Img


NEWS 2018: The Seventh Named Entities Workshop

Shared Task on Named Entity Transliteration

The evaluation phase of NEWS 2018 Shared Task has started on May 7th. Participate at https://competitions.codalab.org/competitions/18905

The test sets are available for download in the Request Datasets page.

Baseline System

You can reproduce the Baseline Transliteration Model results by following this Short Tutorial. The proposed Baseline system is based on Sequitur, a data-driven state-of-the-art open source tool for grapheme to phoneme conversion.

Submission Instructions

Participants can submit several runs for each individual language pair at the competition. However, the system will limit the total number of submissions per language pair to a maximum of 3 submissions per day, with a total maximum of 15 submissions during the competition period: from May 7 to 14, 2018.

From all submissions done to each individual language pair, you must select one to be posted on the corresponding leaderboard. Results posted on the leaderboards by the last day of the competition period (May 14th) will constitute the final offcial results of the Shared Task.

Each submission must be saved in a file named "results.xml" and submitted into the system in a ".zip" compressed file format (please make sure that there is no horizontal tabs in the "results.xml" file, any horizontal tab in the file will cause the submission step to fail). Each "results.xml" file can contain up to 10 output candidates in a ranked list for each corresponding input entry in the test file. Each "results.xml" file should be formated as follows:

<?xml version="1.0" encoding="UTF-8"?>

<TransliterationTaskResults

    SourceLang = "[source-language]"

    TargetLang = "[target-language]"

    GroupID = "[your-institution]"

    RunID = "[your-submission-number]"

    RunType = "Standard"

    Comments = "[your-comments-here]">

<Name ID="1">

    <SourceName>[test-name-1]</SourceName>

    <TargetName ID="1">[your-system-result-1-1]</TargetName>

    <TargetName ID="2">[your-system-result-1-2]</TargetName>

    <TargetName ID="3">[your-system-result-1-3]</TargetName>

    ...

    <TargetName ID="10">[your-system-result-1-10]</TargetName>

</Name>

<Name ID="2">

    <SourceName>[test-name-2]</SourceName>

    <TargetName ID="1">[your-system-result-2-1]</TargetName>

    <TargetName ID="2">[your-system-result-2-2]</TargetName>

    <TargetName ID="3">[your-system-result-2-3]</TargetName>

    ...

    <TargetName ID="10">[your-system-result-2-10]</TargetName>

</Name>

...

<!-- All names in test corpus to follow -->

...

</TransliterationTaskResults>


For more details on how to prepare your submissions, please refer to the Shared Task Whitepaper