Title

Language Technology for the Contact Centers: Opportunities and Challenges

Abstract

Any enterprise that sells a product/service to a customer needs a method to address the requests and challenges faced by their customers. From installation questions to billing inquiries, from product upgrade to troubleshooting, customers expect a frictionless modality to reach the enterprise’s contact center for an expeditious resolution of their issues. Contact center industry is estimated to be $500B, but despite this enormous spend, we have all been victims to poor customer service. In this talk, I will present a few language technologies that have shaped and will continue to transform the industry to deliver a better customer experience.

Bio

Dr. Srinivas Bangalore (CV) is the Senior Vice President of AI at Interactions LLC focused on the complete life cycle of innovation — research to product realization in Conversational AI technologies. Prior to Interactions, Dr. Bangalore was a Principal Research Scientist at AT&T Labs–Research. He has co-edited three books on Supertagging, Natural Language Generation, and Language Translation, has authored over a 100 research publications and holds over 150 patents in the areas of language technology. He is presently a visiting    faculty    at Princeton University    and    Montclair University.

Title

Natural Language Generation (NLG) at LinkedIn: Research & Applications

Abstract

Generative AI models such as ChatGPT and DALL-E, have revolutionized the way we interact with AI systems. Human readable, accurate and personalized content generated by Natural Language Generation (NLG) technologies, have enabled the creation of powerful, personalized experiences for millions of users. In this talk, I will present recent advancements in NLG and discuss model and product innovation directions at the world’s largest professional network – LinkedIn.

Bio

Praveen Kumar Bodigutla is a Senior Staff Applied Machine Learning Researcher, who drives key initiatives in Natural Language Generation (NLG) and Reinforcement Learning at LinkedIn’s Data and AI Foundation team. He has a wealth of diverse experience in machine learning, data science, financial engineering, and software engineering, acquired from working at top firms like LinkedIn, Amazon (Alexa), Credit Suisse and Yahoo! for almost 15 years. He has pushed the boundaries of Conversation AI, Search, NLP and Recommendation Systems, and has written papers in eminent machine learning conferences and workshops. He has masters degrees in Financial Mathematics from Stanford University and Data Sciences from New York University. On the personal front, Praveen is a travel enthusiast and loves having fun and embarking on new experiences.

Title 

Building a dialogue agent for the game of Diplomacy

Abstract 

We recently announced “Cicero,” the first AI capable of playing the game Diplomacy at a human-level. Diplomacy is a complex strategy game involving both cooperation and competition that emphasizes natural language negotiation between 7 players. In this talk, I’ll focus on the language-related components of Cicero and detail our approach to building a negotiation agent.

Bio

Emily Dinan is a Research Engineer at Meta AI (FAIR) in New York. Her research interests include natural language generation and conversational AI, as well as safety and responsibility in these fields. Recently she has focused on methods for controlling generative language models with expert systems.

Title

Creative Text Generation

Abstract

With the advent of large pretrained models, creative generation tasks have garnered much recent interest and have become more viable in practical applications with end users. In this talk, I will describe three lines of creative text generation work from our team at Alexa AI: (1) Pun understanding and generation: We release new datasets for pun understanding and the novel task of context-situated pun generation, and demonstrate the value of our contributions for pun classification and generation tasks. (2) Song lyric generation: We design a hierarchical lyric generation framework that enables us to generate pleasantly-singable lyrics without training on melody-lyric aligned data, and show that our approach is competitive with strong baselines supervised on parallel data. (3) Create with Alexa: We present a multimodal story creation experience recently launched on Alexa devices, which leverages story text generation models in tandem with story visualization and background music generation models to produce multimodal stories for kids.

Bio

Anjali Narayan-Chen is an applied scientist on the Conversational Understanding team at Amazon Alexa AI. Prior to Amazon, she completed her Ph.D. at the University of Illinois at Urbana-Champaign in 2020 advised by Julia Hockenmaier, where she worked on developing collaborative dialogue agents in simulated Minecraft environments. Her current research interests lie in controllable natural language generation and applications to creative AI.

