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

From Apertium
Jump to navigation Jump to search
Line 310: Line 310:
 
| difficulty = hard
 
| difficulty = hard
 
| skills = python, bash, git, XML editing.
 
| skills = python, bash, git, XML editing.
| description = [http://github.com/mlforcada/Appraise], in turn forked from [https://github.com/sereni/Appraise], the work of a GSoC student, contains an adaptation of an old (2014) version of [http://github.com/cfedermann/Appraise] to implement gap-filling evaluation as described in [https://export.arxiv.org/pdf/1809.00315 this WMT2018 paper]. The objective is to bring the gap-filling functionality in [http://github.com/mlforcada/Appraise] to be compatible with the latest versions of Appraise.
+
| description = [http://github.com/mlforcada/Appraise http://github.com/mlforcada/Appraise], in turn forked from [http://github.com/mlforcada/Appraise http://github.com/mlforcada/Appraise], the work of a GSoC student, contains an adaptation of an old (2014) version of [http://github.com/cfedermann/Appraise http://github.com/cfedermann/Appraise] to implement gap-filling evaluation as described in [https://export.arxiv.org/pdf/1809.00315 this WMT2018 paper]. The objective is to bring the gap-filling functionality in [http://github.com/mlforcada/Appraise] to be compatible with the latest versions of Appraise.
 
| rationale = Many language pairs in Apertium are unique, such as Breton-French, and many of them are used for gisting (understanding) purposes. Gap-filling provides a simple way to evaluate the usefulness of machine translation for gisting. The implementation used in recent experiments is based on an outdated version of the Appraise platform.
 
| rationale = Many language pairs in Apertium are unique, such as Breton-French, and many of them are used for gisting (understanding) purposes. Gap-filling provides a simple way to evaluate the usefulness of machine translation for gisting. The implementation used in recent experiments is based on an outdated version of the Appraise platform.
 
| mentors = [[User:Mlforcada|Mikel Forcada]], mentors welcome.
 
| mentors = [[User:Mlforcada|Mikel Forcada]], mentors welcome.

Revision as of 19:08, 4 February 2019

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 ~~~

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.

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:


List

Anaphora resolution for machine translation

  • Difficulty:
    1. Hard
  • Size: default Unknown size
  • Required skills:
    C++, XML, Python
  • Description:
    Write a program to resolve anaphora and include it in the Apertium translation pipeline.
  • Rationale:
    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:
    Francis Tyers
  • read more...

Bring a released language pair up to state-of-the-art quality

  • Difficulty:
    2. Medium
  • Required skills:
    XML, a scripting language (Python, Perl), good knowledge of the language pair adopted.
  • Description:
    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%.
  • Rationale:
    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.
  • Mentors:
    Francis Tyers, Mikel Forcada, Xavi Ivars, Ilnar Salimzianov
  • read more...

Robust tokenisation in lttoolbox

  • Difficulty:
    2. Medium
  • Required skills:
    C++, XML, Python
  • Description:
    Improve the longest-match left-to-right tokenisation strategy in lttoolbox to be fully Unicode compliant.
  • 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.
  • Mentors:
    Francis Tyers, Flammie
  • read more...

Adopt an unreleased language pair

Extend lttoolbox to have the power of HFST

  • Difficulty:
    1. Hard
  • Required skills:
    C++, XSLT, XML
  • Description:
    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.
  • Rationale:
    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.
  • Mentors:
    Mikel Forcada, Tommi A Pirinen, User:Unhammer, mentors wanted
  • read more...


Robust recursive transfer

  • Difficulty:
    1. Hard
  • Required skills:
    Python, XML, linguistics
  • Description:
    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:
    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:
    Francis Tyers, Sortiz, Mikel Forcada, Juan Antonio Pérez
  • read more...

Extend weighted transfer rules

  • Difficulty:
    1. Hard
  • Required skills:
    Python, C++, linguistics
  • Description:
    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.
  • Rationale:
    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.
  • Mentors:
    Francis Tyers, Tommi Pirinen, Sevilay Bayatlı
  • read more...

