Difference between revisions of "Ideas for Google Summer of Code"

From Apertium
Jump to navigation Jump to search
 
(176 intermediate revisions by 16 users not shown)
Line 1: Line 1:
 
{{TOCD}}
 
{{TOCD}}
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, add your name to "Interested mentors" using <nowiki>~~~</nowiki>
+
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.
   
  +
'''Current Apertium contributors''': If you have an idea please add it below, if you think you could mentor someone in a particular area, add your name to "Interested mentors" using <code><nowiki>~~~</nowiki></code>.
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 <code>#apertium</code> on <code>irc.freenode.net</code>, mail the [[Contact|mailing list]], or draw attention to yourself in some other way.
 
   
  +
'''Prospective GSoC contributors''': 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 <code>#apertium</code> on <code>irc.oftc.net</code> ([[IRC|more on IRC]]), mail the [[Contact|mailing list]], or draw attention to yourself in some other way.
Note that, if you have an idea that isn't mentioned here, we would be very interested to hear about it.
 
  +
  +
Note that if you have an idea that isn't mentioned here, we would be very interested to hear about it.
   
 
Here are some more things you could look at:
 
Here are some more things you could look at:
Line 13: Line 15:
 
* Resources that could be converted or expanded in the [[incubator]]. Consider doing or improving a language pair (see [[incubator]], [[nursery]] and [[staging]] for pairs that need work)
 
* Resources that could be converted or expanded in the [[incubator]]. Consider doing or improving a language pair (see [[incubator]], [[nursery]] and [[staging]] for pairs that need work)
 
* Unhammer's [[User:Unhammer/wishlist|wishlist]]
 
* Unhammer's [[User:Unhammer/wishlist|wishlist]]
  +
<!--* The open issues [https://github.com/search?q=org%3Aapertium&state=open&type=Issues on Github] - especially the [https://github.com/search?q=org%3Aapertium+label%3A%22good+first+issue%22&state=open&type=Issues Good First Issues]. -->
* The [http://sourceforge.net/p/apertium/tickets/search/?q=!status%3Awont-fix+%26%26+!status%3Aclosed open tickets] page on SourceForge
 
   
 
__TOC__
 
__TOC__
   
  +
If you're a prospective GSoC contributor trying to propose a topic, the recommended way is to request a wiki account and then go to <pre>http://wiki.apertium.org/wiki/User:[[your username]]/GSoC2023Proposal</pre> and click the "create" button near the top of the page. It's also nice to include <code><nowiki>[[</nowiki>[[:Category:GSoC_2023_student_proposals|Category:GSoC_2023_student_proposals]]<nowiki>]]</nowiki></code> to help organize submitted proposals.
== List ==
 
   
  +
== Language Data ==
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
=== <u>Anaphora resolution for machine translation</u> ===
 
* '''Difficulty''':<br><span style="background-color: #efcdcd">1. Hard</span>
 
* '''Required skills''':<br>C++, XML, Python
 
* '''Description''':<br>Write a program to resolve anaphora and include it in the Apertium translation pipeline.
 
* '''Rationale''':<br>Apertium has a problem with long distance dependencies in terms of agreement and co-reference. For example, deciding which determiner to use when translating from Spanish "su" to English "his, her, its". The objective of this task is to make a system to resolve anaphora and integrate it into a translation pipeline.
 
* '''Mentors''':<br>[[User:Francis Tyers|Francis Tyers]]
 
* '''[[/Anaphora resolution|read more...]]'''
 
</div>
 
   
  +
Can you read or write a language other than English (and we do mean any language)? If so, you can help with one of these and we can help you figure out the technical parts.
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
   
  +
{{IdeaSummary
=== <u>Bring a released language pair up to state-of-the-art quality</u> ===
 
  +
| name = Develop a morphological analyser
* '''Difficulty''':<br><span style="background-color: #cdefcd">2. Medium</span>
 
  +
| difficulty = easy
* '''Required skills''':<br>XML, a scripting language (Python, Perl), good knowledge of the language pair adopted.
 
  +
| size = either
* '''Description''':<br>Take a released language pair, and drastically improve the performance both in terms of coverage, and in terms of translation quality. This will involve working with dictionaries, transfer rules, scripting, corpora. The objective is to make an Apertium language pair state-of-the-art, or close to state-of-the-art in terms of translation quality. This will involve improving coverage to 95-98% on a range of corpora and decreasing [[word error rate]] by 30-50%. For example if the current word error rate is 30%, then it should be reduced to 15-20%.
 
  +
| skills = XML or HFST or lexd
* '''Rationale''':<br>Apertium has quite a broad coverage of language pairs, but few of these pairs offer state-of-the-art translation quality. We think broad is important, but deep coverage is important too.
 
  +
| description = Write a morphological analyser and generator for a language that does not yet have one
* '''Mentors''':<br>[[User:Francis Tyers|Francis Tyers]], [[User:Mlforcada|Mikel Forcada]], [[User:Xavivars|Xavi Ivars]]
 
  +
| rationale = A key part of an Apertium machine translation system is a morphological analyser and generator. The objective of this task is to create an analyser for a language that does not yet have one.
* '''[[/Make a language pair state-of-the-art|read more...]]'''
 
  +
| mentors = [[User:Francis Tyers|Francis Tyers]], [[User:Firespeaker|Jonathan Washington]], [[User: Sevilay Bayatlı|Sevilay Bayatlı]], Hossep, nlhowell, [[User:Popcorndude]]
</div>
 
  +
| more = /Morphological analyser
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = apertium-separable language-pair integration
=== <u>Robust tokenisation in lttoolbox</u> ===
 
  +
| difficulty = Medium
* '''Difficulty''':<br><span style="background-color: #cdefcd">2. Medium</span>
 
  +
| size = small
* '''Required skills''':<br>C++, XML, Python
 
  +
| skills = XML, a scripting language (Python, Perl), some knowledge of linguistics and/or at least one relevant natural language
* '''Description''':<br>Improve the longest-match left-to-right tokenisation strategy in [[lttoolbox]] to be fully Unicode compliant.
 
  +
| description = Choose a language you can identify as having a good number of "multiwords" in the lexicon. Modify all language pairs in Apertium to use the [[Apertium-separable]] module to process the multiwords, and clean up the dictionaries accordingly.
* '''Rationale''':<br>One of the most frustrating things about working with Apertium on texts "in the wild" is the way that the tokenisation works. If a letter is not specified in the alphabet, it is dealt with as whitespace, so e.g. you get unknown words split in two so you can end up with stuff like ^G$ö^k$ı^rmak$ which is terrible for further processing.
 
  +
| rationale = Apertium-separable is a newish module to process lexical items with discontinguous dependencies, an area where Apertium has traditionally fallen short. Despite all the module has to offer, many translation pairs still don't use it.
* '''Mentors''':<br>[[User:Francis Tyers|Francis Tyers]]
 
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Popcorndude]]
* '''[[/Robust tokenisation|read more...]]'''
 
