with processing and displaying non-standard HTML. PHP, C++, XML Francis Tyers Donnek
Conversion of
Anubadok
3. Medium Convert Anubadok to use the Apertium engine for transfer and generation. Anubadok is a GPL'd machine translation system for English to Bengali. Part of the work in converting the bilingual dictionary has already been done (see details in the incubator), but the Bengali generation side needs to be worked on and the tagging needs to be standardised between Apertium and Anubadok. The author (G. M. Hossain) has been contacted and agrees with the idea. Python, shell scripting, knowledge of Bengali definite plus Francis Tyers

Notes

Further reading

Accent and diacritic restoration
Lexical selection
Automated lexical extraction
Support for agglutinative languages
Transfer rule learning
  • Sánchez-Martínez, F. and Forcada, M.L. (2007) "Automatic induction of shallow-transfer rules for open-source machine translation", in Proceedings of TMI 2007, pp.181-190 (paper, poster)
Compounding and de-compounding
Multi-engine machine translation
n-Stage transfer
Trigram Tagging

Natural Language Processing, ANLP. ACL