Improvements to the Apertium website

  • Difficulty:
    3. Entry level
  • Required skills:
    Python, HTML, JS
  • Description:
    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.
  • Rationale:
    https://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 :-)
  • Mentors:
    Jonathan, Sushain
  • read more...

User-friendly lexical selection training

  • Difficulty:
    2. Medium
  • Required 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:
    Unhammer, Mikel Forcada
  • read more...

Light alternative format for all XML files in an Apertium language pair

  • Difficulty:
    1. Hard
  • Required skills:
    Python, C++, shell scripting, XSLT, flex
  • Description:
    Make it possible to edit and develop language data using a format that is lighter than XML
  • Rationale:
    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.
  • Mentors:
    Mikel Forcada, Juan Antonio Pérez, pair.
  • read more...


Bilingual dictionary enrichment via graph completion

  • Difficulty:
    0. Very hard
  • Required skills:
    shell scripting, python, XSLT, XML
  • Description:
    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).
  • Rationale:
    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 ideas that had already been proposed in Apertium or using existing RDF representations of parts of their content, which may benefit from the information coming from being linked to other resources.
  • Mentors:
    Mikel Forcada, Francis Tyers, Jorge Gracia
  • read more... read even more...


UD and Apertium integration

  • Difficulty:
    3. Entry level
  • 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 User:Firespeaker
  • read more...

Add weights to lttoolbox

  • Difficulty:
    1. Hard
  • Required skills:
    c++
  • Description:
    Add support for weighted transducers to lttoolbox
  • Rationale:
    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.
  • Mentors:
    User:Francis Tyers User:Unhammer
  • read more...

Improving language pairs mining Mediawiki Content Translation postedits

  • Difficulty:
    1. Hard
  • Required skills:
    Python, shell scripting, some statistics
  • Description:
    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.
  • Rationale:
    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.
  • Mentors:
    Mikel Forcada, (more mentors to be added)
  • read more...


Unsupervised weighting of automata

  • Difficulty:
    2. Medium
  • Required skills:
    Python, shell scripting, statistics, finite-state transducers
  • Description:
    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:
    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:
    Francis Tyers, Flammie, Unhammer
  • read more...

Improvements to UD Annotatrix

  • Difficulty:
    2. Medium
  • Required skills:
    JavaScript, Jquery, HTML, Python
  • Description:
    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.
  • Rationale:
    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.
  • Mentors:
    Francis Tyers, User:Firespeaker
  • read more...

apertium-separable language-pair integration

  • Difficulty:
    2. Medium
  • 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 newly developed module to process lexical items with discontinguous dependencies, an area where Apertium has traditionally fallen short. Despite all the module has to offer, it has only been put to use in small test cases, and hasn't been integrated into any translation pair's development cycle.
  • Mentors:
    Francis Tyers, User:Firespeaker
  • read more...


Create FST-based module for disambiguating

  • Difficulty:
    2. Medium
  • Size: default Unknown size
  • Required skills:
    XML, a scripting language (Python, Perl), C++, finite-state transducers
  • Description:
    Implement a Constraint Grammar-like module based on finite-state transducers.
  • Rationale:
    Currently, many language pairs use Constraint grammar as a pre-disambiguator for the Apertium tagger, allowing the imposition of more fine grained constraints than would be otherwise possible. However, current implementation of CG is much slower than most of the other modules in the Apertium pipeline, and it's also very different in terms of syntax to other Apertium modules (dictionaries, lexical selection, transfer rules, etc). There have been a few attempts to create FST versions of CG (see User:David_Nemeskey/GSOC_progress_2013), but they haven't succeeded. The hypothesis is that a simpler version of CG that supports the main features that CG support (no need to feature parity) would have better adoption and integration within the Apertium pipeline.
  • Mentors:
    Xavi Ivars, Francis Tyers
  • read more...


