CoNLL 2017 Shared Task System Questionnaire

Request edit access

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

This form elicits information from teams participating in the "CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies". For the overview paper, we need similar information from all teams to be able to summarize the methods, algorithms, resources and software used for all the systems. Please respond carefuly so that we can arrive at meaningful comparisons. Deadline: May 21, 2017.

1. Team name + primary software system ID: *

2. Tokenization, word and sentence segmentation *

My system reads the CoNLL-U files predicted by UDPipe and does not care about segmentation

My system reads the CoNLL-U files predicted by UDPipe but applies its own segmentation tricks on top of it

My system reads the plain text file and segments it on its own

Other:

Required

2a. Comments to Tokenization, word and sentence segmentation (optional)

3. Morphology (lemmas, UPOS, XPOS, features) *

My system uses morphology predicted by UDPipe, provided by the organizers

My system includes a morphological predictor (even UDPipe if you trained it on your own)

I used the morphological transducers from Apertium / Giellatekno.

My system does not care about morphology. We are doing syntax!

Other:

Required

3a. Comments to Morphology (lemmas, UPOS, XPOS, features) (optional)

4. Parsing: describe if your system is based on a single parser or ensamble, identify all parsers used *

5. Word Embeddings *

I used the word embeddings pre-computed by the organizers

I used the crawled raw data from the organizers to compute my own vectors

I used the vectors published by Facebook

I computed my own vectors on other permitted data (please indicate in the comment)

I do not use word vectors

Other:

Required

5a. Comments to Word Embeddings (optional)

6. Additional Data - Apart from the UD 2.0 training data and the raw data mentioned above, my system uses the following data permitted in the shared task (write "None" is no such data is used): *

7. Multilinguality and surprise languages *

I use cross-lingual techniques to build a model for the surprise languages

I trained my model for the surprise languages just on the sample data provided before the test phase

I don’t care about the surprise languages. My system produces a canonical tree without even looking at the input words

I use cross-lingual techniques to get a better model for the small, non-surprise languages (such as Uyghur and Kazakh)

I use domain-adaptation techniques to select the best model for the additional (parallel) test sets

If the additional parallel test set is in a language for which there are multiple training treebanks, I use a model trained on all of them

I simply use one selected model for the language when I encounter an additional parallel test set in a language where multiple training treebanks exist

My system uses one large model for all test sets in the shared task

Other:

Required

7a. Comments to Multilinguality and Surprise Languages (optional)

Submit

Clear form

Never submit passwords through Google Forms.

This content is neither created nor endorsed by Google. - Terms of Service - Privacy Policy

Does this form look suspicious? Report

Forms