The Rasa Summit 2021 is a wrap!
We had a fantastic three days of sessions on making better contextual assistants with Rasa.
See below for the talks, and to stay up-to-date on Rasa events, products, and the Rasa community, sign up for our newsletter!
Below we'll be listing all the talks from Rasa Summit 2021 in their final edited form, plus links to presentation slides where available.
Wednesday 10th February 16:00 UTC
AI Assistants involve machine learning but that doesn’t mean we should throw out everything we know about shipping good software. Rasa CTO and Co-founder Alan Nichol gave the 2021 Rasa Summit welcome keynote and discussed conversational AI teams and moving fast at scale.
Wednesday 10th February 16:50 UTC
The beauty of a conversational AI is the natural way it communicates, which leads to the challenge of handling infinite spontaneous responses in certain contexts. We will discuss the process of continuous improvement of the Conversational AI in production, especially the parts of analyzing user feedback, data training and the lessons learned.
Wednesday 10th February 17:00 UTC
In this presentation, we want to open up the possibilities of designing chatbots for research purposes. Our journey in AI at inChat, began as what we see as the opposite end of practice and design for most chatbot developers. In doing so we have created a methodological approach that we call the Ethnobot (ethnographic chatbot). Bots designed in said method can be used for UX, Service Design, and Ethnographies. Equally important we have learned some best practices for embedding research infrastructures into services and products. These yield the benefits of having a UX researcher on hand. By sharing our story through key case studies, we offer insights towards designing chatbots as a viable resource for teams with varying levels of expertise in user experience and service design.
Wednesday 10th February 17:25 UTC
In this session, I will discuss how we can leverage Rasa to supercharge the command-line and graphical user interfaces. I will walk you through the design and sample implementation of these user interfaces with Rasa and Conversational AI at the core. By the end of the talk, maybe I will convince you (and myself) to call these "Instructional Dialog Interfaces".
Wednesday 10th February 17:50 UTC
How do you design for an interface where people can say and do whatever they want at any given moment? In this talk, Hans van Dam explains the proven conversation design workflow that is helping brands like Vodafone, JP Morgan, Adidas, Daimler, and many others, create better chatbots and voice assistants. You will learn how AI Trainers, Conversation Designers, and Conversational Copywriters can work together to unify the best of technology, psychology, and language. It doesn't matter if you are creating your first chatbot or are managing multi-lingual deployments at scale. This workflow helps you unlock the potential of conversational AI through design. Hans van Dam is founder and CEO of Conversation Design Institute, world's largest training and certification institute for conversation designers. Hans lectures at multiple universities, is a keynote speaker, podcast host, and regarded as one of the most influential people in voice (Voicebot 2020).
Wednesday 10th February 18:15 UTC
Come and see how to automate your chatbot lifecycle and adapt it to your existing workflows. Let’s learn how to programmatically pull your chatbot conversations to retrain your model, automatically build, test and scan your custom actions containers for code quality, security and lhave a happy deployment (and monitoring) of Rasa to Kubernetes Clusters adopting DevOps practices.
Wednesday 10th February 18:40 UTC
With every new interface comes new considerations, constraints, and opportunities – voice and conversation design are no different. Expedited in many ways by the increase of consumer platform adoption, COVID, and burgeoning business use-cases – consumer expectations and business opportunities are rising faster than ever. In this talk, I will break down the foundations of designing and developing for the new world of voice. I will compare and identify the constraints and contexts in which businesses/professionals are thriving and what challenges are naturally tied with channels built on conversational AI. We'll dive into the unique challenges, constraints, and opportunities tied to the conversation design and voice ecosystem – and hope to outline the possibilities of what voice can unlock in the future.
Wednesday 10th February 19:05 UTC
We present STAR, a schema-guided task-oriented dialog dataset consisting of 127,833 utterances and knowledge base queries across 5,820 task-oriented dialogs in 13 domains that is especially designed to facilitate task and domain transfer learning in task-oriented dialog. Furthermore, we propose a scalable crowd-sourcing paradigm to collect arbitrarily large datasets of the same quality as STAR. Moreover, we introduce novel schema-guided dialog models that use an explicit description of the task(s) to generalize from known to unknown tasks. We demonstrate the effectiveness of these models, particularly for zero-shot generalization across tasks and domains.
Thursday 11th February 16:00 UTC
Many enterprises have run dozens of proof of concept for text- and voice-based use cases in conversational AI - from internal to customer-facing use cases. What did they learn? How do they think about choosing the right platform for their enterprise needs? What team and processes do they put in place to fully operationalize conversational AI?