Title

Making Sense of Limited Resources in Cross-Lingual NLP

Abstract

We have witnessed a surge in accurate natural language processing systems for many languages. Such systems heavily rely on annotated datasets. In the absence of such datasets, we should be able to make sense of the available datasets at hand. These include gold-standard annotations from other languages or incidental supervisions available in online resources, such as Wikipedia. We propose a simple but highly effective unsupervised technique to leverage Wikipedia data for creating a highly accurate machine translation model. In some cases, our model’s performance surpasses that of a supervised model. We present our current work on leveraging weakly supervised translation models for cross-lingual transfer, as well as cross-lingual image captioning. We also present our recent work on zero-shot English-centric multilingual machine translation.

Bio

Mohammad Sadegh Rasooli is a senior scientist at Microsoft working on language modeling for speech recognition. He received his Ph.D. in computer science from Columbia University in 2018, advised by Michael Collins. After finishing his Ph.D., he worked at Facebook AI, the University of Pennsylvania, and Microsoft. His research interests lie at the intersection of natural language processing and applied machine learning. He has published many peer-reviewed papers on a variety of NLP tasks, including syntactic and semantic parsing, machine translation, language modeling, sentiment analysis, model transfer, and data annotation.

Title

Building Generalizable, Scalable, and Trustworthy Multimodal Embodied Agents

Abstract 

The development of intelligent agents capable of communicating with humans, perceiving multimodal environments, and acting in the real world has been a long-standing goal of AI research. However, achieving these capabilities presents significant challenges in terms of generalization, scalability, and reliability. 

In this talk, I will introduce three parts aimed at addressing these challenges. First, I will discuss the “last mile problem” of multimodal embodied agent—robotic manipulation following human guidance, and introduce a new compositional benchmark, VLMbench, designed to bridge the gap of vision-and-language robotic manipulation. Second, I will present Counterfactual Prompt Learning (CPL), an efficient transfer learning method for pre-trained vision-and-language models, that enhances generalization and enables agents to understand and reason about complex multimodal environments. Finally, I will introduce the first federated learning framework for scalable and trustworthy embodied agent learning, emphasizing privacy and security considerations for social good. Through these three parts, we aim to advance the development of more generalizable, scalable, and trustworthy multimodal embodied agents.

Bio

Xin (Eric) Wang is an Assistant Professor of Computer Science and Engineering at UC Santa Cruz. His research interests include Natural Language Processing, Computer Vision, and Machine Learning, with a focus on Multimodal and Embodied AI. Before joining UCSC, he obtained his Ph.D. degree from UC Santa Barbara in 2020 and Bachelor’s degree from Zhejiang University in 2015, and interned at Google AI, Facebook AI Research, Microsoft Research, and Adobe Research. 

Xin has served as Area Chair for conferences such as ACL, NAACL, EMNLP, ICLR, and NeurIPS, as well as Senior Program Committee for AAAI and IJCAI. He has also organized numerous workshops and tutorials at conferences such as ACL, NAACL, CVPR, and ICCV. He has received several awards and recognitions for his work, including a CVPR Best Student Paper Award (2019), a Google Research Faculty Award (2022), and three Amazon Alexa Prize Awards (2022-2023).

Title

Grounded Conversational AI with Large Language Models

Abstract

Following the recent advances in large language models and availability of large natural language datasets, conversational AI research has become the center of attention during the last few years, and especially after generative models like chatGPT. The progress also helped to emphasize the importance of grounding over a diverse set of external knowledge and task completion resources for forming relevant, informative, and accurate responses. In this talk, I will discuss some recent work on integrating knowledge to conversation responses from such a diverse set of resources, challenges associated with these, and progress made so far.

Bio

Dr. Gokhan Tur is a leading artificial intelligence expert, especially on human/machine conversational language understanding systems. He has been involved with Apple Siri, Microsoft Cortana, Google Assistant, and Amazon Alexa systems. He co-authored more than 200 papers published in journals or books and presented at conferences. He is the editor of the book entitled “Spoken Language Understanding” by Wiley in 2011.

