Conversational AI & NLP Summit schedule

Jinjin is a Senior Applied/ML Scientist at Amazon with 6+ years of research and practical application experience in a few domains (Supply Chain, Retail, Advertising and recommender systems, Voice assistant, AI Education). Jinjin has been devoted to AI Education since early 2018. During the past 2-3 years, she was able to publish 10 research papers at various conferences (ACM [email protected], AAAI, EducationDataMining, etc) and there are more under double-blind peer review. She is enthusiastic at research and bringing researchers together for bigger wins and also leads the dWdt initiative for mentoring and guiding junior scientists and engineers for their growth in research and practical AI applications.

NLP for Music and Audio: Challenges and Opportunities

A vast amount of music information available on social media, web pages, online forums, and digital libraries, etc., is represented in natural language. Making sense of this information is challenging due to the unstructured nature of the data. Meanwhile, the AI landscape is becoming increasingly multimodal. In this talk, I will bring together the recent developments at the intersection of NLP, Music Information Retrieval (MIR) and audio AI. We will walk through the challenges of applying NLP in MIR that enables machines to make sense of the world through multimodal music and sound data, including building a music voice assistant. I will also report on the recent developments in the audio AI community and discuss how NLP plays a role in it.

Shuo Zhang is a Senior ML Research Engineer at Bose Corp., where he works on machine learning and deep learning for audio signal processing and NLP applications. Prior to Bose, he worked at the Music Technology Group, Universitat Pompeu Fabra in Barcelona. Shuo received his PhD from Georgetown University with a focus on computational linguistics. In 2016, he co-taught a tutorial on the application of NLP in Music Information Retrieval at the ISMIR conference. This year, Shuo serves as the Co-Chair of Industry Liaisons of the DCASE 2021 conference - the leading conference of audio AI.

PRESENTATION: NLU & Computer Vision

The explosion of user-generated content, endlessly growing interactions between organizations and customers, regulators and companies, federal institutions and citizens, employees and employers are creating both Big Content problems and new business opportunities. Individuals are expressing their needs, suggestions and concerns across a wide variety of content types that can help organizations to make optimal data-driven decisions. At the same time, the amount of available content can be overwhelming and no longer possible to monitor by Content Managers. This is the content that might quickly prove to be harmful for the organization, creating very serious issues including legal consequences as any discriminatory, sexist, racist language used in emails or corporate social media along with inappropriate pieces of text or images should not have a place in the workplace or digital communities. In this session, I'll give an overview of how Natural Language Understanding (NLU) gives a manner to analyze, identify and regroup these opportunities and risks through automated classification, named-entity extraction as well as the analysis of subjectivity, tonality, emotions & intentions within the textual content, and how NLU combined with Computer Vision applications can identify high-risk content.

Key Takeaways: *The boundary between NLP and NLU is not that obvious

*Voice of CCE and Risk Assessment are great examples of Text Analytics business use cases

*NLP is not enough for text analytics; other technologies are needed such as Computer Vision

ROUNDTABLE: Extended Q&A & Demo with Robert

Join this session to ask your pressing questions and see NLU & Computer Vision applied in action. Demos will be shown of applications leveraging these new technologies and the use cases they support, including:

*Analyzing the voice of the customer

*Uncovering high-risk text and visual content

*Behind the scenes of preconfigured REST APIs

Robert Kapitan is the Lead Product Manager at OpenText for the AI & Analytics content analytics platform, Magellan Text Mining. Robert has been working with Text Mining and Content Analytic applications for over 20 years helping to build software solutions that will understand human language. He holds an M.A. in Theoretical Linguistics and a PhD in Cognitive Semantics.

Generic ASR Will Never be Accurate Enough for Conversational AI

Most conversational AI use cases don’t actually need text, they need intent. That is why there are challenges in providing ASR to conversational models. Was that speaker part of the conversation or were they just background that should be ignored? Should the model have transcribed ‘Hey Frank! Make sure pineapple is on there!’ or is that an error? A core issue is that ASR not only provides a transcript, it also decides things like what was part of the conversation and what wasn’t. Learn why one-size-fits-all ASRs lack understanding of intent and context and why a domain-targeted ASR is essential for Conversational AI.

Jeff "Susan" Ward is a Research Engineer at Deepgram, where for the past four years, he has been exploring and innovating technological solutions in the realm of automatic speech recognition. His work has focused on automating the entire training pipeline with the intent to enable rapid customization across a variety of ASR use cases. He also has experience in automatic alignment, transcript cleaning, large-scale data management, automated training, and model design. Before joining Deepgram, Susan earned his master's from the University of Edinburgh and his pilot wings from the US Navy.

Conversational AI in Finance: Personalized Customer Engagement at Scale

From Chatbots, Contact Centers and Speech & Text analytics to Financial Advice, Conversational AI plays a center piece in banks’ strategies to for providing personalized services, product and advice to their customers at scale and to improve internal efficiencies. In this presentation I will describe a series of applications of Conversational AI in banking, the rewards, challenges, and some solutions to overcome them.

