Page 24 - Realcomm13-Final-LR

BUSINESS SOLUTIONS

AI Meets CRE: The Power of Deep Learning

Emilio Matthaei
CEO
LEVERTON

ven though deep learning, a branch of artificial intelli- detection and drug discovery research. Nevertheless, with the
gence, is still in the early stages of its evolution, it is a pow- rise of deep learning technologies, complex algorithms are
E erful technology expected to disrupt global industries. The applied to more and more areas. There are in fact various ways
market for deep learning technology is estimated to exceed over in which this technology creates efficiency in the commercial &
$100 billion in annual revenue over the next several years and corporate real estate industry.
the interest in this technology, which
is based on the idea of programming Increased transparency and higher
networks to imitate functions of neu- data quality
rons in the human brain, is drastically Global corporations with offices
increasing. around the world must keep
Amongst many use cases, deep track of thousands of real estate
learning technology can be used to documents containing legally
train computational models to identify binding information—across various
and understand text. Once fed with jurisdictions and in multiple languages.
information, the model begins to When negotiating lease terms,
interpret text and extract verbiage in companies face the challenge of being
the form of text, numerical values and dependent solely on local resources.
other complex fields. The more data a Once finalized, only a speaker of the
system is fed with and the more train- native language can understand the
ing the model is given, the more negotiated terms. Deep
it learns. The learning happens “With deep learning technologies, tedious learning technology can
through repeated exposure to data aggregation exercises can be accelerated be trained to ‘understand’
the item the model is supposed significantly. The technology allows for any language, enabling it
to learn about. Rather than to extract key information
writing rules for how a system automated information extraction from the from all global contracts
should learn to recognize words, original document and thereby delivers more and convert it into a target
e.g. how a letter looks and which accurate and structured data. Reporting and language, regardless of the
letters a word is composed of, forecasting suddenly become significantly easier original language type. Thus,
the system learns by recog- the application of intelligent
nition. It gets exposed to one and more accurate.” technologies allows you to
term over and over again until it understand the contents
learns to make sense out of that term. Researchers have com- of your international portfolio in any language. By overcoming
pared this learning process to that of a child learning to commu- these language barriers, a global portfolio overview and a new
nicate for the first time. level of transparency can be achieved.
Another challenge global corporations are facing is the difficul-
Why is deep learning important for the real estate industry? ty of reporting and forecasting. In large enterprises, accurate
Deep learning will—and has already begun to—change reports are crucial, yet, they are time consuming to compile as
industries. For most people, the industries that come to mind they are based on unstructured data, hidden amongst thou-
first when thinking about deep learning are probably finance sands of documents. With deep learning technologies, tedious
and health care. We have indeed seen a number of ways data aggregation exercises can be accelerated significantly. The
in which it is applied in those sectors, for example in fraud technology allows for automated information extraction from

22 Realcomm