Page 162 - Leaders in Legal Business - PDF - Final 2018
P. 162
algorithm, or neural nets prediction is a common output. For example, given a few dozen
instances of a type of lawsuit, any of those machine-learning algorithms could predict the likely
cost of a new matter once sufficient information is available and tell you how probable that cost
would be.
As another example of machine learning, a regression model might explain and forecast
how fees and hours devoted to five common litigation tasks are associated with outcomes and
therefore can predict the likely outcomes for the next case that can identify the corresponding
data. Moreover, the machine-learning software can tell you which of the five tasks underpin the
strongest association with the outcomes as well as how confident you can be that your prediction
is correct.
Extracting insights from text.
Words in documents can be handled statistically by software as text mining. When a
survey returns free-text comments, for example, software can pick out not only which terms are
used most frequently, but also assess the sentiment (the positive or negative vibes of the
comments). Even more powerful are the algorithms that can assemble words from the survey
comments into topics. A person has to examine the words and identify the actual topic, but the
laborious work of parsing all the documents and doing the math can be done quickly by the
computer. If you want to show off, mention latent Dirichlet allocation (LDA) as your topic-
modeling algorithm of choice!
A second form of machine learning would be at work when text mining software takes
thousands of emails, identifies patterns in words such as repetition or proximity to each other, or
pores through email messages to tag possible indicators of insider trading.
Classification and clustering.
Whenever a law firm or law department has collected a set of data, it can use a range of
software tools to cluster the observations. This means that the software brings together related
clients, matters, or law firms based on the information available to it about them. Once the
software has clustered the observations, managers can more easily detect patterns and understand
similarities and differences. A chart known as a dendrogram can depict the clustering of data and
how clusters relate to each other. Somewhat similarly, software can classify observations into
similar groups. Both of these types of analytics help partners see patterns that they could not
otherwise detect from a massive set of data.
Other models can also classify new observations into the most appropriately fitting
group. With several types of algorithms, including K-Nearest Neighbor or Support Vector
Machines, you can classify clients or other data. You would be able, for example, to identify
publicly-traded clients or clients likely to reach a certain realization level.
Other varieties of machine-learning software do not require labels. Their models cluster
the data into groupings that will reveal something. For example, they might cluster a firm’s
clients by profitability. The K-Means algorithm can do this, and with the Principal Components
Analysis you can aggregate “variables” to find out which of them is more influential.
148
instances of a type of lawsuit, any of those machine-learning algorithms could predict the likely
cost of a new matter once sufficient information is available and tell you how probable that cost
would be.
As another example of machine learning, a regression model might explain and forecast
how fees and hours devoted to five common litigation tasks are associated with outcomes and
therefore can predict the likely outcomes for the next case that can identify the corresponding
data. Moreover, the machine-learning software can tell you which of the five tasks underpin the
strongest association with the outcomes as well as how confident you can be that your prediction
is correct.
Extracting insights from text.
Words in documents can be handled statistically by software as text mining. When a
survey returns free-text comments, for example, software can pick out not only which terms are
used most frequently, but also assess the sentiment (the positive or negative vibes of the
comments). Even more powerful are the algorithms that can assemble words from the survey
comments into topics. A person has to examine the words and identify the actual topic, but the
laborious work of parsing all the documents and doing the math can be done quickly by the
computer. If you want to show off, mention latent Dirichlet allocation (LDA) as your topic-
modeling algorithm of choice!
A second form of machine learning would be at work when text mining software takes
thousands of emails, identifies patterns in words such as repetition or proximity to each other, or
pores through email messages to tag possible indicators of insider trading.
Classification and clustering.
Whenever a law firm or law department has collected a set of data, it can use a range of
software tools to cluster the observations. This means that the software brings together related
clients, matters, or law firms based on the information available to it about them. Once the
software has clustered the observations, managers can more easily detect patterns and understand
similarities and differences. A chart known as a dendrogram can depict the clustering of data and
how clusters relate to each other. Somewhat similarly, software can classify observations into
similar groups. Both of these types of analytics help partners see patterns that they could not
otherwise detect from a massive set of data.
Other models can also classify new observations into the most appropriately fitting
group. With several types of algorithms, including K-Nearest Neighbor or Support Vector
Machines, you can classify clients or other data. You would be able, for example, to identify
publicly-traded clients or clients likely to reach a certain realization level.
Other varieties of machine-learning software do not require labels. Their models cluster
the data into groupings that will reveal something. For example, they might cluster a firm’s
clients by profitability. The K-Means algorithm can do this, and with the Principal Components
Analysis you can aggregate “variables” to find out which of them is more influential.
148