Key Takeaways:

*Demand for Conversational AI is boosted by the exploding need for virtual client engagement solutions during the Covid-19 pandemic

*Conversational AI applications range from chatbots and contact centers to improve quality of service without a cost explosion, to financial advice applications

*Success and adoption depend on business buy-in, on frictionless customer experience, omnichannel end to end integration, and agile continuous improvement

Simona Gandrabur has been working in the general field of AI for close to 20 years, most notably in areas related to processing of human languages – such as automatic speech recognition, natural language understanding, machine translation and conversational AI. Her experience ranges from many years in research, in the development of intelligent assistant applications, to defining strategy of AI-based offers. After 10 years at Nuance Communications she joined the National Bank of Canada in 2018 as AI & Data Business Strategy Lead.

Extracting Conversation Highlights using Dialog Acts

Online messaging platforms and virtual meetings have been a dominant mode of communication for many years now and more so in recent times. However, processing and condensing this data into useful snippets of information is still an on ongoing research problem. In this talk, I will introduce the concept of dialog acts to understand some semantics of these conversations and use them to highlight useful, actionable information. We will talk about dialog act datasets, models and their applications in different modes of conversation like multi-party meetings and chat messages.

Key Takeaways:

*Meeting summaries are subjective and collecting annotations are expensive, but research shows participants find actionable-items to be key takeaways.

*Speech acts are used to understand the structure of a conversation and a subset of them can serve as a proxy for identifying these actionable-items.

*Pretrained models fine-tuned even on small datasets can achieve good accuracies in identifying speech acts that are useful for the task.

Varsha Embar is a Senior Machine Learning Engineer at MindMeld, Cisco, where she builds production level conversational interfaces. She works on improving the core Natural Language Processing platform, including features and algorithms for low-resource settings, and tackles challenging problems such as summarization and action item detection in noisy meeting transcripts. Prior to MindMeld, Varsha earned her Master’s degree in Machine Learning and Natural Language Processing from Carnegie Mellon University.

Cross-Lingual Transfer Learning

Research in natural language processing (NLP) has seen striking advances in recent years, mainly driven by large pretrained language models. However, most of these successes have been achieved in English and a small set of other high-resource languages. In this talk, I will highlight methods that enable us to scale NLP models to more of the world's 7,000 languages, challenges, and promising future directions.

Key takeaways: *Multilingual models are necessary in order to democratise access to language technology.

*Large language models trained on unlabelled data in many languages learn a surprising amount of cross-lingual information.

*Nevertheless, there are many open challenges such as generalisation to languages with limited data.

Sebastian Ruder is a research scientist in the Language team at DeepMind, London. He completed his PhD in Natural Language Processing and Deep Learning at the Insight Research Centre for Data Analytics, while working as a research scientist at Dublin-based text analytics startup AYLIEN. Previously, he studied Computational Linguistics at the University of Heidelberg, Germany and at Trinity College, Dublin.

PRESENTATION: NLU & Computer Vision

The explosion of user-generated content, endlessly growing interactions between organizations and customers, regulators and companies, federal institutions and citizens, employees and employers are creating both Big Content problems and new business opportunities. Individuals are expressing their needs, suggestions and concerns across a wide variety of content types that can help organizations to make optimal data-driven decisions. At the same time, the amount of available content can be overwhelming and no longer possible to monitor by Content Managers. This is the content that might quickly prove to be harmful for the organization, creating very serious issues including legal consequences as any discriminatory, sexist, racist language used in emails or corporate social media along with inappropriate pieces of text or images should not have a place in the workplace or digital communities. In this session, I'll give an overview of how Natural Language Understanding (NLU) gives a manner to analyze, identify and regroup these opportunities and risks through automated classification, named-entity extraction as well as the analysis of subjectivity, tonality, emotions & intentions within the textual content, and how NLU combined with Computer Vision applications can identify high-risk content.

Key Takeaways: *The boundary between NLP and NLU is not that obvious

*Voice of CCE and Risk Assessment are great examples of Text Analytics business use cases

*NLP is not enough for text analytics; other technologies are needed such as Computer Vision

ROUNDTABLE: Extended Q&A & Demo with Robert

Join this session to ask your pressing questions and see NLU & Computer Vision applied in action. Demos will be shown of applications leveraging these new technologies and the use cases they support, including:

*Analyzing the voice of the customer

*Uncovering high-risk text and visual content

*Behind the scenes of preconfigured REST APIs

Robert Kapitan is the Lead Product Manager at OpenText for the AI & Analytics content analytics platform, Magellan Text Mining. Robert has been working with Text Mining and Content Analytic applications for over 20 years helping to build software solutions that will understand human language. He holds an M.A. in Theoretical Linguistics and a PhD in Cognitive Semantics.

