Dash Open 05: Makeskill Design Kit, the Open Source Multimodal Rapid Prototyping Suite for Alexa

Ashley Wolf: Hi everyone, and welcome to the Dash Open Podcast. Dash Open is your place for interesting conversations about Open Source and other technologies from our Open Source Program Office. We're home to many leading brands including Yahoo, AOL, Tumblr, TechCrunch, and many more. My name is Ashley and I'm a Technical Program Manager on the Open Source team at Oath. Ashley Wolf: Can you tell us a little bit about yourselves, your roles at Oath, and how you got started with Makeskill? Anna Shainskaya: Hi, I'm Anna, and I'm a product designer. I currently work on Yahoo Mail. Previously, I've been working on several chatbot initiatives. Lauren Tsung: I'm Lauren, I'm also a designer in the Yahoo Mail team. I'm just interested in exploring the new paradigm of voice, and I did chatbots previously too. Ashley Wolf: Oh great. Can you tell us a little bit more about how chatbots are related to voice assistance and Makeskill? Lauren Tsung: What chatbots and voice technology have in common, is that all of them have this element of conversation. You need to understand how to organize conversational design. We saw that there are a lot of learnings that we gathered from designing for chatbots, or just devices that you're maybe chatting or texting with an assistant on, but not necessarily speaking to it. But there are a lot of commonalities between how we can organize conversations between voice and chatbots, so we've learned a lot. Ashley Wolf: That's great. Is that how you got interested in putting together a project called Makeskill? Can you tell us a little bit more about that - specifically what it does? Lauren Tsung: Sure. Use a Google Sheet and make a working voice assistant prototype in seconds. I was really inspired by Google Sheets, I guess. Well I just, I didn't see any sort of products out there that met my need as a designer to really just quickly prototype. I did experiment with noncoding tools, but they weren't really nuanced, and I dug into the Alexa SDK and over time I realized, well, for every new iteration, I have to code this up. Can I just make a tool? What's cool about Google Sheets, I've seen this in the chatbot world, and I'm sure Anna echo that sentiment. How can we basically use a model as a Google Sheet and see how can we, it's a very familiar interface, how can we make voice first, and multimodal experiences with Makeskill? Anna Shainskaya: Yeah. The main pain point is that currently with voice technology, you can't really get started unless you're an engineer. Me being a designer, there is no way that I can start experimenting and playing with different voice ideas. In the previous world when we were working on chatbots, we worked a lot with content engineers. They've been using different tools on how to manage all these different complex conversations. When we started doing Makeskill, we just wanted a tool that would allow everybody to be able to be part of this process from day one. It's content engineers who are responsible for building the voice and the mood, and polishing all these conversations and words. Then it's designers who are responsible for building the right experience. Product managers who guide the direction, and researchers. We need to find the right gaps in the right problems, and the right intents and use cases. Anna Shainskaya: The main idea was to enable collaboration and provide access to all of these extended product teams to be part of this process. Lauren Tsung: So, while I am a product designer, my specialty is actually in technical prototyping, hence I got into the Alexa SDK and I just like nerding out with new technologies all the time. I realized, well, when I'm trying to design for voice, I don't really have a good discernment as to what is too long of a response, what's too short. What's really cool about a Google spreadsheet, you can compare and contrast different copy, right? It's like, what Alexa should say, her tone, et cetera. By utilizing the different columns in the spreadsheet and trying to see how we can do parallel experiments and exporting batches of Alexa prototypes, rather than coding every single one from scratch, every single time. Lauren Tsung: I felt that this could be a new way of prototyping. Anna Shainskaya: Yes, if you think about, today there are no best practices. How do you design for voice? What copy is too long, what's too short? What's a visual card that can meaningfully enhance a voice experience and what's too much information? We didn't really know the answers to this, we all are trying to figure out what's the right formula. Ashley Wolf: It sounds like Makeskill is built by designers for designers in hopes of setting up a best practice. Anna Shainskaya: Yes. So, not just for designers, I would say it's built by designers for larger, non-technical teams and actually for engineers as well, but not limited to engineers. So everybody can participate. Lauren Tsung: And I'd say for the engineers, being an ex-engineer, I think they could appreciate scaffolding a prototype rather than coding from scratch. Of course, when it comes to refactoring, adding their Unit tests, and adding the real API, I'm sure that the onus should be on the developer eventually. The point of Makeskill is to be able to, in an agile fashion, test out ideas very quickly. This is why we believe Makeskill can assist the throwaway sort of mode, where we're brainstorming, figuring out the right use cases, talking to users and understanding the feel of the skill, before it can be evolved by engineers who are getting acclimated with developing the skill. Ashley Wolf: What are you hoping to see happen with Makeskill, now that it's open sourced? Lauren Tsung: Well, I certainly would appreciate all kinds of new ideas, I think I've been really set in a certain frame of mind with prototyping, but I'd love to see, for example, more complex conversations and more modalities. It currently is very Alexa focused, I'm open-minded to other assistants, such as Google Home. I know they're interested in voice first multimodal experiences, too. Ashley Wolf: So, if we have anybody that's listening who is interested in getting more involved in chatbots or voice assistants, do you have any recommendations for them? Lauren Tsung: Yeah. First and foremost, live with a voice assistant. I actually have a smart apartment. I bought a bunch of Internet of Things, from light bulbs to my air filter, to a vacuum cleaner, and a cooking device. And I just wanted to see what utterances or what things made sense in my daily life, with my voice assistant. Lauren Tsung: Of course, learning about chatbots is cool, too. There's really, I mean, I think going through the exercise of designing conversations. Anna Shainskaya: Yeah, exactly. So I think that's really important, to be on the side of the user because today there are a lot of experiences are not fully polished. Being the user of a voice assistant and having it in your home, you can actually spot like, a few improvements. You can say like, "Well, I hear this response. Maybe it could be a little bit shorter" or "Maybe like, this response could be beautifully enhanced with a visual card." I think when you are a user and you're on the receiving end of this experience, I think you can, if you have all these devices at your home, you can start thinking, how can I improve it? How can I make it better and what other use cases can I have here that would work beautifully with voice? Lauren Tsung: I would also say, be more aware of your context, as a user. Because maybe it could be that there's a preference for a longer answer, or it's just a very passive experience where you're letting it play in the background while you're multi-tasking. Or, you're in the moment and you need instructions step by step and you'd really like a very concise answer. So just pay attention to your physical context and how that plays with multi-tasking or not multi-tasking. Anna Shainskaya: And then think about multi-modality. What should be said with voice and what is better to be shown with a card? I think that's starting to pay attention to those experience and just think, "Okay, if I had to answer this question, and I have those two modalities, I have voice and I have a screen, what's the optimal response that I could receive that would actually solve this informational need that I have? I think just on a daily basis, thinking about those experiences and having them in your home and experiencing them, I think is the best way to get started. Ashley Wolf: Awesome. Well, thank you so much Anna and Lauren, it was so great talking to you both today. Anna Shainskaya: Thank you. Lauren Tsung: Thank you. Happy hacking. Ashley Wolf: If you enjoyed this episode of Dash Open and would like to learn more about our Open Source program and other technologies, visit developer.yahoo.com. You can also find us on Twitter at YDN.

More Episodes: