We are always in touch
Ukraine: +380674822702
email@singularika.com

4 days with chatbots: Chatbot Days 2017

One day the author of this article was improving skills at the copywriting workshop in the welcoming IQSpace. Everyone should present their company, as a practice. Surely, the company’s specialization was to be mentioned.
– My company makes chatbots.
– Makes what?
Well, there was one person who really understood what are the bots and another one who had heard something about them.

Six months later, Singularika in cooperation with IQSpace educational institution organized Chatbot Days 2017. This seminar included lectures and a workshop and attracted developers, managers, marketers and other specialists interested in the trendy technology.
For the first three days, Singularika representatives made presentations and answered the questions. To make it more convenient for the participants we divided the program into sections:
December 5
For project managers, startupers, marketers, business analysts.
December 6
For developers, QA-engineers, project managers.
December 7
Natural Language Processing Technologies. For developers, mathematicians, NLP-engineers.
December 8
There was a workshop where participants have built chatbots themselves using a platform for developing without coding.

It was great that some people attended all four days!

“We have a business and we have been thinking about creating a bot, but couldn’t make out how to start, what and how to do… Now we see what it is, what it can do, we can draw the chat flow on our own to explain to developers what we need, and even, for a beginning, create our own bot on Facebook Messenger”.
“I am a university professor and I can say that as an introduction it is really good. And that was a good idea to make it several days with different topics for the different audience. But surely I would like to go further, to code a bot. It was like we’ve just come to the most interesting thing – and then the seminar was just over! It’s clear, that not everyone wants and needs the practice that includes coding. But it really needs a continuation.”
“I am a programming engineer but didn’t know anything about chatbots before. Now I know the basics, and I want to code my own bot, and I am going to move forward on my own. And by the way, at the workshop I was not doing exactly the same as others, I just explored that platform and began constructing the bot for my page with other features.”
“I love such events that charge you with energy and the feeling that you know nothing except some crumbs of information. These events are worth the time spent.
The third day of #chatbotdays kind of opened our minds with linguistic analysis and neural network. Nobody managed to provide me with so much information in two hours. There is so much energy after this seminar as if I was sleeping all my life and now I’m awake. I want to learn, develop and create.”
Many participants told that they would like to attend other such events. Some of them even hurried us to conduct them as soon as possible: give us more, more practice!

The first day: overview, Technical Specification, the future

Singularika VP of Business development Alexey Ilchishin presented an overview, provided the definition for the term chatbot, bots classifications and description of key messaging apps. The speaker explained how to choose the messaging platform depending on the target audience and bot tasks. At the coffee break, Alexey was surrounded by intrigued participants asking: What is the difference between the selling bots and the lead generating ones? Have you done the projects for WeChat? What would be better for a certain company, Viber or Telegram?

Singularika PM Alex Petlya revealed how to write Technical Specification for a chatbot. He also touched upon the topics which are important not only for Technical Specification writing. For example, to create specification, you have to know what the customer wants in the admin panel. And here Alex explained what functions and modules can be there. Also, he described the services with which a bot can be integrated: NLP services, analytics, etc. It is important to ask the client about them in advance.

CEO Vladimir Koval spoke about using bots in lead generation, marketing, and as automated operators. The benefit of chatbot is that as soon as customers start a conversation with a bot, we already have their contact information and a possibility to interact with them, whereas on the website people often leave without giving their contacts. Can we spam people through bot? No (except Viber). But we have the chance to reach them and attract them. We shouldn’t repeat the functionality of our app or the website, it’s better to automate just some of the functions or create additional features in the bot. In general, the speaker outlined the future of chatbots: you can’t expect a miracle of this technology but this is an effective tool for solving many problems and soon a bot will be a “must have” for any business.

The second day: UX design and details on development for different platforms

UX/UI designer Valeriia Maliarenko told about interface elements in key messaging apps, their limitations, and differences. Also, she provided main principles of chatbot UX design, real-life examples and shared company’s best practices. The questions showed the audience was interested in the first place in such things as bot personality development, technical aspects and tools for chat flow composition, and the problems of creation maximum user-friendly design.

Vadim Sharafin, Singularika leading developer, gave a presentation on the development of chatbots for Facebook Messenger, Telegram, Skype, Viber. He explained in details how to register chatbot in messaging apps, how to integrate it with third-party services, user data storage and assign statuses to users. The most relevant seemed to be the information on opportunities of the opt-in messaging. The lecture went on in an interactive format, very close to the dialogue with the audience.

