Manas Gaur

3rd KDD Workshop on Knowledge-infused Learning

2023-04-10T00:00:00+00:00

Dear Colleagues and Students,

We are organizing the 3rd International Workshop on Knowledge-infused Learning in conjunction with KDD 2023.

Link: https://aiisc.ai/kiml2023

This year we are having a special theme session on NeuroSymbolic AI. We invite contributions along novel methods, datasets, and evaluation metrics.

Important Dates:

Paper submission deadline: May 23, 2023 (possible extension because of delayed decision from KDD)
Author notification: June 23, 2023
Camera-ready: June 27, 2023
Workshop Date: August 6-10, 2023

Easychair

Submission Link

Call for Papers

Topics of interest, for the technical, poster, and demo sessions, include, but are not limited to:

Topics of interest include but are not limited to, the following:

* Methods:

    * NeuroSymbolic AI

    * Knowledge-injected foundational language models 

    * Injecting Knowledge-constraints in Deep Neural Networks

    * Automated Rule Learning and Inference

    * Knowledge-guided Active Learning

    * Knowledge-injected Reinforcement Learning

    * Multimodal Knowledge-injected Learning

* Responsible AI:

    * User-Explainable Machine Learning and Deep Learning

    * User-Safety in Conversational Systems

* Evaluation:

    * Metric to assess Safe AI

    * Metric to assess User-Explainable AI

    * Metric to assess Knowledge-infusion in AI

* Application domains (not limited to):

    * Healthcare

    * Legal

    * Cybersecurity

    * Autonomous Systems

Submissions:

* Full Papers (9 Pages)

* Short Papers (6 Pages)

* Position Papers (4-6 Pages)

* Demo Paper (4 Pages)

* Industry Paper (4-9 Pages)

All the limits include references. The workshop submission template follows ACM Sigconf and accepted paper will be published as a part of CEUR Workshop Proceedings.

Committee

Steering Committee:

Amit Sheth, AI Institute, University of South Carolina, USA
Srinivasan Parathasarathy, Ohio State University, USA
Biplav Srivastava, AI Institute, University of South Carolina, USA

Organizing Committee:

Manas Gaur, University of Maryland Baltimore County, USA (prime contact; manas@umbc.edu)
Efthymia Tsamoura, Samsung AI and Alan Turing Institute, UK
Sarath Sreedharan, Colorado State University, USA
Sudip Mittal, Mississippi State University, USA
Kaushik Roy, AI Institute, University of South Carolina, USA

PS: Please forward this weblink to your colleagues and students.

Information Disguise (A New Domain in NLP)

2023-03-15T00:00:00+00:00

Author’s identity is important when their content is used to showcase some effective and useful research direction in domains like health and well-being, cyber social threats, and others.

Excited to be a part of this new research with Joseph Reagle and Precog Lab in IIIT Hyderabad

Disguising Reddit sources and the efficacy of ethical research (starting point of the research on Information disguise)
Spinning words as disguise: Shady services for ethical research?
Towards Effective Paraphrasing for Information Disguise in ECIR 2023

Explainable and Safe Chatbot in AAAI 2023 and UMBC SURFF Award

2023-03-15T00:00:00+00:00

Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable. from “Letter on Pause Giant AI”

This trust must be well-founded because guaranteeing system safety calls for more than just increasing precision, efficacy, and scalability. It also calls for making sure that systems are resilient to extreme occurrences and keeping an eye out for unusual and risky behavior.

ALLEVIATE came up in AAAI 2023 Demo
Awarded UMBC seed faculty award for developing a Safe and Explainable Conversational Agent with focus on question generation.
Amazing work from my student Surjodeep Sarkar in putting together the first critical review of AI-based conversational agents, conversational datasets, and evaluation metrics, from the perspective of safety and explainability.

Samsung Research Patent on Question Generation

2023-03-01T00:00:00+00:00

Question Generation is important in AI if it is from an AI agent. This is because it shows a sense of curiosity and ensures that an AI agent is following a conceptual flow while maintaining contextual interaction with the user.

Better things can happen with such a method:

Workflow-driven language Modeling
Minimize generation which has no trace to factual information ( A work in progress)

Excited about my first and upcoming Patent with Samsung Research America on “Dynamic question generation for information-gathering”.

Award from RAAPID.AI, New Beginnings @ UMBC, Tutorial at ICON and AI/ML Systems Conference

2022-09-06T00:00:00+00:00

Started as an Assistant Professor in the Department of Computer Science and Electrical Engineering at The University of Maryland Baltimore County.

Fortunate to recieve unrestricted gift from RAAPID for research on Conversational Systems.
Current Research Directions:
- Personalized Dialog Systems in Mental Healthcare
- Explainable AI
- Knowledge-infused Learning
- Knowledge Graphs
For UMBC Students: I am actively looking students for Directed Study or Independent Study at UMBC.
For External Students: I would be able to support you after we have accomplished atleast 3 months in collaborative research.
Do learn about excited projects in CSEE @ UMBC
Excited to share that following tutorials will be delivered in ICON 2022 and AI/ML Systems Conferences.
- Neuro-symbolic AI for Mental Health at AI/ML Systems 2022
- Process Knowledge-infused Reinforcement Learning at ICON 2022

AAAI New Faculty Highlights, Safety Data Initiative Talk, Carolina TV, Samsung Research

2022-09-06T00:00:00+00:00

Excited to be selected for the AAAI New Faculty Highlights 2023!! See you in Washington DC where I will be talking about Towards Process Knowledge-infused Learning For Safety and Explainability.

