Sitemap

EuroPython 2019

4 min readDec 30, 2022

--

[Originally published internally at work in July, 2019.]

Last week, I ran a workshop at EuroPython with Cheuk Ting Ho on building a Natural Language Processing (NLP) powered chatbot. I have mentored before at codebar and Code Your Future; however, this was my first workshop delivered at a conference. It was also my first EuroPython. Both of these turned out to be wonderful experiences.

Here are some photos of Cheuk and I helping attendees during our workshop:

Press enter or click to view image in full size
Press enter or click to view image in full size

🤖 NLP powered chatbot workshop 🎉

Through hands-on coding exercises, workshop participants learned how to build a chatbot with Rasa, an open source machine learning framework for automated conversations. Rasa uses NLP to understand messages and hold conversations with users. The goal of the workshop was to build a small chatbot demonstrating the role of machine learning (ML) in assessing the sentiment of a user’s input and in generating the chatbot’s response.

Rasa has two main components, both of which rely on machine learning and the workshop was structured so that participants would be able to see how they connect and work together to form a responsive chatbot. The first main component of Rasa consists of natural language understanding (NLU), a subset of NLP. Rasa’s NLU extracts information from the user’s message to predict the meaning of a user’s message. For our workshop, we used a pre-trained spaCy model in our NLU pipeline. We fine-tuned our spaCy model on example data to create NLU capabilities that were appropriate for the chatbot we were building.

The second main component of Rasa is its dialogue engine, or Core. Rasa Core uses a machine learning model trained on example conversations to decide how to respond to a user. This enables more complexity in the conversation. We wrote our own example conversations for the chatbot, including custom actions for capturing specific user data.

Get Kara de la Marck’s stories in your inbox

Join Medium for free to get updates from this writer.

In the third part of the workshop we used Natural Language Toolkit (NLTK), a suite of libraries for natural language processing, to analyze the sentiment of a user’s feedback to an event — such as EuroPython! — to determine if it was positive or negative and we displayed the results in a rudimentary dashboard using the lightweight web framework CherryPy.

Additional notes on the conference

🐍 First EuroPython! 🇪🇺

EuroPython was fantastic! As its name implies it was a highly technical Python conference, including a PyData track with loads of talks on machine learning and artificial intelligence. For the purpose of this summary, I will focus on my impressions of how the Python community is grappling with CI/CD for cloud native applications and also community initiatives at the conference.

Continuous Integration / Continuous Delivery (CI/CD) at EuroPython

A number of talks entirely or partially focused on how enterprises were wiring up their CI/CD process. It was clear that people are using a number of different tools to put together their CI/CD process and that there is still a lot of hard work being done to discover and glue together these processes. Each enterprise seemed to be inventing and reinventing the wheel: best practices are not standardized and many people are doing a lot of discovery and fine tuning. Overall, in these Python centric talks, there was much sharing of own best practices for CI/CD and a ton of interest from the audience as demonstrated by high attendance and many audience questions.

The Continuous Delivery Foundation (CDF) has a fantastic role to play in creating an open space for people to share best practices and helping to condense best practices into a set of recommendations. This sharing of best practices happened at EuroPython for the Python community, but in an informal way. People are left reinventing the wheel on how to do CI/CD in an enterprise grade, production grade way. And they are not comfortable doing this on their own: they are actively looking at what others are doing, comparing notes, and looking for best practices.

Community at EuroPython

EuroPython is a conference governed by the community for the community. My first glimpse of the inclusive and rollicking EuroPython community was on the main, incredibly busy communication channel for the conference. Of course, people had side channels, but a lot of information was shared in the main channel, including many types of social events outside of the conference. Due to the endless sharing of plans and social events in and around the conference, the week felt like one big rolling, friendly gathering.

More formally, there were lots of planned ways for the community to contribute towards the conference and share what they were working on or passionate about. There was an open track, poster sessions, and lightning talks. These sessions were motivated by the realization that no matter how good the speakers or selection committee for a conference, no format will be perfect. Open spaces and lightning talks let attendees choose the subjects that matter most to them. Attendees are given space and invited to share with each other: their passions, their concerns, their experiences. These openings in a conference format are a fantastic way to give trust and space to the community and let people step forward to share.

--

--

Kara de la Marck
Kara de la Marck

Written by Kara de la Marck

FOSS, Python, JavaScript | Director @codebar