Thursday 11th February 16:40 UTC
At Alpaca, we have built a Facebook Messenger-based virtual assistant to help people find a home. Users can search through thousands of listings and request viewings in the same channel they use every day to talk with their friends. By taking advantage of Rasa's open-source framework, we are able to fully utilize all the advanced features of the Messenger platform and to fully integrate them into the user's conversational flows. For instance, users can see apartment previews directly in a carousel within the conversation, edit their preferences through webviews, and many more! The introduction of the Rule Policy in Rasa 2.0 greatly facilitates these interactions, opening the door to even better user experiences.
Thursday 11th February 16:50 UTC
Testing is a crucial enabler for the success of chatbots and virtual assistants. Doing it manually requires enormous time and efforts. As DevOps and furthermore AIOps grow in importance, automated testing will remain critical to ensure that bots actually do what their designers intend. Unlike traditional software where the application follows a predefined flow, a chatbot runs without any restrictions. Talking to a bot has no barriers. Combining this with an unpredictable user behavior, it becomes utmost difficult to verify the correctness of conversational AI. Training data and test sets are infinitely large. In fact, quantity plays a major role in quality assurance for bots, but makes it impossible to test manually. The main questions to answer are "Why are bots failing?", "What and how should you test?" and of course "How to automate?". We will showcase the setup of a test automation pipeline for a Rasa based chatbot to continuously check conversation flows and NLP performance. And we will take it even further by adding full End-to-End testing from API over Web & Mobile to Voice.
Thursday 11th February 17:15 UTC
In this talk, we will present our approach to building a platform that supports the fast delivery of chatbots and AI assistants without compromising on performance and customer experience. We will discuss the main components of the solution and the inherent challenges, from both the architecture and AI science sides.
Thursday 11th February 17:40 UTC
Human-to-human electronic communication has moved from text (email) to voice (VoIP) to augmented video (Zoom/Skype). Similarly, the medium for human-to-machine conversation has moved from text (chatbots) to voice, with voice-enabled chatbots in wide use today. The next step in this evolution is a video-enabled conversational experience. Each medium change brings its own technical challenges. Creating a good voice experience involves more than just hooking up a chatbot to a text-to-speech and speech-to-text service. Vocinity has developed a platform for voice-enabled chatbots that has been in production for almost 2 years. We're updating our platform to support a multimedia experience where the bot communicates via video, voice and text messages and images. Using Rasa to provide the conversational logic for the immersive multimedia bot enables us to meet the challenges in voice/video communication. Rasa’s power and flexibility enabled us to extend it to support voice and video.
Thursday 11th February 18:05 UTC
Thursday 11th February 18:30 UTC
Clearly understanding your users’ intent, sentiment and background information is critical in helping them achieve what they want. Hence, we built our platform Iris around Conversation Driven Development even before we knew what it was. The challenge now is the need for a huge collection of data to get close intent matches. In fact, you have to design, build and test your chatbot with INTENT CLASHES in mind. We’ve used Rasa to build SELECTIVE CONFIDENCE, addressing intent clashes. If we know that the user is going through a sales pathway, we won’t match them with content from, let’s say, customer support. Or if the user says something irrelevant, we can still engage them with the identified flow without losing the conversation. This enables you to build a much richer user experience. Global brands such as TWITTER and FLIGHT CENTRE are taking advantage of this. We’d be happy to share these creative and innovative projects using Rasa.
Thursday 11th February 18:55 UTC
New algorithms may get the press, but the real heart of any AI project is data collection and curation. This talk will show you why getting to know your data is so important and provide best practices for improving your data curation and annotation.
Friday 12th February 16:25 UTC
QA has always been under-rated and thus it is important to consider this equally important as the Dev. If we look at the Chatbot QA, it had been considered as a highly challenging work specially when you do not know where your bot may break while you sequentially will be only running your flow (stories). Most of the companies / tools only check the flow which are coded in a fixed format which often breaks while testing. There may be cases where bot are migrated to new version and it breaks. The presentation will discuss the possibilities to test the bots by helping folks to create their coverage matrix for your stories, efficiently looking at the logs and mine information and most importantly what to test and which components to test.
