Database Reference
In-Depth Information
Gutierrez:
How would you describe your work to someone who is not
familiar with it but is familiar with data science?
Heineike:
From a data science perspective, the way I explain it is that what
we're doing is Natural Language Processing [NLP] plus machine learning plus
network science plus data visualization—which is a really weird combination
of things to combine, actually. I don't think there are many who would include
network visualization with NLP. So there's kind of a uniquely mixed stack. And
for technical people, there are a lot of different bits that they might engage
with from their experience.
Gutierrez:
Why is putting these techniques together powerful?
Heineike:
A lot of work in the information retrieval and search space focuses
on the problem of highlighting the one document that you should read.
Similarly, if you look within NLP, there's a lot of work around how to extract
entities or topics from a block of content. But again, the output is often simply
a list of things, hopefully, sorted by relevance. The focus is on finding the single
best thing. We differ in that Quid is tackling the question of “can we expose all
of what's in the data and make it understandable so that anyone can interact
with?” So we switched to finding unique signals that exist in the data and then
figuring out how those indicators all interact with each other—giving a totally
different level of perspective on markets, consumers, innovation, etc. So the
questions people answer using Quid are actually posed slightly differently than
the normal—it's not about a single signal, it's about a holistic and nuanced
understanding. This is why we end up doing the network visualizations of
topics, which isn't normal.
Gutierrez:
How do you describe your work to someone who has very little
knowledge of mathematics?
Heineike:
There are a lot of people in the research, investment, and strategy
analysis space with challenging and complex questions about what's happening
or likely to happen in the world. This set of questions encompasses questions
like: What do people think about my business? What is the next technology
bubble? What is a major market player's long term IP strategy? What do
people mean when they say terms like
happiness
, family, love? The way
people have historically tried to answer these questions was that they hired
consultants, some interns, they sat down themselves and googled things, or
they had this epically long RSS or Twitter feed that they read and read and
read, so that they could eventually summarize their findings into maybe a
spreadsheet or a PowerPoint slide. This is a fundamental but inefficient pro-
cess done for a huge number of big decisions that are made by organizations
all over the world. That's just how the system works.