He received the Ph.D. degree in Computer Science from Bilkent University, Turkey in 2000. Between 1997 and 1999, he was a visiting scholar at the CMU LTI, then the Johns Hopkins University, and the Speech Lab of SRI, CA. At AT&T Research (formerly Bell Labs), NJ (2001-2006) he worked on pioneering conversational systems like “How May I Help You?”. He worked for the DARPA GALE and CALO projects at SRI, CA (2006-2010). He was a founding member of the Microsoft Cortana team, and later the Conversational Systems Lab at Microsoft Research (2010-2016). He worked as the Conversational Understanding Architect at Apple Siri team (2014-2015) and as the Deep Conversational Understanding TLM at Google Research (2016-2018). He was a founding area director at Uber AI (2018-2020). He is currently a Senior Principal Scientist at Amazon Alexa AI.

Dr. Tur is the organizer of the HLT-NAACL 2007 Workshop on Spoken Dialog Technologies, and the HLT-NAACL 2004 and AAAI 2005 Workshops on SLU, and the editor of the Speech Communication Issue on SLU in 2006. Dr. Tur is also the recipient of the IEEE SPS Best Paper Award for 2020, Speech Communication Journal Best Paper awards by ISCA for 2004-2006 and by EURASIP for 2005-2006. He is also the spoken language processing area chair for IEEE ICASSP 2007, 2008, and 2009 conferences and IEEE ASRU 2005 workshop, spoken dialog area chair for HLT-NAACL 2007 conference, NLP applications area chair for HLT-NAACL 2021 conference, and organizer of SLT 2010 workshop.

Dr. Tur is a Fellow of IEEE, and member of ACL and ISCA. He was a member of IEEE Speech and Language Technical Committee (SLTC) (2006-2008), member of the IEEE SPS Industrial Relations Committee (2013-2014) and an associate editor for the IEEE Transactions on Audio, Speech, and Language Processing (2010-2014), and Multimedia Processing (2014-2016) journals.

Title

Foundation Models for NLP and Beyond

Abstract

I will cover the fundamentals behind foundation models (started large language models, or pre-trained language models, but then evolved beyond text) in NLP and a very brief history of foundation models. I will then discuss foundation models for other modalities than NLP and show some very promising results in material science, sensor networks, cybersecurity, and other models we are building at IBM Research.

Bio

Radu Florian is a Distinguished Research Scientist at IBM Research for the last 20 years, where he is also a Senior Manager of the Multilingual Natural Language Technologies department. He has graduated Summa cum Laude from University of Bucharest with a BS in Informatics and has received a PhD from Johns Hopkins University in computational linguistics with the thesis “Transformation Based Learning and Data-Driven Lexical Disambiguation: Syntactic and Semantic Ambiguity Resolution” in 2002, and has been working at IBM Research ever since. He has published over 100 publications in computational linguistics (according to Semantic Scholar), and is an active member of ACL, being senior chair and area chair in several NLP and AI conferences, including AAAI, ACL, NAACL, EMNLP, EACL, AACL. Two true facts and one lie about him: he played water polo for the Romanian Junior team, loves to eat fish, and bungy-jumped at least once in his life.

Multilingual Taxonomic Web Page Classification for Contextual Targeting at Yahoo

Abstract

As we move toward a cookie-less world, the ability to track users’ online activities for behavior targeting will be drastically reduced, making contextual targeting an appealing alternative for advertising platforms. Category-based contextual targeting displays ads on web pages that are relevant to advertiser-targeted categories, according to a pre-defined taxonomy. Accurate web page classification is key to the success of this approach. In this talk, I will discuss the challenges and introduce how we build multilingual Transformer-based transfer learning models to classify web pages at scale. I will also discuss how we apply the proposed models to support category-based contextual targeting at Yahoo.

Bio

XIAO BAI is a Principal Research Scientist at Yahoo Research. She received her PhD in Computer Science from INRIA, France. Her research is primarily focused on information retrieval, natural language processing, and their applications in online advertising. Her contributions to various domains of research have been published in top venues where she regularly serves as PC member, such as SIGIR, CIKM, WWW and KDD.

