EVENT - Tuesday
Location50 Finsbury Square, London, EC2A 1HD
When6:00 PM - 9:00 PM
DSF Meetup with Busuu
Hi team, in light of whats happening with Coronavirus this event has been postponed. We will keep you up to date when we have more details. We have place-marked a date in May but will keep you in the loop once the day has been confirmed based on the next few weeks. Thank you for your continued support at this time!
Due to the popularity of Data Science Festival events, we are now allocating event tickets via a random ballot. Registering here enters you into the ticket ballot for the Data Science Festival Event with Busuu.
If you get an allocated Universe ticket, please bring a copy of your paper ticket or your ticket on your phone to the event to check in with your QR code. Tickets are non-transferable.
PLEASE NOTE REGISTERING ON MEETUP DOES NOT GUARANTEE YOU ENTRY TO THIS EVENT.
Please click here to apply for a ticket: GET TICKETS
6:00: Doors open
6:30: Tom Richardson
7:00: Oscar Knagg
9:00: Leave building
Tom Richardson – A machine learning approach to crediting marketing campaigns with multi-touch attribution modelling.
Summary: Why was today such a strong day for revenue – was it CRM or an increase in facebook spend? Often the answer depends on who you ask because in reality the revenue from a user that installs from a facebook ad and clicks a 50% promotion email should credit both campaigns. In this talk we discuss an ML solution to deciding how much credit to assign to each marketing campaign in a multi-touch chain based on historical campaign performance.
Bio: Tom has a PhD in physics and moved to data science 6 years ago. He is now the data lead at Busuu heading a team of data scientists and engineers. Since joining Busuu he’s applied his experience to bring machine learning to Busuu; spearheading projects such as Busuu’s Smart Review feature that predicts which vocabulary a user should practice next.
Oscar Knagg – Phoneme level speech recognition
Summary: Modern ML-powered speech recognition systems are highly performant but are trained to be insensitive to subtle pronunciation differences. A British English speaker would omit the “r” sound in the word “market” while an American English speaker would retain it but typical speech recognition systems would try to map both of these inputs to the same output. In this talk I will share how Busuu built a phoneme level speech recognition model in-house overcoming hurdles such as the paucity of training data compared to regular speech-to-text (5 vs 1000 hours).
Bio: Oscar completed his Masters in Natural Sciences 3 years ago and dived straight into the data side of technology. Starting out training phishing website detection models for an internet security startup he now is part of Busuu’s data team as a data engineer where he works with Busuu’s batch and realtime data pipelines.