PAIL-2021

Introduction

Annotated corpora are vital resources for deep learning application development and linguistic analyses. In the case of Indian languages, sufficient annotated quality data are not available publicly for researchers and developers to build upon. Although there are different levels of annotations, syntactically annotated corpora or treebank are very resourceful especially for Indian languages, which are morphosyntactically rich. They are incorporated in downstream applications for various information extraction.

Treebanks are being created for only a few Indian languages and still, there is a high requirement for building more data in different domains while involving other languages. Indian languages do show language-specific complexities that require special attention. When a handful of languages representing different language families of India are ready with quality treebanks, those can be used to build resources for other languages using different approaches like transfer learning and multilingual learning.

There is no way to start with or progress without annotated data; unsupervised approaches are not yet convincing enough, even for resourceful languages like English. Apart from all these justifications, in a task like parsing and treebanking, we need to work together to make resources public and acceptable. Otherwise, we cannot create a meaningful and quality impact.

Objectives

• Bring researchers and developers together who work on treebanks, parsing, and related downstream natural language processing applications

• Provide a platform for researchers to discuss Indian language-specific issues in the morphosyntactic analysis

• Encourage researchers to collaborate and create more annotated resources