  +
| more = /Apertium separable
</div>
 
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Bring an unreleased translation pair to releasable quality
  +
| difficulty = Medium
  +
| size = large
  +
| skills = shell scripting
  +
| description = Take an unstable language pair and improve its quality, focusing on testvoc
  +
| rationale = Many Apertium language pairs have large dictionaries and have otherwise seen much development, but are not of releasable quality. The point of this project would be bring one translation pair to releasable quality. This would entail obtaining good naïve coverage and a clean [[testvoc]].
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Seviay Bayatlı|Sevilay Bayatlı]], [[User:Unhammer]], [[User:hectoralos|Hèctor Alòs i Font]]
  +
| more = /Make a language pair state-of-the-art
  +
}}
   
  +
{{IdeaSummary
=== <u>Adopt an unreleased language pair</u> ===
 
  +
| name = Develop a prototype MT system for a strategic language pair
* '''Difficulty''':<br><span style="background-color: #cdcdef">3. Entry level</span>
 
  +
| difficulty = Medium
* '''Required skills''':<br>XML, a scripting language (Python, Perl), good knowledge of the language pair adopted.
 
  +
| size = large
* '''Description''':<br>Take on an orphaned unreleased 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 &mdash; which are specified in a declarative language &mdash; and possibly [[Constraint Grammar]] rules if that is relevant)
 
  +
| skills = XML, some knowledge of linguistics and of one relevant natural language
* '''Rationale''':<br>Apertium has a few pairs of languages (e.g. mt-he, ga-gd, ur-hi, pl-cs, sh-ru, 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. See also [[Incubator]]
 
  +
| description = Create a translation pair based on two existing language modules, focusing on the dictionary and structural transfer
* '''Mentors''':<br>[[User:Francis Tyers|Francis Tyers]], [[User:Jimregan|Jimregan]], [[User:Kevin Scannell|Kevin Scannell]], [[User:Trondtr|Trondtr]], [[User:Unhammer|Unhammer]], [[User:Darthxaher|Darthxaher]], [[User:Firespeaker|Firespeaker]], [[User:Hectoralos|Hectoralos]], [[User:Krvoje|Hrvoje Peradin]], [[User:Jacob Nordfalk|Jacob Nordfalk]], [[User:Mlforcada|Mikel Forcada]], [[User:Vin-ivar|Vinit Ravishankar]], [[User:Aida|Aida Sundetova]], [[User:Xavivars|Xavi Ivars]]
 
  +
| rationale = Choose a strategic set of languages to develop an MT system for, such that you know the target language well and morphological transducers for each language are part of Apertium. Develop an Apertium MT system by focusing on writing a bilingual dictionary and structural transfer rules. Expanding the transducers and disambiguation, and writing lexical selection rules and multiword sequences may also be part of the work. The pair may be an existing prototype, but if it's a heavily developed but unreleased pair, consider applying for "Bring an unreleased translation pair to releasable quality" instead.
* '''[[/Adopt a language pair|read more...]]'''
 
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Sevilay Bayatlı| Sevilay Bayatlı]], [[User:Unhammer]], [[User:hectoralos|Hèctor Alòs i Font]]
</div>
 
  +
| more = /Adopt a language pair
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Add a new variety to an existing language
  +
| difficulty = easy
  +
| size = either
  +
| skills = XML, some knowledge of linguistics and of one relevant natural language
  +
| description = Add a language variety to one or more released pairs, focusing on the dictionary and lexical selection
  +
| rationale = Take a released language, and define a new language variety for it: e.g. Quebec French or Provençal Occitan. Then add the new variety to one or more released language pairs, without diminishing the quality of the pre-existing variety(ies). The objective is to facilitate the generation of varieties for languages with a weak standardisation and/or pluricentric languages.
  +
| mentors = [[User:hectoralos|Hèctor Alòs i Font]], [[User:Firespeaker|Jonathan Washington]],[[User:piraye|Sevilaybayatlı]]
  +
| more = /Add a new variety to an existing language
  +
}}
   
  +
{{IdeaSummary
=== <u>Extend lttoolbox to have the power of HFST</u> ===
 
  +
| name = Leverage and integrate language preferences into language pairs
* '''Difficulty''':<br><span style="background-color: #efcdcd">1. Hard</span>
 
  +
| difficulty = easy
* '''Required skills''':<br>C++, XSLT, XML
 
  +
| size = medium
* '''Description''':<br>Extend lttoolbox (perhaps writing a preprocessor for it) so that it can be used to do the morphological transformations currently done with HFST. And yes, of course, writing something that translates the current HFST format to the new lttolbox format. Proof of concept: Come up with a new format that can express all of the features found in the Kazakh transducer; implement this format in Apertium; Implement the Kazakh transducer in this format and integrate it in the English--Kazakh pair.
 
  +
| skills = XML, some knowledge of linguistics and of one relevant natural language
* '''Rationale''':<br>Some language pairs in Apertium use HFST where most language pairs use Apertium's own lttoolbox. This is due to the fact that writing morphologies for languages that have features such as the vowel harmony found in Turkic languages is very hard with the current format supported by lttoolbox. The mixture of HFST and lttoolbox makes it harder for people to develop some language pairs.
 
  +
| description = Update language pairs with lexical and orthographical variations to leverage the new [[Dialectal_or_standard_variation|preferences]] functionality
* '''Mentors''':<br>[[User:mlforcada|Mikel Forcada]], [[User:TommiPirinen|Tommi A Pirinen]], [[User:Unhammer]], [[User:Mlforcada|Mikel Forcada]], mentors wanted
 
  +
| rationale = Currently, preferences are implemented via language variant, which relies on multiple dictionaries, increasing compilation time exponentially every time a new preference gets introduced.
* '''[[/Extend lttoolbox to have the power of HFST|read more]]'''
 
  +
| mentors = [[User:Xavivars|Xavi Ivars]] [[User:Unhammer]]
</div>
 
  +
| more = /Use preferences in pair
  +
}}
   
  +
{{IdeaSummary
<!--
 
  +
| name = Add Capitalization Handling Module to a Language Pair
  +
| difficulty = easy
  +
| size = small
  +
| skills = XML, knowledge of some relevant natural language
  +
| description = Update a language pair to make use make use of the new [[Capitalization_restoration|Capitalization handling module]]
  +
| rationale = Correcting capitalization via transfer rules is tedious and error prone, but putting them in a separate set of rules should allow them to be handled in a more concise and maintainable way. Additionally, it is possible that capitalization rule could be moved to monolingual modules, thus reducing development effort on translators.
  +
| mentors = [[User:Popcorndude]]
  +
| more = /Capitalization
  +
}}
   
  +
== Data Extraction ==
DANGER TERROR HORROR !!!!!!
 
   
  +
A lot of the language data we need to make our analyzers and translators work already exists in other forms and we just need to figure out how to convert it. If you know of another source of data that isn't listed, we'd love to hear about it.
The task above has subsumed these two
 
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = dictionary induction from wikis
=== <u>Flag diacritics in lttoolbox</u> ===
 
  +
| difficulty = Medium
* '''Difficulty''':<br><span style="background-color: #efcdcd">1. Hard</span>
 
  +
| size = either
* '''Required skills''':<br>C++ or Java, XML, Knowledge of FSTs
 
  +
| skills = MySQL, mediawiki syntax, perl, maybe C++ or Java; Java, Scala, RDF, and DBpedia to use DBpedia extraction
* '''Description''':<br>Adapt [[lttoolbox]] to elegantly use flag diacritics. Flag diacritics are a way of avoiding transducer size blow-up by discarding impossible paths at runtime as opposed to compile time. Some work has already been done, see [[Flag diacritics]].
 
  +
| description = Extract dictionaries from linguistic wikis
* '''Rationale''':<br>This will involve designing some changes to our XML dictionary format (see [[lttoolbox]], and implementing the associated changes in the FST compiling processing code. The reason behind this is that many languages have prefix inflection, and we cannot currently deal with this without either making paradigms useless, or overanalysing (e.g. returning analyses where none exist). Flag diacritics (or constraints) would allow us to restrict overanalysis without blowing up the size of our dictionaries.
 
