Another month, another launch of a highly requested product feature! In this month’s product update, we’re pleased to bring you GST/VAT Tax Extraction in Australia, New Zealand, and the United Kingdom!
In March of 2018, we announced our latest version of machine learning-powered data extraction. It brought better, faster data extraction (less than five seconds!) on ~25% of uploaded documents, with the promise of increasing its coverage of documents and extracted data fields (GST/VAT Tax extraction, as an example) as it learned over time. I’m excited to report that roughly 50% of your uploaded bills, receipts, and invoices are extracted using the machine learning model we launched less than one year ago today!
In addition to automatically extracting:
- Supplier name
- Total amount
It now extracts (wherever possible):
- GST/VAT in Australia, New Zealand, and the United Kingdom – learn how to enable the feature here
- Due Date
- Invoice Number
We know how important accurate, fast data extraction is for your bookkeeping workflow, which is why we invest a lot of resources and time into increasing our coverage and improving our accuracy. Our goal is to help you stop chasing financial documents and doing manual data entry. The launch of GST/VAT tax extraction gets us one step closer to that goal!
We look forward to keeping you posted on our progress. In the meantime, we thought you might want to meet the humans behind all this amazing work!
The humans behind the robots
Hubdoc’s Machine Learning (ML) team works tirelessly to improve our product’s data extraction features. We decided to ask the team a few questions about what they do in and outside of Hubdoc, what excites them about machine learning, and more.
Without further ado, let’s meet the team!
Lucian Mustatea, Chief Technology Officer & ML Team Lead
What does the ML team aim to achieve for Hubdoc partners?
“We are aiming to accurately and quickly extract financial information from documents uploaded by our users. This should enable our partners to spend more time on higher-value tasks which ought to result in higher profitability and happier clients.”
Dina Jankovic, Data Analyst
Can you define “machine learning” in simple terms?
“Machine learning is putting effort into making the machine intelligent enough to learn by itself, when it is humanly impossible. Once the algorithms are designed and their maximum potential is achieved, humans are no longer expected to perform those tasks themselves, so they can move on to a new, more complex task.”
Jerome Gleyzes, Machine Learning Developer
What excites you most about machine learning?
“What excites me most about machine learning is how it can free us from many boring, repetitive tasks. This way we can focus on more rewarding activities, such as building relationships or coming up with new ideas to solve problems.”
Salim Fakhouri, Machine Learning Engineer
What do you enjoy doing outside of Hubdoc?
“I go to the gym, play soccer with friends every now and then, and I also work on my AI fashion research. I am working on building an AI model that is capable of generating compatible, good looking outfits using your own clothes in the closet. It is basically answering the question: ‘What do I wear today?’. It can also work as a shopping assistant – when you take a picture of something you want to buy, it can generate compatible outfits including this item, so you know exactly what you can wear it with.”
Bill Wu, Data Scientist
How does machine learning most benefit you in everyday life?
“It enables better access to information in the world via better results from search engines that incorporate machine learning.”
Mohamed Khodeir, Machine Learning Engineer
What does the machine learning team aim to achieve for Hubdoc partners?
“More productivity, less work. We want to offload all of the friction/repetition from daily tasks and leave Hubdoc partners with more time to scale their practice or add strategic value to their clients' businesses.”
Robby Jennings, Product Manager
What excites you most about the power of machine learning?
“I get excited by machine learning for two reasons: its potential to distill complexity into simplicity, and because it feels only limited by the imagination of those applying it.”
About the AuthorFollow on Twitter More Content by Geoffrey Gualano