On December 07, 2022, I will be talking to Public.io on Online Safety Data Initiative. Looking forward to it!!

Thank you to University of South Carolina for considering me for USC Today Web Story.

Will be joining Samsung Research America’s Knowledge and Dialog Team as Visiting Faculty in 2023!!

CLPsych Shared Task @ NAACL 2022

2022-03-01T00:00:00+00:00

Excited to be a part of CLPsych Shared Task @ NAACL 2022, along with Adam Tsakilidis, Maria Liakata, Iman Bilal, Dana Atzil-Slonim, Ayah Zirikly, Federico Nanni, Philip Resnik, and Kaushik Roy

About the Task: Assessing user-level temporal mood changes based on users’ online content

Description: The shared task is comprised of two sub-tasks:

Identification of segments within posts at user-level that shows mood changes;
Following (1) assess the risk level of the user

More Details: https://lnkd.in/dQ24-YwQ

Email for Questions: clpsych22-shared-task-organizers@googlegroups.com

PhD Defense March 25, 2022!! PhinishedD

2022-03-01T00:00:00+00:00

SLIDES, VIDEO

Thesis Statement

Knowledge-infused Learning is a class of Neuro-Symbolic AI techniques that incorporate broader forms of knowledge (lexical, domain-specific, common-sense, and constraint-based) into addressing limitations of either symbolic or statistical AI approaches, such as model interpretations and user-level explanations. Compared to powerful statistical AI that exploit data, KiL benefit from data as well as knowledge.

Abstract

My Ph.D. Defense investigate the knowledge-infusion strategy in two important ways. The first is to infuse knowledge to make any classification task explainable. The second is to achieve explainability in any natural language generation tasks. The defense demonstrated the effective strategies of knowledge infusion that bring five characteristic properties in any statistical AI model:

Context Sensitivity,
Handling Uncertainty and Risk,
Interpretable in learning,
User-level Explainability,

across natural language understanding (NLU) tasks. Along with proven methodological contributions in AI made by the dissertation, it also introduced Knowledge-intensive Language Understanding tasks, a variant of General Language Understanding (GLUE) tasks that challenges AI and NLU research on explainability and interpretability.

Furthermore, the Defense showcased the utility of incorporating diverse forms of knowledge: linguistic, commonsense, broad-based, and domain-specific. As the Defense illustrated the success in various domains, achieving state-of-the-art in specific applications, and significant contributions towards improving the state of machine intelligence, it also highlighted careful steps to prevent errors arising due to knowledge infusion. The Defense also laid out future research direction towards Deep Knowledge Infusion, which would be pivotal in propelling machine understanding.

Accepted Paper in ACL 2022!!

2022-02-26T00:00:00+00:00

With advances in Natural Language Processing strategies, it is now possible to design automated systems to assess suicide risk. However, such systems may generate uncertain predictions, leading to severe consequences. We hence reformulate suicide risk assessment as a selective prioritized prediction problem over the Columbia Suicide Severity Risk Scale (C-SSRS). We propose SASI, a risk-averse and self-aware transformer-based hierarchical attention classifier, augmented to refrain from making uncertain predictions. We show that SASI is able to refrain from 83% of incorrect predictions on real-world Reddit data. Furthermore, we discuss the qualitative, practical, and ethical aspects of SASI for suicide risk assessment as a human-in-the-loop framework.

ACL 2022 Paper

How this study contribute to the chain of research on Mental Health, Knowledge Graphs, Knowledge-infused Learning, and Human-AI Collaboration?

Accepted Tutorial in Knowledge Graph Conference 2022!!

2022-02-24T00:00:00+00:00

Excited to collaborate with Kaushik Roy, Qi Zhang, and Amit Sheth on Knowledge Graph Conference Tutorial on Knowledge-infused Reinforcement Learning

Abstract

This tutorial seeks to showcase AI strategies that provide medical context to patient data with the help of a knowledge graph. This supports personalization through a personalized knowledge graph that captures the patient’s personalized health management objectives within the context of the clinical guidelines and care plan. The continuous capture of this information through the analysis of patient-VHA interactions, and the strategy of creating engaging interactions (conversations) can further augment the personalized knowledge graph. These operations are required to support self-appraisal and self-management, and when necessary perform fail-safe tasks such as connecting the patient to a crisis help-line or professional help. The core innovation is the use of a novel knowledge-infused reinforcement learning method. The by-product of this approach leads to transparency in decision-making with the ability to offer a user understandable explanation.

Illustration of Knowledge-infused Reinforcement Learning using knowledge graph and process knowledge in clinical guidelines to support patients in their goal management.