|Full Title:||Workshop on Building Linguistically Generalizable NLP Systems (at EMNLP 2017)|
|Start Date:||08-Sep-2017 - 08-Sep-2017|
|Contact:||Hal Daumé III|
|Meeting Email:||click here to access email|
|Meeting Description:||Machine learning techniques have had immeasurable positive impact on the field of natural language processing, to the point that we now have systems for many NLP problems that work incredibly well, at least when the NLP problem is carefully designed, and these systems are tested on data that looks like their training data. Especially with the influx of deep learning approaches to NLP, we find ourselves more and more in the situation that we have systems that work well under some conditions, but have little idea what those conditions are.
We believe that linguistic knowledge is indispensable in many phases of the NLP pipeline, including:
- Task design and choice of language(s)
- Annotation schema design
- System architecture design and/or feature design
- Evaluation design and error analysis
- Generalization beyond training data
Our goal in this workshop is to bring together researchers from NLP and Linguistics through a carefully designed shared task. This shared task is designed to test the true generalization ability of NLP systems beyond the distribution of data on which they may have been trained. Details will be released in early 2017.
This workshop will take place at EMNLP 2017, in Copenhagen, Denmark.
|Linguistic Subfield:||Computational Linguistics; General Linguistics|
|Calls and Conferences main page|