  +
| rationale = Wiki dictionaries and encyclopedias (e.g. omegawiki, wiktionary, wikipedia, dbpedia) contain information (e.g. bilingual equivalences, morphological features, conjugations) that could be exploited to speed up the development of dictionaries for Apertium. This task aims at automatically building dictionaries by extracting different pieces of information from wiki structures such as interlingual links, infoboxes and/or from dbpedia RDF datasets.
* '''Mentors''':<br>[[User:Francis Tyers|Francis Tyers]] (C++), [[User:Jacob Nordfalk|Jacob Nordfalk]] (Java)
 
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Popcorndude]]
* '''[[/Flag diacritics in lttoolbox|read more...]]'''
 
  +
| more = /Dictionary induction from wikis
</div>
 
  +
}}
   
  +
{{IdeaSummary
  +
| name = Dictionary induction from parallel corpora / Revive ReTraTos
  +
| difficulty = Medium
  +
| size = medium
  +
| skills = C++, perl, python, xml, scripting, machine learning
  +
| description = Extract dictionaries from parallel corpora
  +
| rationale = Given a pair of monolingual modules and a parallel corpus, we should be able to run a program to align tagged sentences and give us the best entries that are missing from bidix. [[ReTraTos]] (from 2008) did this back in 2008, but it's from 2008. We want a program which builds and runs in 2022, and does all the steps for the user.
  +
| mentors = [[User:Unhammer]], [[User:Popcorndude]]
  +
| more = /Dictionary induction from parallel corpora
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Extract morphological data from FLEx
=== <u>Weights in lttoolbox</u> ===
 
  +
| difficulty = hard
* '''Difficulty''':<br><span style="background-color: #cdefcd">2. Medium</span>
 
  +
| size = large
* '''Required skills''':<br>C++, XML, FSTs
 
  +
| skills = python, XML parsing
* '''Description''':<br>[[lttoolbox]] is a set of tools for building finite-state transducers. As part of Apertium's long-term strategy we would like to include probabilistic information into more stages of the pipeline to allow generic tools to be optimised for machine translation. This task involves adding the possibility of weighting lexemes and analyses in our finite-state transducer toolbox.
 
  +
| description = Write a program to extract data from [https://software.sil.org/fieldworks/ SIL FieldWorks] and convert as much as possible to monodix (and maybe bidix).
* '''Rationale''':<br>Weighting information for lexical forms will be useful for morphological disambiguation, and for work on [[spellchecking]].
 
  +
| rationale = There's a lot of potentially useful data in FieldWorks files that might be enough to build a whole monodix for some languages but it's currently really hard to use
* '''Mentors''':<br>[[User:Francis Tyers|Francis Tyers]], [[User:TommiPirinen|Tommi Pirinen]]
 
  +
| mentors = [[User:Popcorndude|Popcorndude]], [[User:TommiPirinen|Flammie]]
* '''[[/Add weights to lttoolbox|read more...]]'''
 
  +
| more = /FieldWorks_data_extraction
</div>
 
  +
}}
-->
 
   
  +
== Tooling ==
   
  +
These are projects for people who would be comfortable digging through our C++ codebases (you will be doing a lot of that).
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
=== <u>Robust recursive transfer</u> ===
 
* '''Difficulty''':<br><span style="background-color: #efcdcd">1. Hard</span>
 
* '''Required skills''':<br>Python, XML, linguistics
 
* '''Description''':<br>The purpose of this task would be to create a module to replace the apertium-transfer module(s) which will parse and allow transfer operations on an input.
 
* '''Rationale''':<br>Currently we have a problem with very distantly related languages that have long-distance constituent reordering, because we can only do finite-state chunking.
 
* '''Mentors''':<br>[[User:Francis Tyers|Francis Tyers]], [[User:Sortiz|Sortiz]], [[User:Mlforcada|Mikel Forcada]], [[User:Japerez|Juan Antonio Pérez]]
 
* '''[[/Robust recursive transfer|read more...]]'''
 
</div>
 
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Python API for Apertium
=== <u>Extend weighted transfer rules</u> ===
 
  +
| difficulty = medium
* '''Difficulty''':<br><span style="background-color: #efcdcd">1. Hard</span>
 
  +
| size = medium
* '''Required skills''':<br>Python, C++, linguistics
 
  +
| skills = C++, Python
* '''Description''':<br>The purpose of this task is to extend weighted transfer rules to all transfer files and to allow conflicting rule patterns to be handled by combining (lexicalised) weights.
 
  +
| description = Update the Python API for Apertium to expose all Apertium modes and test with all major OSes
* '''Rationale''':<br>Currently our transfer rules are applied longest-match left-to-right ([[LRLM]]). When two rule patterns conflict the first one is chosen. We have a prototype for selecting based on lexicalised weights, but it only applies to the first stage of transfer.
 
  +
| rationale = The current Python API misses out on a lot of functionality, like phonemicisation, segmentation, and transliteration, and doesn't work for some OSes <s>like Debian</s>.
* '''Mentors''':<br>[[User:Francis Tyers|Francis Tyers]], [[User:TommiPirinen|Tommi Pirinen]]
 
  +
| mentors = [[User:Francis Tyers|Francis Tyers]]
* '''[[/Weighted transfer rules|read more...]]'''
 
  +
| more = /Python API
</div>
 
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Robust tokenisation in lttoolbox
  +
| difficulty = Medium
  +
| size = large
  +
| skills = C++, XML, Python
  +
| description = Improve the longest-match left-to-right tokenisation strategy in [[lttoolbox]] to handle spaceless orthographies.
  +
| rationale = One of the most frustrating things about working with Apertium on texts "in the wild" is the way that the tokenisation works. If a letter is not specified in the alphabet, it is dealt with as whitespace, so e.g. you get unknown words split in two so you can end up with stuff like ^G$ö^k$ı^rmak$ which is terrible for further processing. Additionally, the system is nearly impossible to use for languages that don't use spaces, such as Japanese.
  +
| mentors = [[User:Francis Tyers|Francis Tyers]], [[User:TommiPirinen|Flammie]]
  +
| more = /Robust tokenisation
  +
}}
   
  +
{{IdeaSummary
=== <u>Improvements to the Apertium website</u> ===
 
  +
| name = rule visualization tools
* '''Difficulty''':<br><span style="background-color: #cdcdef">3. Entry level</span>
 
  +
| difficulty = Medium
* '''Required skills''':<br>Python, HTML, JS
 
  +
| size = either
* '''Description''':<br>Our web site is pretty cool already, but it's missing things like dictionary/synonym lookup, support for several variants of one language, reliability visualisation, (reliable) webpage translation, feedback, etc.
 
  +
| skills = python? javascript? XML
* '''Rationale''':<br>[https://apertium.org https://apertium.org] / [http://beta.apertium.org http://beta.apertium.org] is what most people know us by, it should show off more of the things we are capable of :-)
 
  +
| description = make tools to help visualize the effect of various rules
* '''Mentors''':<br>[[User:Firespeaker|Jonathan]], [[User:Sushain|Sushain]]
 
  +
| rationale = TODO see https://github.com/Jakespringer/dapertium for an example
* '''[[/Apertium website improvements|read more...]]'''
 