The third day (mind-blowing): NLP, Machine Learning

The first presentation, by Singularika NLP-developer Roman Briazgalov, was on an NLP service called Dialogflow. It made us, as the organizing party, consider in future adding 10-15 minutes break to such lectures for the attendees to test the service and ask more of their questions. We could notice that as soon as Roman had finished presenting and began answering, guys opened Dialogflow on their laptops and one of them even on a phone! The flood of questions was out of control and even during several minutes of the next presentation people were still studying the service.

With the neurolinguist Inga Velkova, the listeners plunged into the complex topic — “How to
obtain vector representations with various NLP (Natural Language Processing) methods.” The seminar participants learned about distributional semantics models, their advantages and drawbacks. Inga illustrated the concepts by the example of Singularika project, the Q&A system for internet providers.

Philip Marchenko, neural network and machine learning specialist, told about text classification, provided examples and schemes. He demonstrated, that, considering machine learning, you cannot point out “the best model ever”, the choice depends on the specific tasks. For small data amount, traditional approach is more effective than sensational neural networks.
When the presentation was over, the speaker asked “Are there any questions?”, and there was a silence for a minute or so. Then one of the participants said “Bang!” and gestured toward his head imitating brain explosion. And then the questions showered, mostly applied ones: how to choose a model, what’s the difference between a sigmoid and non-sigmoid, how much is “enough” data for a neural network (it turned out to be no less than 600 000 samples). There was only one case, when Phil found it difficult to give a single-valued answer, and the question was “Why is Alexa so stupid?”. Considering the secrecy of the development process, our specialist could only venture scientifically sound suggestions.

The fourth day: create a bot in two hours

This day was the most tense. Under the sensitive guidance of Vadim Sharafin and with the assistance of the second leading developer of Singularika Vlad Dariev the workshop participants created their own chatbots for Facebook Messenger on Chatfuel platform. As there was not so much time, all the attendees created the bot following one scheme (chat flow) prepared in advance. There was a main menu with five items: Resume, Portfolio, Social Media, Weather and Exchange rates. These items were also presented in the persistent menu. When you press Resume button the chatbot returns series of messages about its owner (each participant wrote whatever he or she wanted), after each message you can go back to the menu or continue reading. In Portfolio section, we placed a slider with images and links to the website. In the Social Media, obviously, there were links to the profiles. It was not that easy with Weather and Exchange Rate sections. The moderators wanted the participants to add JSON API module to send the user weather forecast after getting a location. The part with location feature which can be easily added through Chatfuel went well. But when Vadim started explaining how to deal with a JSON and showed corresponding Facebook Messenger documentation, non-programmers became uneasy. So we had to put aside the Weather section and work with the Exchange rates menu item, where bot asks users which currency are they interested in, records the data as an attribute and allows sending targeted messages (creating filters). In the meanwhile, Vadim found a solution to the coding problem: he prepared and gave everyone access to the link returning code in the format required by Chatfuel, so each participant could easily complete the task and add weather forecast feature to the bot.
Voilà – the chatbots were tasted, most of them were working properly, and moderators quickly helped to fix those which was not operating correctly. While constructing their bots, the participants learned a lot of useful things about both Chatfuel and Facebook Messenger, for example, that Buy button works only if you have a PayPal or Stripe account, and that you can switch off or allow user text input for a bot, but you can’t do it for a separate block.

There was no such thing as official closing of the seminar, no closing ceremony. The event ending time was 9 p.m. but almost everybody stayed after that to ask additional questions and polish their demo chatbots, someone just for several minutes, someone almost for an hour.
The participants were expressing their gratitude, saying good bye and leaving quietly not to disturb their colleagues. The last ones went away together with moderators about 10 p.m. (we wonder, did participants dream of bots conquering the world after that?). We guess, it was a sign and the closing ceremony just wasn’t necessary because we will meet again soon at Singularika workshops and lectures.
We are grateful to IQ Space for the comfort and the great organization, and for the tasty cookies in the coffee breaks. And most of all we are grateful to the seminar participants, who are interested in new technologies, who are ready to learn and explore the world of chatbots!

Contact us!

We respond within 2 hours