Here I want to share with you the nuggets of knowledge from Berlin’s datanatives conference which is probably THE data-driven conference in Germany’s capital. The topics vary but are always focused on data so that you will find lots of interesting talks about your data topic of choice.
Those are just my personal nuggets of course because I couldn’t attend all the talks and sessions.
Big Data is dead - Data Thinking is on!
The current Big Data approach to monetize data is flawed and non-efficient.
Big Data approach:
Pump all the data you can get into your warehouse (or lake or whatever) and try to make money from it by somehow getting value out of the data mountains. This approach doesn’t work most of the time.
Now try this Data-Thinking approach:
In an iterative approach, figure out what the most valuable use cases are that you could tackle. Then target the data that you really need and solve truly existing problems with your algorithms and analytics.
On starting your own data science consultancy
You were an engineer? Now you are an entrepreneur.
If you love to be an engineer, you might have to step back from this calling of yours because you most probably will turn into a full-time entrepreneur
Define your niche
Unless you want to fight for small margins against all the other legion consultancies.
Love the founders team
You will sometimes spend more time together with your co-founders than with your family.
Are you planning to grow or do you want to stay a small boutique consultancy?
Whole set of different rules applies here. For example with hiring. If you aim for rapid growth you won’t have the time to hire best experts in your field.
We live in the modern era of text analytics!
Wild West 2000 - 2010
Lots of counting of words basically.
(2010 - 2015)
Mostly, this era was about figuring out if people like or dislike things.
The Modern Era (2015 - 2018)
Advertising experience is becoming much better, analytics lead to way better targeting.
It was exciting to see how fast the technologies and possibilities in NLP are moving.
I also recently skimmed through a paper from 2008 that described some “eras” from 1940s to 2008. The time from 2000 to 2008 was called “The Rise of Machine Learning”. Similarly optimistic as Jay Krall was in his talk about today’s advancements. Probably in a few years, 2000 - 2020 will be known as the dark age of text analytics. But let’s enjoy Today of course :)
A lot has changed since 1992! Learn modern SQL!
SQL is more hip than you think! It can handle JSON and multidimensional arrays(soon). SQL can handle non-relational models. (the only bummer being that SQL distributions are not adapting fast to all the cool features that the SQL standard already offers)
For further investigations on this revolutionary topic, check this:
These summaries are not extremely detailed. Just let me know in the comments if you need more information on any of the topics.