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Sevilay Bayatlı|Sevilay Bayatlı]], [[User:Popcorndude]]
</div>
 
  +
| more = /Visualization tools
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Extend Weighted transfer rules
  +
| difficulty = Medium
  +
| size = medium
  +
| skills = C++, python
  +
| description = The weighted transfer module is already applied to the chunker transfer rules. And the idea here is to extend that module to be applied to interchunk and postchunk transfer rules too.
  +
| rationale = As a resource see https://github.com/aboelhamd/Weighted-transfer-rules-module
  +
| mentors = [[User: Sevilay Bayatlı|Sevilay Bayatlı]]
  +
| more = /Make a module
  +
}}
   
  +
{{IdeaSummary
=== <u>User-friendly lexical selection training</u> ===
 
  +
| name = Automatic Error-Finder / Pseudo-Backpropagation
* '''Difficulty''':<br><span style="background-color: #cdefcd">2. Medium</span>
 
  +
| difficulty = Hard
* '''Required skills''':<br>Python, C++, shell scripting
 
  +
| size = large
* '''Description''':<br>Make it so that training/inference of lexical selection rules is a more user-friendly process
 
  +
| skills = python?
* '''Rationale''':<br>Our lexical selection module allows for inferring rules from corpora and word alignments, but the procedure is currently a bit messy, with various scripts involved that require lots of manual tweaking, and many third party tools to be installed. The goal of this task is to make the procedure as user-friendly as possible, so that ideally only a simple config file would be needed, and a driver script would take care of the rest.
 
  +
| description = Develop a tool to locate the approximate source of translation errors in the pipeline.
* '''Mentors''':<br>[[User:Unhammer|Unhammer]], [[User:Francis Tyers|Francis Tyers]], [[User:Mlforcada|Mikel Forcada]]
 
  +
| rationale = Being able to generate a list of probable error sources automatically makes it possible to prioritize issues by frequency, frees up developer time, and is a first step towards automated generation of better rules.
* '''[[/User-friendly lexical selection training|read more...]]'''
 
  +
| mentors = [[User:Popcorndude]]
</div>
 
  +
| more = /Backpropagation
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = More Robust Recursive Transfer
=== <u>Light alternative format for all XML files in an Apertium language pair</u> ===
 
  +
| difficulty = Hard
* '''Difficulty''':<br><span style="background-color: #efcdcd">1. Hard</span>
 
  +
| size = large
* '''Required skills''':<br>Python, C++, shell scripting, XSLT, flex
 
  +
| skills = C++
* '''Description''':<br>Make it possible to edit and develop language data using a format that is lighter than XML
 
  +
| description = Ensure [[Apertium-recursive#Further_Documentation|Recursive Transfer]] survives ambiguous or incomplete parse trees
* '''Rationale''':<br>In most Apertium language pairs, monolingual dictionaries, bilingual dictionaries, post-generation rule files and structural transfer rule files are all written in XML. While XML is easy to process due to explicit tagging of every element, it is tedious to deal with, particularly when it comes to structural transfer rules. Apertium's precursor, interNOSTRUM, had lighter text based formats. The task involves: (a) designing and documenting an interNOSTRUM-style format for all of the XML language data files in a language pair; (b) writing converters to XML and from XML that are fully roundtrip-compliant: (c) designing a way to synchronize changes when both the XML and the non-XML format are used simultaneously in a specific language pair.
 
  +
| rationale = Currently, one has to be very careful in writing recursive transfer rules to ensure they don't get too deep or ambiguous, and that they cover full sentences. See in particular issues [https://github.com/apertium/apertium-recursive/issues/97 97] and [https://github.com/apertium/apertium-recursive/issues/80 80]. We would like linguists to be able to fearlessly write recursive (rtx) rules based on what makes linguistic sense, and have rtx-proc/rtx-comp deal with the computational/performance side.
* '''Mentors''':<br>[[User:Mlforcada|Mikel Forcada]], [[User:Japerez|Juan Antonio Pérez]], pair.
 
  +
| mentors =
* '''[[/Plain-text_formats_for_Apertium_data|read more...]]'''
 
  +
| more = /More_robust_recursive_transfer
</div>
 
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = CG-based Transfer
=== <u>Eliminate dictionary trimming</u> ===
 
  +
| difficulty = Hard
* '''Difficulty''':<br><span style="background-color: #ffbdbd">0. Very Hard</span>
 
  +
| size = large
* '''Required skills''':<br>C++, Finite-State Transducers
 
  +
| skills = C++
* '''Description''':<br>Eliminate the need for trimming the monolingual dictionaries, in order to preserve and take advantage of maximal source language analysis.
 
  +
| description = Linguists already write dependency trees in [[Constraint Grammar]]. A following step could use these to reorder into target language trees.
* '''Rationale''':<br>[[Why we trim]] mentions several technical reasons for why trimming away monolingual information is currently needed. Unfortunately, this limitation means that a lot of useful contextual information is lost. It would be ideal if the source language could be fully analyzed independent of target language, with any untranslated part fed back into the source language generator.
 
  +
| mentors =
* '''Mentors''':<br> [[User:TommiPirinen|Tommi Pirinen a.k.a. Flammie]]
 
  +
| more =
* '''Work around everything in [[Why we trim]]'''
 
  +
}}
</div>
 
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Language Server Protocol
  +
| difficulty = Medium
  +
| size = medium
  +
| skills = any programming language
  +
| description = Build a [https://microsoft.github.io/language-server-protocol/|Language Server] for the various Apertium rule formats
  +
| rationale = We have some static analysis tools and syntax highlighters already and it would be great if we could combine and expand them to support more text editors.
  +
| mentors = [[User:Popcorndude]]
  +
| more = /Language Server Protocol
  +
}}
   
  +
{{IdeaSummary
=== <u>Bilingual dictionary enrichment via graph completion</u> ===
 
  +
| name = WASM Compilation
* '''Difficulty''':<br><span style="background-color: #ffbdbd">0. Very hard</span>
 
  +
| difficulty = hard
* '''Required skills''':<br>shell scripting, python, XSLT, XML
 
  +
| size = medium
* '''Description''':<br>Generate new entries for existing or new bilingual dictionaries using graphic representations of bilingual correspondences as found in all existing dictionaries (note that this idea defines a rather open-ended task to be discussed in detail with mentors).
 
  +
| skills = C++, Javascript
* '''Rationale''':<br>Apertium bilingual dictionaries establish correspondences between lexical forms in a number of language pairs. Connections among them may be used to infer new entries for existing or new language pairs using graphs. The graphs may be directly generated from Apertium bidixes and exploiting using [[Bilingual_dictionary_discovery|ideas that had already been proposed in Apertium]] or using existing [http://linguistic.linkeddata.es/apertium/ RDF representations] of parts of their content, which may benefit from the information coming from being linked to other resources.
 
  +
| description = Compile the pipeline modules to WASM and provide JS wrappers for them.
* '''Mentors''':<br>[[User:Mlforcada|Mikel Forcada]], [[User:Francis Tyers|Francis Tyers]], [[User:Jorge Gracia|Jorge Gracia]]
 
  +
| rationale = There are situations where it would be nice to be able to run the entire pipeline in the browser
* '''[[Bilingual_dictionary_discovery|read more...]]''' '''[[/Bilingual_dictionary_enrichment_via_graph_completion|read even more...]]'''
 
