Central topics of the workshop are knowledge extraction and integration for deep learning architectures. Following the success of deep learning methods in NLP, the focus of DeeLIO is to explore the linguistic and real-world knowledge that can be extracted from deep learning models, and to identify potentially useful sources of language-related information that can make the models “smarter” and better in natural language understanding tasks.

The Deep Learning Inside Out (DeeLIO) workshop aims precisely to bring together the knowledge interpretation, extraction and integration lines of research in deep learning, and cover the area in between. It will explore the linguistic and world knowledge neural networks encode, how this knowledge is exploited by the models for performing specific tasks, as well as ways to enhance the quality of this knowledge by leveraging external hand-crafted resources.

We especially encourage works on structurally diverse languages and low-data regimes. We hope that DeeLIO will inspire novel variation-aware transfer learning and multilingual solutions on how to use the knowledge from resource-rich languages to inform deep learning architectures where external repositories are scarce or missing.

Topics of interest include but are not limited to:

• Interpretation and explanation methodology (such as probing, adversarial techniques, counterfactual methodology);

• Exploration of the types of linguistic and world knowledge neural models, architectures and representations encode;

• Exploration of the knowledge neural models actually use for performing specific tasks;

• Extraction of linguistic and world knowledge from deep learning models;

• Analysis of the limitations of the knowledge acquired by current neural models;

• Integration of external knowledge into deep learning models (under the form of semantic specialization of embeddings, retrofitting, joint modeling, or other);

• Benefits of using external versus internally encoded knowledge, and their combination, for knowledge enhancement in neural networks;

• Development and enrichment of lexico-semantic knowledge resources using deep learning models;

• Using external knowledge in resource-lean languages through transfer techniques or joint multilingual modelling;

• Analysis of the interplay between prompts and language models in few-shot learning.

Important Dates

• February 28, 2022: Workshop Paper Due Date

• March 21, 2022: Commitment deadline for DeeLIO

• March 26, 2022: Notification of Acceptance

• April 10, 2022: Camera-ready papers due

• May 27, 2022: Workshop

All deadlines are 11:59PM (UTC-12:00 / Anywhere on Earth).