ROUNDTABLE: AI Ethics and Model Risk Q&A with Angelina Yang & Mehdi Allahyari

Join this session to ask your questions and see how financial institutions identify and manage risks relating to AI ethics, including transparency and explainability, in the development, deployment and ongoing use of AI systems.

Angelina Yang leads the data science and engineering teams for conversational and voice AI development at Well Fargo. Her focus is to research, develop and deploy machine learning solutions for the next generation of AI assistants across the enterprise. Her teams are committed to the advancement of AI driven by not only efficiency and performance, but also ethical principles that put customers first. She drives and fosters innovation in responsible AI research across her organization. In the past year, she led more than 10 patent applications in the area of AI transparency, fairness analysis and interpretability for machine learning models. Previously, she led the NLP Corporate Model Risk Management team and led the development of model risk governance policy for machine learning/NLP models. She is a buckeye from the Ohio State University majoring in statistics. She has a wealth of experience in the space of statistical and machine learning, management consulting and policy research at Bank of Tokyo, KPMG Advisory and the United Nations Department of Economics and Social Affairs.

ROUNDTABLE: AI Ethics and Model Risk Q&A with Angelina Yang & Mehdi Allahyari

Join this session to ask your questions and see how financial institutions identify and manage risks relating to AI ethics, including transparency and explainability, in the development, deployment and ongoing use of AI systems.

Mehdi Allahyari is a Lead Data Scientist at Wells Fargo. His focus is on natural language processing, knowledge graphs, question answering, and conversational AI. He has More than 15 years of software development and research experience in industry and academia. Before joining Wells Fargo, Mehdi was an assistant professor in computer science at Georgia Southern University. He worked on topics ranging from Semantic Web, topic modeling, to information extraction for natural language text applications such as recommendation, summarization, and chatbot development.

Maria is a data journalist and subject matter expert on conversational interfaces. She is responsible for exploring new conversational A.I. technologies for 80+ Nestlé global markets. Maria works directly with senior business stakeholders to design, develop and launch multiple consumer-focused chatbots on the Nespresso, Nescafé Dolce Gusto, Maggi and Nestlé Infant Nutrition brands platforms.

Private Distributed Learning in a Byzantine World

The ever-growing number of edge devices (e.g., smartphones) and the exploding volume of sensitive data they produce, call for distributed machine learning techniques that are privacy-preserving. Given the increasing computing capabilities of modern edge devices, these techniques can be realized by pushing the sensitive-data-dependent tasks of machine learning to the edge devices and thus avoid disclosing sensitive data.

I will present two important challenges in this new computing paradigm along with an overview of our proposed solutions to address them. First, for many applications, such as news recommenders, data needs to be processed fast, before it becomes obsolete. Second, given the large amount of uncontrolled edge devices, some of them may undergo arbitrary (Byzantine) failures and deviate from the distributed learning protocol with potentially negative consequences such as learning divergence or even biased predictions.

Key Takeaways:

*Our data is extremely valuable and vulnerable => let's push it to the "Edge"

*Machine Learning at the Edge is possible yet challenging due to (a) temporality of the data and (b) unreliability of the machines

Georgios is a machine learning engineer at Facebook London, focusing on natural language processing. He received his Ph.D. from EPFL in September 2020, where he worked under the supervision of Rachid Guerraoui. Before joining EPFL, he received his MEng in Electrical and Computer Engineering from NTUA. His research focuses on distributed machine learning techniques that are privacy-preserving and robust against arbitrary failures (such as adversarial attacks). He is mainly a practitioner but also studies algorithmic tools from a theoretical perspective. His work has led to publications in multiple premier conferences such ICML and AAAI while he has also won several awards including the EPFL Ph.D. fellowship and the best paper award in Middleware 2020. More about Georgios: https://gdamaskinos.com/

As a product owner for the Autonomous Digital Assistant, Shyamala currently leads the company-wide Ford's Autonomous Digital Assistant innovation for voice and chatbots by owning the roadmap and overall vision and working with cross-functional teams to bring that vision to reality.

She is a self-driven evangelist for UX and voice technology and recognizes the societal and cultural benefits of voice tech evolution. She possesses the insight and knowledge that comes with experience and leadership from her technical, development, and design roles. Her design and research work is presented nationally and internationally to general and field-specific audiences, and she has 18 years of experience designing for mobile, web, desktop, and smart TV interfaces.

Shyamala has more than six years of experience designing voice interfaces for Connected Home Experience, Automotive, and Wearables. Publications: Her work on research into usability, accessibility, speech recognition, multimodal voice user interfaces, and user interfaces for research and educational use cases has been published internationally, and her interviews have been published with several magazines and podcasts, including Forbes, the healthy code, Automotive, and TU-Auto magazines.