  +
| mentors = [[User:Tino Didriksen|Tino Didriksen]]
</div>
 
  +
| more = /WASM
  +
}}
   
  +
== Web ==
   
  +
If you know Python and JavaScript, here's some ideas for improving our [https://apertium.org website]. Some of these should be fairly short and it would be a good idea to talk to the mentors about doing a couple of them together.
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
   
  +
{{IdeaSummary
=== <u>Transfer rule induction from comparable parsed corpora</u> ===
 
  +
| name = Web API extensions
* '''Difficulty''':<br><span style="background-color: #ffbdbd">1. Hard</span>
 
  +
| difficulty = medium
* '''Required skills''':<br>shell scripting, python, XSLT, XML
 
  +
| size = small
* '''Description''':<br>A system to infer transfer rules from comparable corpora that have both been deeply parsed (with e.g. CG)
 
  +
| skills = Python
* '''Rationale''':<br>Many languages have good CG's and fairly large monolingual corpora, but little parallel material. Given a small bidix, fairly large monolingual corpora and good analysers/CG's, we should be able to parse both corpora, translate lemmas and look for similar sentences, turning the differences in their parses into transfer rules.
 
  +
| description = Update the web API for Apertium to expose all Apertium modes
* '''Mentors''':<br>[[User:Unhammer]]
 
  +
| rationale = The current Web API misses out on a lot of functionality, like phonemicisation, segmentation, transliteration, and paradigm generation.
* '''[[Transfer_induction_from_comparable_parsed_corpora|read more...]]'''
 
  +
| mentors = [[User:Francis Tyers|Francis Tyers]], [[User:Firespeaker|Jonathan Washington]], [[User:Xavivars|Xavi Ivars]]
</div>
 
  +
| more = /Apertium APY
<div>
 
  +
}}
   
  +
{{IdeaSummary
  +
| name = Website Improvements: Misc
  +
| difficulty = Medium
  +
| size = small
  +
| skills = html, css, js, python
  +
| description = Improve elements of Apertium's web infrastructure
  +
| rationale = Apertium's website infrastructure [[Apertium-html-tools]] and its supporting API [[APy|Apertium APy]] have numerous open issues. This project would entail choosing a subset of open issues and features that could realistically be completed in the summer. You're encouraged to speak with the Apertium community to see which features and issues are the most pressing.
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Xavivars|Xavi Ivars]]
  +
| more = /Website improvements
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Website Improvements: Dictionary Lookup
=== <u>UD and Apertium integration</u> ===
 
  +
| difficulty = Medium
* '''Difficulty''':<br><span style="background-color: #cdcdef">3. Entry level</span>
 
  +
| size = small
* '''Required skills''':<br>python, javascript, HTML, (C++)
 
  +
| skills = html, css, js, python
* '''Description''':<br>Create a range of tools for making Apertium compatible with Universal Dependencies
 
  +
| description = Finish implementing dictionary lookup mode in Apertium's web infrastructure
* '''Rationale''':<br>Universal dependencies is a fast growing project aimed at creating a unified annotation scheme for treebanks. This includes both part-of-speech and morphological features. Their annotated corpora could be extremely useful for Apertium for training models for translation. In addition, Apertium's rule-based morphological descriptions could be useful for software that relies on Universal dependencies.
 
  +
| rationale = Apertium's website infrastructure [[Apertium-html-tools]] and its supporting API [[APy|Apertium APy]] have numerous open issues, including half-completed features like dictionary lookup. This project would entail completing the dictionary lookup feature. Some additional features which would be good to work would include automatic reverse lookups (so that a user has a better understanding of the results), grammatical information (such as the gender of nouns or the conjugation paradigms of verbs), and information about MWEs.
* '''Mentors''':<br>[[User:Francis Tyers]] [[User:Firespeaker]]
 
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Xavivars|Xavi Ivars]], [[User:Popcorndude]]
* '''[[/UD and Apertium integration|read more...]]'''
 
  +
| more = https://github.com/apertium/apertium-html-tools/issues/105 the open issue on GitHub
</div>
 
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Website Improvements: Spell checking
  +
| difficulty = Medium
  +
| size = small
  +
| skills = html, js, css, python
  +
| description = Add a spell-checking interface to Apertium's web tools
  +
| rationale = [[Apertium-html-tools]] has seen some prototypes for spell-checking interfaces (all in stale PRs and branches on GitHub), but none have ended up being quite ready to integrate into the tools. This project would entail polishing up or recreating an interface, and making sure [[APy]] has a mode that allows access to Apertium voikospell modules. The end result should be a slick, easy-to-use interface for proofing text, with intuitive underlining of text deemed to be misspelled and intuitive presentation and selection of alternatives. [https://github.com/apertium/apertium-html-tools/issues/390 the open issue on GitHub]
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Xavivars|Xavi Ivars]]
  +
| more = /Spell checker web interface
  +
}}
   
  +
{{IdeaSummary
=== <u>Add weights to lttoolbox</u> ===
 
  +
| name = Website Improvements: Suggestions
* '''Difficulty''':<br><span style="background-color: #ffbdbd">1. Hard</span>
 
  +
| difficulty = Medium
* '''Required skills''':<br>c++
 
  +
| size = small
* '''Description''':<br>Add support for weighted transducers to lttoolbox
 
  +
| skills = html, css, js, python
* '''Rationale''':<br>This will either involve implementing it from scratch or adding OpenFST as a backend. We would like to be able to use it both in the bilingual dictionaries, and in the morphological analysers, to be able to order analyses/translations by their probability/weight instead of by the random topological order.
 
  +
| description = Finish implementing a suggestions interface for Apertium's web infrastructure
* '''Mentors''':<br>[[User:Francis Tyers]] [[User:Unhammer]]
 
  +
| rationale = Some work has been done to add a "suggestions" interface to Apertium's website infrastructure [[Apertium-html-tools]] and its supporting API [[APy|Apertium APy]], whereby users can suggest corrected translations. This project would entail finishing that feature. There are some related [https://github.com/apertium/apertium-html-tools/issues/55 issues] and [https://github.com/apertium/apertium-html-tools/pull/252 PRs] on GitHub.
* '''[[/Add weights to lttoolbox|read more...]]'''
 
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Xavivars|Xavi Ivars]]
</div>
 
  +
| more = /Website improvements
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
  +
| name = Website Improvements: Orthography conversion interface
  +
| difficulty = Medium
  +
| size = small
  +
| skills = html, js, css, python
  +
| description = Add an orthography conversion interface to Apertium's web tools
  +
| rationale = Several Apertium language modules (like Kazakh, Kyrgyz, Crimean Tatar, and Hñähñu) have orthography conversion modes in their mode definition files. This project would be to expose those modes through [[APy|Apertium APy]] and provide a simple interface in [[Apertium-html-tools]] to use them.
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Xavivars|Xavi Ivars]]
  +
| more = /Website improvements
  +
}}
   
  +
{{IdeaSummary
=== <u>Improving language pairs mining Mediawiki Content Translation postedits</u> ===
 
  +
| name = Add support for NMT to web API
* '''Difficulty''':<br><span style="background-color: #ffbdbd">1. Hard</span>
 
  +
| difficulty = Medium
* '''Required skills''':<br>Python, shell scripting, some statistics
 
  +
| size = medium
* '''Description''':<br>Implement a toolkit that allows mining existing machine translation postediting data in [Mediawiki Content Translation https://www.mediawiki.org/wiki/Content_translation] to generate (as automatically as possible, and as complete as possible) monodix and bidix entries to improve the performance of an Apertium language pair. Data is available from Wikimedia content translation through an [API https://www.mediawiki.org/wiki/Content_translation/Published_translations#API] or in the form of [Dumps https://dumps.wikimedia.org/other/contenttranslation/] available in JSON and TMX format. This project is rather experimental and involves some research in addition to coding.
 