ChatGPT and Large Language Models

Abstract

Large language models (LLMs) such as ChatGPT have transformed natural language processing by enabling machines to generate human-like language and engage in natural conversations. In this talk, we will begin with a brief overview of the history of language models, and then discuss the latest developments of LLMs. We will explore the capabilities of ChatGPT, how it works, the challenges involved in developing and training such models, and the opportunities ahead.

Bio

Dr. Bing Liu is an Engineering Manager at Meta, where he leads the Natural Language Understanding team in the development of Meta’s AI Assistant for Augmented Reality and Virtual Reality. His team develops cutting-edge NLP and conversational AI technologies to enable the voice AI Assistant on Meta VR Headsets, Smart Displays, and AR Glasses. Before joining Meta, he received his PhD from Carnegie Mellon University, where he worked on deep learning and reinforcement learning for spoken dialog systems. Dr. Liu is an expert in machine learning and natural language processing. He has served on the Senior Program Committees for major NLP, Speech, and AI conferences, including ACL, ICASSP, and AAAI. He also served as the General Chair for the NLP for Conversational Al Workshop at ACL 2022. Dr. Liu has over 30 professional publications and patents.

Title

Conversational AI Becoming Mainstream

Abstract

After decades in the lab, artificial intelligence is becoming part of users’ everyday lives. Neural networks run every time you unlock your iPhone with your face, or when you interact with Siri, the first mainstream intelligent assistant. This talk will introduce how advances in research were needed to improve the user experience.

Bio

Alex Acero is Sr. Director at Apple leading speech recognition, speech synthesis, language understanding, and dialog for Siri, Apple’s personal assistant for iPhone, iPad, Apple Watch, Apple TV, AirPods, Carplay, Macintosh, and HomePod. Prior to joining Apple in 2013, he spent 20 years at Microsoft Research managing teams in speech, audio, multimedia, computer vision, natural language processing, machine translation, machine learning, and information retrieval. His team at Microsoft Research built Bing Translator, contributed to Xbox Kinect, and pioneered the use of deep learning in large vocabulary speech recognition. From 1991-1993 he managed the speech team for Spain’s Telefonica. His first stint at Apple started in 1990. He is Affiliate Faculty at the University of Washington.

Dr. Acero is a Fellow of IEEE and ISCA. He received the 2017 Norbert Wiener Society Award, the 2013 Best Paper Award, and the 2006 Distinguished Lectureship from the IEEE Signal Processing Society. Alex received an engineering degree from the Polytechnic University of Madrid, a Masters from Rice University, and a PhD from Carnegie Mellon. Alex is author of the textbook “Spoken Language Processing”, over 250 technical papers and 150 US patents. Alex has served in a number of advisory boards, including University of Washington’s ECE Advisory Board, IDIAP’s International Advisory Council, Stony Brook University’s Industry Advisory Board, and EPSRC Natural Speech Technology Advisory board.

Title

Enabling domain experts to create their own NLP models: Notes from our journey

 Abstract

Creating AI models for Natural Language Processing (NLP) tasks remains a daunting task for domain experts that have the need but lack the technical expertise and resources to create such models. To address this issue, at IBM Research we have been designing and developing human-in-the-loop systems that enable domain experts to interactively build their own NLP models. In this talk I will summarize our experience with building two such systems focusing on text classification and information extraction, respectively. After a description of the systems, I will focus on the lessons learned during this process and interesting research challenges that lie ahead in our journey to lowering the barrier of entry to NLP.

Bio

Yannis Katsis is a Senior Research Scientist at IBM Research, Almaden with expertise in the management, integration, and extraction of knowledge from structured, semi-structured, and unstructured data. In his recent work, Yannis focuses on lowering the barrier of entry to knowledge extraction by designing, analyzing, and building human-in-the-loop systems that enable domain experts to interactively generate knowledge extraction AI models that serve their needs. Yannis received his PhD in Computer Science from UC San Diego. His work has appeared in top conferences and journals in the areas of data management, natural language processing, and human-computer interaction, and has been leveraged for multiple IBM products.