Python API/library for Apertium

  • Difficulty:
    2. Medium
  • Size: default Unknown size
  • Required skills:
    Python, C++, SWIG
  • Description:
    Implement a Python library for Apertium and Lttoolbox.
  • Rationale:
    Lots of people use Python, they like to use it in their Jupyter notebooks and on Microsoft Windows™. Apertium is really hard to get going in these kind of environments. So it would be cool if we could make Apertium work for them too. It has a lot of nice language processing tools that we would like more people to use. I'm sure people would love to "pip install apertium" or "pip install apertium-ava". The API/implementation should be pythonistic.
  • Mentors:
    Francis Tyers, Unhammer, Xavi Ivars, Sushain
  • read more...


TIPP functionality for Apertium

  • Difficulty:
    2. Medium
  • Size: default Unknown size
  • Required skills:
    Python, C++, XML, XLIFF (a subset called XLIFF:doc)
  • Description:
    TIPP, the TMS Interoperability Protocol Package (where TMS means translation management system), [1], currently in version 1.5 but being upgraded to 2.0, specifies a container (package format) that allows the interchange of information along a translation value chain. There are various container varieties for different tasks. One such variety, called Translate-Strict-Bitext, represents a bilingual translation job. The 'request' TIPP would contain an XLIFF:doc ([2], a subset of XLIFF 1 [3]) file with the document to be translated and the corresponding metadata, and the corresponding 'response' TIPP would contain the results of Apertium MT applied to it, but taking into account the translation memory provided in the TIPP, if any (using Apertium's -m switch). Apertium should be endowed with the capacity to manage TIPP packages: unpack the request package, parse it, process it, and repack it.
  • Rationale:
    {{{rationale}}}
  • Mentors:
    Mikel Forcada
  • read more...


Bring Apertium Occitan--French closer to posteditable quality.

  • Difficulty:
    2. Medium
  • Size: default Unknown size
  • Required skills:
    GNU/Linux advanced user, bash, git, XML editing, standard Occitan, French.
  • Description:
    The idea is to make Occitan output easier to postedit and French output easier to understand. This entails increasing the monolingual and bilingual dictionaries, improving disambiguation, and writing new structural transfer rules.
  • Rationale:
    The Occitan--French language pair has been recently published. This language pair is of strategic importance for the Occitan language, as Apertium offers the only machine translation system for this language pair.
  • Mentors:
    Mikel Forcada, mentors welcome.
  • read more...



Create a usable version of one of these language pairs: English--Igbo, English--Yoruba, English--Tigrinya, English--Swahili, English-Hausa

  • Difficulty:
    2. Medium
  • Size: default Unknown size
  • Required skills:
    GNU/Linux advanced user, bash, git, XML editing, English, Igbo/Yoruba/Tigrinya/Swahili/Hausa
  • Description:
    The objective is to start these language pairs (which haven't been started or have currentlu very little data in Apertium) and write an usable version which provides intelligible output.
  • Rationale:
    African languages are not particularly well served by Apertium. The four languages listed are quite important, and are only currently served by commercial machine translation companies such as Google, which makes these language communities dependent on a specific commercial provider.
  • Mentors:
    Mikel Forcada, mentors welcome.
  • read more...


Set up gap-filling machine-translation-for-gisting evaluation on a recent version of Appraise

  • Difficulty:
    1. Hard
  • Size: default Unknown size
  • Required skills:
    python, bash, git, XML editing.
  • Description:
    http://github.com/mlforcada/Appraise, in turn forked from http://github.com/mlforcada/Appraise, the work of a GSoC student, contains an adaptation of an old (2014) version of http://github.com/cfedermann/Appraise to implement gap-filling evaluation as described in this WMT2018 paper. The objective is to bring the gap-filling functionality in [4] to be compatible with the latest versions of Appraise.
  • Rationale:
    Many language pairs in Apertium are unique, such as Breton-French, and many of them are used for gisting (understanding) purposes. Gap-filling provides a simple way to evaluate the usefulness of machine translation for gisting. The implementation used in recent experiments is based on an outdated version of the Appraise platform.
  • Mentors:
    Mikel Forcada, mentors welcome.
  • read more...