  +
| skills = python, NMT
* '''Rationale''':<br>Apertium is used to generate new Wikipedia content: machine-translated content is postedited (and perhaps adapted) before publishing. Postediting information may contain information that can be used to help improve the lexical components of an Apertium language pair.
 
  +
| description = Add support for a popular NMT engine to Apertium's web API
* '''Mentors''':<br> [[User:Mlforcada|Mikel Forcada]], (more mentors to be added)
 
  +
| rationale = Currently Apertium's web API [[APy|Apertium APy]], supports only Apertium language modules. But the front end could just as easily interface with an API that supports trained NMT models. The point of the project is to add support for one popular NMT package (e.g., translateLocally/Bergamot, OpenNMT or JoeyNMT) to the APy.
* '''(more soon)'''<!---[[/Shallow-function labeller|read more...]]'''-->
 
  +
| mentors = [[User:Firespeaker|Jonathan Washington]], [[User:Xavivars|Xavi Ivars]]
</div>
 
  +
| more =
  +
}}
   
  +
== Integrations ==
   
  +
In addition to incorporating data from other projects, it would be nice if we could also make our data useful to them.
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
=== <u>Unsupervised weighting of automata</u> ===
 
* '''Difficulty''':<br><span style="background-color: #cdefcd">2. Medium</span>
 
* '''Required skills''':<br>Python, shell scripting, statistics, finite-state transducers
 
* '''Description''':<br>Implement a collection of methods for weighting finite-state transducers, the methods should include an implementation of a simple method of supervised training, and a number of methods for unsupervised training. The objective being to get the analysis ranking given a set of a analyses for a given surface form as close to the result given by supervised training as possible.
 
* '''Rationale''':<br>Apertium struggles with ambiguity, we have had many attempts to write better part of speech taggers. This would complement those attempts by providing a generic method to weight automata.
 
* '''Mentors''':<br> [[User:Francis Tyers|Francis Tyers]], [[User:TommiPirinen|Flammie]], [[User:Unhammer|Unhammer]]
 
* '''[[/Unsupervised weighting of automata|read more...]]'''
 
</div>
 
   
  +
{{IdeaSummary
  +
| name = OmniLingo and Apertium
  +
| difficulty = medium
  +
| size = either
  +
| skills = JS, Python
  +
| description = OmniLingo is a language learning system for practicing listening comprehension using Common Voice data. There is a lot of text processing involved (for example tokenisation) that could be aided by Apertium tools.
  +
| rationale =
  +
| mentors = [[User:Francis Tyers|Francis Tyers]]
  +
| more = /OmniLingo
  +
}}
   
  +
{{IdeaSummary
<div style="background-color: #f9f9f9; border: 1px solid black; padding: 1ex; margin-bottom: 2ex;">
 
=== <u>Improvements to UD Annotatrix</u> ===
+
| name = Support for Enhanced Dependencies in UD Annotatrix
  +
| difficulty = medium
* '''Difficulty''':<br><span style="background-color: #cdefcd">2. Medium</span>
 
  +
| size = medium
* '''Required skills''':<br>JavaScript, Jquery, HTML, Python
 
  +
| skills = NodeJS
* '''Description''':<br>UD Annotatrix is an interface by Apertium for annotating dependency trees in CoNLL-U format. The system is currently in beta, but is getting traction as more people start using it.
 
  +
| description = UD Annotatrix is an annotation interface for Universal Dependencies, but does not yet support all functionality.
* '''Rationale''':<br>Universal Dependencies is a very widely used standard for annotating data, the kind of annotated data that can be used to train part of speech taggers. There is still a lot of work that could be done to improve it.
 
  +
| rationale =
* '''Mentors''':<br> [[User:Francis Tyers|Francis Tyers]], [[User:Firespeaker]]
 
  +
| mentors = [[User:Francis Tyers|Francis Tyers]]
* '''[[/UD Annotatrix|read more...]]'''
 
  +
| more = /Annotatrix enhanced dependencies
</div>
 
  +
}}
   
  +
<!--
  +
This one was done, but could do with more work. Not sure if it's a full gsoc though?
  +
  +
{{IdeaSummary
  +
| name = User-friendly lexical selection training
  +
| difficulty = Medium
  +
| skills = Python, C++, shell scripting
  +
| description = Make it so that training/inference of lexical selection rules is a more user-friendly process
  +
| rationale = Our lexical selection module allows for inferring rules from corpora and word alignments, but the procedure is currently a bit messy, with various scripts involved that require lots of manual tweaking, and many third party tools to be installed. The goal of this task is to make the procedure as user-friendly as possible, so that ideally only a simple config file would be needed, and a driver script would take care of the rest.
  +
| mentors = [[User:Unhammer|Unhammer]], [[User:Mlforcada|Mikel Forcada]]
  +
| more = /User-friendly lexical selection training
  +
}}
  +
-->
   
  +
{{IdeaSummary
  +
| name = UD and Apertium integration
  +
| difficulty = Entry level
  +
| size = medium
  +
| skills = python, javascript, HTML, (C++)
  +
| description = Create a range of tools for making Apertium compatible with Universal Dependencies
  +
| rationale = Universal dependencies is a fast growing project aimed at creating a unified annotation scheme for treebanks. This includes both part-of-speech and morphological features. Their annotated corpora could be extremely useful for Apertium for training models for translation. In addition, Apertium's rule-based morphological descriptions could be useful for software that relies on Universal dependencies.
  +
| mentors = [[User:Francis Tyers]], [[User:Firespeaker| Jonathan Washington]], [[User:Popcorndude]]
  +
| more = /UD and Apertium integration
  +
}}
   
 
[[Category:Development]]
 
[[Category:Development]]

Latest revision as of 09:15, 4 March 2024

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.

Current Apertium contributors: If you have an idea please add it below, if you think you could mentor someone in a particular area, add your name to "Interested mentors" using ~~~.

Prospective GSoC contributors: 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.oftc.net (more on IRC), mail the mailing list, or draw attention to yourself in some other way.

Note that if you have an idea that isn't mentioned here, we would be very interested to hear about it.

Here are some more things you could look at:


If you're a prospective GSoC contributor trying to propose a topic, the recommended way is to request a wiki account and then go to

http://wiki.apertium.org/wiki/User:[[your username]]/GSoC2023Proposal

and click the "create" button near the top of the page. It's also nice to include [[Category:GSoC_2023_student_proposals]] to help organize submitted proposals.

Language Data[edit]

Can you read or write a language other than English (and we do mean any language)? If so, you can help with one of these and we can help you figure out the technical parts.


Develop a morphological analyser[edit]

  • Difficulty:
    3. Entry level
  • Size: Multiple lengths possible (discuss with the mentors which option is better for you)
  • Required skills:
    XML or HFST or lexd
  • Description:
    Write a morphological analyser and generator for a language that does not yet have one
  • Rationale:
    A key part of an Apertium machine translation system is a morphological analyser and generator. The objective of this task is to create an analyser for a language that does not yet have one.
  • Mentors:
    Francis Tyers, Jonathan Washington, Sevilay Bayatlı, Hossep, nlhowell, User:Popcorndude
  • read more...


apertium-separable language-pair integration[edit]

  • Difficulty:
    2. Medium
  • Size: Small
  • Required skills:
    XML, a scripting language (Python, Perl), some knowledge of linguistics and/or at least one relevant natural language
  • Description:
    Choose a language you can identify as having a good number of "multiwords" in the lexicon. Modify all language pairs in Apertium to use the Apertium-separable module to process the multiwords, and clean up the dictionaries accordingly.
  • Rationale:
    Apertium-separable is a newish module to process lexical items with discontinguous dependencies, an area where Apertium has traditionally fallen short. Despite all the module has to offer, many translation pairs still don't use it.
  • Mentors:
    Jonathan Washington, User:Popcorndude
  • read more...