Friday 12th February 16:35 UTC
According to Home Office data, the proportion of crimes solved by police in England and Wales has fallen to the lowest level recorded. Crime investigations are inherently expensive and each case takes a lot of man-hours and resource-poor cities are less able to reduce investigators' caseloads. With the growing amount of information held in siloed systems and people’s minds, it’s becoming increasingly difficult to grasp all the links and see the relations. Data is overwhelming. What if by using latest technology such as AI and NLP we could reduce time in solving crimes and the response times at the scene? What if more lives can be solved because AI can help automatically make a link between information faster than any human could do? What if that same technology could help us become more efficient and satisfied employees while decreasing costs for the companies? In this talk I would share the specific formula governments companies can adopt to successfully employ AI while keeping humans in the loop making them more efficient, productive and happy.
Friday 12th February 17:00 UTC
Future of Digital Experience would be driven by the intersection of Content & Context, where Context would take the lead. Search is changing, and so is the way consumers choose to engage with businesses locally or globally. There is a distinct move away from screens and keyboards, and into voice-based interactions. Voice search is becoming a fast-growing habit across consumer segments and fundamentally transforming how people and businesses transact on the internet. Consider this: - -> In 2020, there will be 4.2 billion digital voice assistants being used in devices around the world. Forecasts suggest that by 2024, the number of digital voice assistants will reach 8.4 billion units – a number higher than the world’s population. - -> Sales revenue from wearable devices is projected to grow from around 16 billion U.S. Dollars in 2016 to around 73 billion U.S. dollars by 2022. - -> McDonalds is adding 1000 kiosks per quarter for self-ordering and checkout since 2018 and acquired Appente to boost voice tech on these devices. In the talk which includes hands-on demo, we will talk about how the current CMS ecosystem is structured and how the new-age headless CMS is changing how we create content. We will also look at the Schema.org usefulness. Here are some key takeaways: - Why a voice content strategy is critical for enterprises - How and Why to make your content future proof - The differences between voice-based and web-based content, and how that affects the user experience - The basics of optimizing your content for voice search - Why bots should be your next strategic investment Demos: - A quick view of the schemas important for the VSO - Example of sites ranking in the Voice
Friday 12th February 17:25 UTC
Not all beginner coders know what is a FOSS software nor understand how most FOSS communities work or interact. Onboarding newcomers is challenging to most communities. Underrepresented groups and non english speakers face additional difficulties to contribute to FOSS, such as language barrier and confidence gap. To overcome these barriers, BOSS (Big Open Source Sister) was created. The program has three major objectives: introduce newcomers to FOSS, to capacitate the under-represented in the coding community into technologies demanded by the software industry, and improve participants’ belief in their own competence. To do that, BOSS based their mentorship in chatbot development, using a Rasa boilerplate project, created by LAPPIS. This program won the Gnome Engagement Challenge second phase (5 finalists).
Friday 12th February 17:50 UTC
You know the feeling when you ask for something and you’re pretty sure “no” will be the answer, but you still do it, because why not try? Well… the story of end-to-end is exactly this! Before starting on it, we read several papers about the technology not being ready for end-to-end dialogues in production. So, when we started working on it as a research project, “negative results are also interesting results” was our mantra. Suddenly, the results started to look more and more promising. Then, we developed the end-to-end training further – so that one can combine the classic Rasa format with intents and actions with the new end-to-end and gradually get rid of intents they don’t need. In short, I will tell you a story of how end-to-end grew from a little internship project into an experimental feature of Rasa (and spanned far beyond the internship).
Friday 12th February 18:15 UTC
Ever since launch, four years ago, the Autodesk Virtual Agent has been one of the cornerstones of our support strategy. However, as customer demands have evolved and internal stakeholder interest has grown, we have seen our original scope expand. After the initial novelty of the solution wears off, the continued adoption of a conversational AI solution depends on continuous improvement. I will share our approach to building a platform that prioritizes learning from customer input and supports personalized problem solving. I will discuss the evolution of our architecture over the last year, the technical decisions we made and the creation of a platform powered by natural language processing.
Friday 12th February 18:40 UTC
As a machine learning engineer specializing in natural language processing, I used to scoff at out-of-the-box NLU solutions, extolling the virutes of using hand-crafted models exclusively. In just a year, the AI @ T-Mobile team has gone from creating models exclusively in keras or tensorflow with no supports to fully embracing Rasa - and not just for it's chatbot functionality. In this talk, I will go over how we've used Rasa to stand up a beefy customer service chatbot in a company that highly prioritizes human-to-human interactions - as well as the surprise lift our data science team has found from leveraging Rasa models outside of chatbot use cases.
Friday 12th February 19:05 UTC
Launching conversational AI in the enterprise depends on many things that are essential for successful adoption and delivering value. This talk will walk through why enterprises need conversational teams, real user data and insights, and a unique approach to launch assistants that can handle mission-critical tasks.
