LREC launches a new initiative around these Shared LRs, organising an Open Challenge.
Firmly believing in the power of sharing resources and in the creativeness of the public, encouraging reusability of LRs, reproducibility of language technology (LT) research/experiments, etc., this initiative promotes the "culture" of sharing and re-using LRs and technology for the common good.
We propose a general, open-ended LREC challenge, not privileging a given, pre-defined task, encouraging everyone to participate.
We ask everyone to "use the Shared LRs for whatever use/task they are interested/able to do", from the simplest to the more sophisticated.
We wish to promote and reward creative and original uses of available LRs.
The only condition is that participants should use at least a tool or resource which they did not create themselves.
Challenge contributions will be divided into the following categories:
- production of LRs
- interoperability of LRs
- application of shared LRs
Some (non exhaustive) examples of tasks/usages are:
- using your own tools on third-party shared LRs
- using others’ shared tools for processing your own data, extracting new information, etc.
- testing shared tools on new languages or language variants
- integrating one or more third-party shared LRs/tools into a new application
- merging shared resources, producing new, larger or different LRs
- converting LRs to RDF and/or integrating into the (Linguistic) Linked Open Data cloud
- deploying a new LR platform, with new service interfaces to utilize the LRs, and creating workflows to combine services
- comparing/evaluating similar shared resources (e.g. focusing on sentiment lexicons, reporting on the best performing/simple to use/best documented/easy to replicate/etc. LR)
- using a shared corpus to train a new model
- creative, unforeseen usage of shared LR/tool
- finding uses emphasizing LRs for languages other than English
- exploiting shared LRs for compiling another one, for training a tool, etc.
- application of LRs/tools to benefit less resourced languages.
Along with descriptions of successful experiments and case studies, we also encourage reporting negative results and problems encountered when re-using shared LRs.
We aim to coordinate the Open Challenge with other challenges/hackathons in different LREC workshops in order to facilitate synergies between them and these shared LRs. We thus encourage also the use of shared LRs for replicating/improving results already submitted to shared challenges promoted by different workshops at LREC.
In any case, we do not mean to replace or compete with other shared tasks organized at LREC and its workshops. While double submission is welcome, the broadly defined criteria of the Open Challenge and the more focused shared tasks shall remain sufficiently distinct to guarantee an independent evaluation.
What do we offer?
At LREC we will award prizes based on a number of criteria to be announced (e.g., creativity and originality of the approach, number of resources involved, number of languages, depth and breadth of LR integration, etc. …).
Prizes will also be awarded to the LR that was most exploited!
Depending on the quality and quantity of responses and accompanying reports, we may produce a special LREC publication: the LREC Shared LRs Challenge Proceedings.
We shall also organise one or more special Poster sessions during the Conference
We will also select papers to invite for submissions (of extended version) to the LRE Journal.
Schedule and Report
Participants should send a two-page description of what they have done with the third-party shared LRs by 2 May 2014 using the following format:
Title (related to the use), Author(s), Affiliation
- identifying which Shared LR(s) have been used
- type of usage/applications (see examples above, or other)
- motivation of the usage/experiment
- evaluation of the result
- difficulties in usage/integration
- also negative experiences, well founded
- if the result of the usage is a new LR that can be shared, please provide access information