Bring an unreleased translation pair to releasable quality[edit]

  • Difficulty:
    2. Medium
  • Size: Large
  • Required skills:
    shell scripting
  • Description:
    Take an unstable language pair and improve its quality, focusing on testvoc
  • Rationale:
    Many Apertium language pairs have large dictionaries and have otherwise seen much development, but are not of releasable quality. The point of this project would be bring one translation pair to releasable quality. This would entail obtaining good naïve coverage and a clean testvoc.
  • Mentors:
    Jonathan Washington, Sevilay Bayatlı, User:Unhammer, Hèctor Alòs i Font
  • read more...


Develop a prototype MT system for a strategic language pair[edit]

  • Difficulty:
    2. Medium
  • Size: Large
  • Required skills:
    XML, some knowledge of linguistics and of one relevant natural language
  • Description:
    Create a translation pair based on two existing language modules, focusing on the dictionary and structural transfer
  • Rationale:
    Choose a strategic set of languages to develop an MT system for, such that you know the target language well and morphological transducers for each language are part of Apertium. Develop an Apertium MT system by focusing on writing a bilingual dictionary and structural transfer rules. Expanding the transducers and disambiguation, and writing lexical selection rules and multiword sequences may also be part of the work. The pair may be an existing prototype, but if it's a heavily developed but unreleased pair, consider applying for "Bring an unreleased translation pair to releasable quality" instead.
  • Mentors:
    Jonathan Washington, Sevilay Bayatlı, User:Unhammer, Hèctor Alòs i Font
  • read more...


Add a new variety to an existing language[edit]

  • Difficulty:
    3. Entry level
  • Size: Multiple lengths possible (discuss with the mentors which option is better for you)
  • Required skills:
    XML, some knowledge of linguistics and of one relevant natural language
  • Description:
    Add a language variety to one or more released pairs, focusing on the dictionary and lexical selection
  • Rationale:
    Take a released language, and define a new language variety for it: e.g. Quebec French or Provençal Occitan. Then add the new variety to one or more released language pairs, without diminishing the quality of the pre-existing variety(ies). The objective is to facilitate the generation of varieties for languages with a weak standardisation and/or pluricentric languages.
  • Mentors:
    Hèctor Alòs i Font, Jonathan Washington,Sevilaybayatlı
  • read more...


Leverage and integrate language preferences into language pairs[edit]

  • Difficulty:
    3. Entry level
  • Size: Medium
  • Required skills:
    XML, some knowledge of linguistics and of one relevant natural language
  • Description:
    Update language pairs with lexical and orthographical variations to leverage the new preferences functionality
  • Rationale:
    Currently, preferences are implemented via language variant, which relies on multiple dictionaries, increasing compilation time exponentially every time a new preference gets introduced.
  • Mentors:
    Xavi Ivars User:Unhammer
  • read more...


Add Capitalization Handling Module to a Language Pair[edit]

  • Difficulty:
    3. Entry level
  • Size: Small
  • Required skills:
    XML, knowledge of some relevant natural language
  • Description:
    Update a language pair to make use make use of the new Capitalization handling module
  • Rationale:
    Correcting capitalization via transfer rules is tedious and error prone, but putting them in a separate set of rules should allow them to be handled in a more concise and maintainable way. Additionally, it is possible that capitalization rule could be moved to monolingual modules, thus reducing development effort on translators.
  • Mentors:
    User:Popcorndude
  • read more...

Data Extraction[edit]

A lot of the language data we need to make our analyzers and translators work already exists in other forms and we just need to figure out how to convert it. If you know of another source of data that isn't listed, we'd love to hear about it.


dictionary induction from wikis[edit]

  • Difficulty:
    2. Medium
  • Size: Multiple lengths possible (discuss with the mentors which option is better for you)
  • Required skills:
    MySQL, mediawiki syntax, perl, maybe C++ or Java; Java, Scala, RDF, and DBpedia to use DBpedia extraction
  • Description:
    Extract dictionaries from linguistic wikis
  • Rationale:
    Wiki dictionaries and encyclopedias (e.g. omegawiki, wiktionary, wikipedia, dbpedia) contain information (e.g. bilingual equivalences, morphological features, conjugations) that could be exploited to speed up the development of dictionaries for Apertium. This task aims at automatically building dictionaries by extracting different pieces of information from wiki structures such as interlingual links, infoboxes and/or from dbpedia RDF datasets.
  • Mentors:
    Jonathan Washington, User:Popcorndude
  • read more...


Dictionary induction from parallel corpora / Revive ReTraTos[edit]

  • Difficulty:
    2. Medium
  • Size: Medium
  • Required skills:
    C++, perl, python, xml, scripting, machine learning
  • Description:
    Extract dictionaries from parallel corpora
  • Rationale:
    Given a pair of monolingual modules and a parallel corpus, we should be able to run a program to align tagged sentences and give us the best entries that are missing from bidix. ReTraTos (from 2008) did this back in 2008, but it's from 2008. We want a program which builds and runs in 2022, and does all the steps for the user.
  • Mentors:
    User:Unhammer, User:Popcorndude
  • read more...


Extract morphological data from FLEx[edit]

  • Difficulty:
    1. Hard
  • Size: Large
  • Required skills:
    python, XML parsing
  • Description:
    Write a program to extract data from SIL FieldWorks and convert as much as possible to monodix (and maybe bidix).
  • Rationale:
    There's a lot of potentially useful data in FieldWorks files that might be enough to build a whole monodix for some languages but it's currently really hard to use
  • Mentors:
    Popcorndude, Flammie
  • read more...

Tooling[edit]

These are projects for people who would be comfortable digging through our C++ codebases (you will be doing a lot of that).


Python API for Apertium[edit]

  • Difficulty:
    2. Medium
  • Size: Medium
  • Required skills:
    C++, Python
  • Description:
    Update the Python API for Apertium to expose all Apertium modes and test with all major OSes
  • Rationale:
    The current Python API misses out on a lot of functionality, like phonemicisation, segmentation, and transliteration, and doesn't work for some OSes like Debian.
  • Mentors:
    Francis Tyers
  • read more...


Robust tokenisation in lttoolbox[edit]

  • Difficulty:
    2. Medium
  • Size: Large
  • Required skills:
    C++, XML, Python
  • Description:
    Improve the longest-match left-to-right tokenisation strategy in lttoolbox to handle spaceless orthographies.
  • Rationale:
    One of the most frustrating things about working with Apertium on texts "in the wild" is the way that the tokenisation works. If a letter is not specified in the alphabet, it is dealt with as whitespace, so e.g. you get unknown words split in two so you can end up with stuff like ^G$ö^k$ı^rmak$ which is terrible for further processing. Additionally, the system is nearly impossible to use for languages that don't use spaces, such as Japanese.
  • Mentors:
    Francis Tyers, Flammie
  • read more...


rule visualization tools[edit]


Extend Weighted transfer rules[edit]


Automatic Error-Finder / Pseudo-Backpropagation[edit]

  • Difficulty:
    1. Hard
  • Size: Large
  • Required skills:
    python?
  • Description:
    Develop a tool to locate the approximate source of translation errors in the pipeline.
  • Rationale:
    Being able to generate a list of probable error sources automatically makes it possible to prioritize issues by frequency, frees up developer time, and is a first step towards automated generation of better rules.
  • Mentors:
    User:Popcorndude
  • read more...


