Our submission to the MADAR shared task on Arabic dialect identification employed a language modeling technique called Prediction by Partial Matching, an ensemble of neural architectures, and sources of additional data for training word embeddings and auxiliary language models.
This work considers a task from traditional literary criticism: annotating a structured, composite document with information about its sources.
There is a long-standing interest in understanding the internal behavior of neural networks.
Twitter is a ubiquitous source of micro-blog social media data, providing the academic, industrial, and public sectors real-time access to actionable information.
It combines a multitude of analytics together with a flexible environment for customizing the workflow for different users.