Competition

Evaluation and Submission Details

In this shared task, we provide train/dev/test data for 11 languages (i.e. English, Spanish, Dutch, Russian, Turkish, Korean, Farsi, German, Chinese, Hindi, and Bangla), and two language settings: code-mixed and multilingual. As a summary, we provide 13 train/dev/test sets, respectively. This Codalab competition is the practice phase, where you are allowed to submit prediction files for all dev sets. The evaluation framework is divided in three broad types of tracks.

1. Mono-lingual (Tack 1-11): In this track, the participants have to train a model that works only for one language and use that to predict the evaluation sets. For example, when participating Track 1 English, participants can use en_train.conll to train their model and the model will be evaluated by en_dev.conll in practice phase, and en_test.conll in evaluation phase. Predictions from any multi-lingual model is not allowed in this track.
2. Multilingual (Track 12): In this track, the participants have to train a single multi-lingual NER model using training data for all the languages (i.e. multi_train.conll). The training data contains sentences in 11 languages. Note that a sentence in this data only has one language. This trained model should be used to predict multilingual evaluation sets, i.e. multi_dev.conll or multi_test.conll. Predictions from any mono-lingual model is not allowed in this track. Therefore, please do not submit predictions from mono-lingual models in this track. The data does not identify the language of each sentence.
3. Code-mixed (Tack 13): A sentence in this track is a mixture of two languages. The participants can use the given training data (i.e. mix_train.conll) or use the training data of Track 1-12 to train their model. The trained model will be evaluated on mix_dev.conll and mix_test.conll. The data does not identify the languages used in the sentences.

Submission Instructions

Your submissions will be evaluated by macro- Precision, Recall and F1 over 6 entity classes, i.e. LOC, PER, PROD, GRP, CW and CORP. The leaderboard per track will contains these tree metrics and the performance is ranked by macro-F1. To check more detailed evaluation scores after submission, you can either “View scoring output log” or “Download output from scoring step” by checking scores.json.

1. Format of prediction file

The prediction file should follow conll format but only contain tags. i.e. each line contains only the predicted tags of the tokens and sentences are separated by a blank line. Make sure your tags in your prediction file are exactly aligned with the provided dev/test sets. For example,

• Given en_dev.conll/en_test.conll. Note that the 4th column in the data contains entity tags, which will be hidden (replaced by _) in test set.

        # id f423a88e-02b7-4d61-a546-4a1bd89cfa15    domain=dev
        it _ _ O 
        originally _ _ O
        operated _ _ O
        seven _ _ O
        bus _ _ O
        routes _ _ O
        which _ _ O
        were _ _ O
        mainly _ _ O
        supermarket _ _ O
        routes _ _ O
        for _ _ O
        asda _ _ B-CORP
        and _ _ O
        tesco _ _ B-CORP
        . _ _ O
  
        ...
  

• You will need generate the prediction file in the follow format

        # (You can either delete the sentence id or keep it)
        O
        O
        O
        O
        O
        O
        O
        O
        O
        O
        O
        O
        B-CORP
        O
        B-CORP
        O
  
        ...
  

2. Preparing Submission Files

Follow the below instructions to submit your prediction files for a track. Codalab requires all submissions in zip format

• Use your trained specific model to generate a prediction file for a specific tack and name it in this format: <language_code>.pred.conll.
  • For example, when you participate in the English track, you will need to generate a prediction file for en_dev.conll (or en_test.conll in the testing phase) and name it as en.pred.conll.
  • The language_code values for Track 12 Multilingual and Track 13 code mixed are mix and multi, i.e. you will need to name the prediction file as mix.pred.conll or multi.pred.conll.
• Compress it to a zip file by using zip my_submission.zip <language_code>.pred.conll (or your favorite zip utility), and the submit the zip file to the right track on Codalab.

More info and FAQ: https://multiconer.github.io/faq

Terms and Conditions

By submitting results to this competition, you consent to the public release of your scores at the SemEval-2022 workshop and in the associated proceedings, at the task organizers' discretion. Scores may include, but are not limited to, automatically and manually calculated quantitative judgements, qualitative judgements, and such other metrics as the task organizers see fit. You accept that the ultimate decision of metric choice and score value is that of the task organizers.

You further agree that if your team has several members, each of them will register to the competition and build a competition team (as described on the 'Overview' page) and that if you are a single participant you will build a team with a single member.

You further agree that the task organizers are under no obligation to release scores and that scores may be withheld if it is the task organizers' judgement that the submission was incomplete, erroneous, deceptive, or violated the letter or spirit of the competition's rules. Inclusion of a submission's scores is not an endorsement of a team or individual's submission, system, or science.

You further agree that your system may be named according to the team name provided at the time of submission, or to a suitable shorthand as determined by the task organizers.

The data of this competition is released under the CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) license (see lowcontext-ner-gaz (https://registry.opendata.aws/lowcontext-ner-gaz/) and code-mixed-ner (https://registry.opendata.aws/code-mixed-ner/)). Attribution shall be provided by citing:

• Meng, T., Fang, A., Rokhlenko, O., & Malmasi, S. (2021). GEMNET: Effective Gated Gazetteer Representations for Recognizing Complex Entities in Low-context Input (https://www.amazon.science/publications/gazetteer-enhanced-named-entity-recognition-for-code-mixed-web-queries). In Proceedings of NAACL.
• Fetahu, B., Fang, A., Rokhlenko, O., & Malmasi, S. (2021). Gazetteer Enhanced Named Entity Recognition for Code-Mixed Web Queries (https://www.amazon.science/publications/gemnet-effective-gated-gazetteer-representations-for-recognizing-complex-entities-in-low-context-input). In Proceedings of SIGIR.

Organizer List

• Shervin Malmasi, Amazon, USA.

• Besnik Fetahu, Amazon, USA.

• Anjie Fang, Amazon, USA.

• Sudipta Kar, Amazon, USA.

• Oleg Rokhlenko, Amazon, USA.

Contact Us

• Join us in Slack

• Subscribe to the task mailing list

• Contact the organizers