More Robust Recursive Transfer[edit]

  • Difficulty:
    1. Hard
  • Size: Large
  • Required skills:
    C++
  • Description:
    Ensure Recursive Transfer survives ambiguous or incomplete parse trees
  • Rationale:
    Currently, one has to be very careful in writing recursive transfer rules to ensure they don't get too deep or ambiguous, and that they cover full sentences. See in particular issues 97 and 80. We would like linguists to be able to fearlessly write recursive (rtx) rules based on what makes linguistic sense, and have rtx-proc/rtx-comp deal with the computational/performance side.
  • Mentors:
  • read more...


CG-based Transfer[edit]

  • Difficulty:
    1. Hard
  • Size: Large
  • Required skills:
    C++
  • Description:
    Linguists already write dependency trees in Constraint Grammar. A following step could use these to reorder into target language trees.
  • Rationale:
    {{{rationale}}}
  • Mentors:
  • [[|read more...]]


Language Server Protocol[edit]


WASM Compilation[edit]

  • Difficulty:
    1. Hard
  • Size: Medium
  • Required skills:
    C++, Javascript
  • Description:
    Compile the pipeline modules to WASM and provide JS wrappers for them.
  • Rationale:
    There are situations where it would be nice to be able to run the entire pipeline in the browser
  • Mentors:
    Tino Didriksen
  • read more...

Web[edit]

If you know Python and JavaScript, here's some ideas for improving our website. Some of these should be fairly short and it would be a good idea to talk to the mentors about doing a couple of them together.


Web API extensions[edit]

  • Difficulty:
    2. Medium
  • Size: Small
  • Required skills:
    Python
  • Description:
    Update the web API for Apertium to expose all Apertium modes
  • Rationale:
    The current Web API misses out on a lot of functionality, like phonemicisation, segmentation, transliteration, and paradigm generation.
  • Mentors:
    Francis Tyers, Jonathan Washington, Xavi Ivars
  • read more...


Website Improvements: Misc[edit]

  • Difficulty:
    2. Medium
  • Size: Small
  • Required skills:
    html, css, js, python
  • Description:
    Improve elements of Apertium's web infrastructure
  • Rationale:
    Apertium's website infrastructure Apertium-html-tools and its supporting API Apertium APy have numerous open issues. This project would entail choosing a subset of open issues and features that could realistically be completed in the summer. You're encouraged to speak with the Apertium community to see which features and issues are the most pressing.
  • Mentors:
    Jonathan Washington, Xavi Ivars
  • read more...


Website Improvements: Dictionary Lookup[edit]

  • Difficulty:
    2. Medium
  • Size: Small
  • Required skills:
    html, css, js, python
  • Description:
    Finish implementing dictionary lookup mode in Apertium's web infrastructure
  • Rationale:
    Apertium's website infrastructure Apertium-html-tools and its supporting API Apertium APy have numerous open issues, including half-completed features like dictionary lookup. This project would entail completing the dictionary lookup feature. Some additional features which would be good to work would include automatic reverse lookups (so that a user has a better understanding of the results), grammatical information (such as the gender of nouns or the conjugation paradigms of verbs), and information about MWEs.
  • Mentors:
    Jonathan Washington, Xavi Ivars, User:Popcorndude
  • [the open issue on GitHub|read more...]


Website Improvements: Spell checking[edit]

  • Difficulty:
    2. Medium
  • Size: Small
  • Required skills:
    html, js, css, python
  • Description:
    Add a spell-checking interface to Apertium's web tools
  • Rationale:
    Apertium-html-tools has seen some prototypes for spell-checking interfaces (all in stale PRs and branches on GitHub), but none have ended up being quite ready to integrate into the tools. This project would entail polishing up or recreating an interface, and making sure APy has a mode that allows access to Apertium voikospell modules. The end result should be a slick, easy-to-use interface for proofing text, with intuitive underlining of text deemed to be misspelled and intuitive presentation and selection of alternatives. the open issue on GitHub
  • Mentors:
    Jonathan Washington, Xavi Ivars
  • read more...


Website Improvements: Suggestions[edit]

  • Difficulty:
    2. Medium
  • Size: Small
  • Required skills:
    html, css, js, python
  • Description:
    Finish implementing a suggestions interface for Apertium's web infrastructure
  • Rationale:
    Some work has been done to add a "suggestions" interface to Apertium's website infrastructure Apertium-html-tools and its supporting API Apertium APy, whereby users can suggest corrected translations. This project would entail finishing that feature. There are some related issues and PRs on GitHub.
  • Mentors:
    Jonathan Washington, Xavi Ivars
  • read more...


Website Improvements: Orthography conversion interface[edit]

  • Difficulty:
    2. Medium
  • Size: Small
  • Required skills:
    html, js, css, python
  • Description:
    Add an orthography conversion interface to Apertium's web tools
  • Rationale:
    Several Apertium language modules (like Kazakh, Kyrgyz, Crimean Tatar, and Hñähñu) have orthography conversion modes in their mode definition files. This project would be to expose those modes through Apertium APy and provide a simple interface in Apertium-html-tools to use them.
  • Mentors:
    Jonathan Washington, Xavi Ivars
  • read more...


Add support for NMT to web API[edit]

  • Difficulty:
    2. Medium
  • Size: Medium
  • Required skills:
    python, NMT
  • Description:
    Add support for a popular NMT engine to Apertium's web API
  • Rationale:
    Currently Apertium's web API Apertium APy, supports only Apertium language modules. But the front end could just as easily interface with an API that supports trained NMT models. The point of the project is to add support for one popular NMT package (e.g., translateLocally/Bergamot, OpenNMT or JoeyNMT) to the APy.
  • Mentors:
    Jonathan Washington, Xavi Ivars
  • [[|read more...]]

Integrations[edit]

In addition to incorporating data from other projects, it would be nice if we could also make our data useful to them.


OmniLingo and Apertium[edit]

  • Difficulty:
    2. Medium
  • Size: Multiple lengths possible (discuss with the mentors which option is better for you)
  • Required skills:
    JS, Python
  • Description:
    OmniLingo is a language learning system for practicing listening comprehension using Common Voice data. There is a lot of text processing involved (for example tokenisation) that could be aided by Apertium tools.
  • Rationale:
  • Mentors:
    Francis Tyers
  • read more...


Support for Enhanced Dependencies in UD Annotatrix[edit]

  • Difficulty:
    2. Medium
  • Size: Medium
  • Required skills:
    NodeJS
  • Description:
    UD Annotatrix is an annotation interface for Universal Dependencies, but does not yet support all functionality.
  • Rationale:
  • Mentors:
    Francis Tyers
  • read more...


UD and Apertium integration[edit]

  • Difficulty:
    3. Entry level
  • Size: Medium
  • Required skills:
    python, javascript, HTML, (C++)
  • Description:
    Create a range of tools for making Apertium compatible with Universal Dependencies
  • Rationale:
    Universal dependencies is a fast growing project aimed at creating a unified annotation scheme for treebanks. This includes both part-of-speech and morphological features. Their annotated corpora could be extremely useful for Apertium for training models for translation. In addition, Apertium's rule-based morphological descriptions could be useful for software that relies on Universal dependencies.
  • Mentors:
    User:Francis Tyers, Jonathan Washington, User:Popcorndude
  • read more...