How does this technology compare to what
a human can do?
When
reviewing a relatively small volume of uncomplicated
data a human could analyze the same data that our system
does and likely discover the same or similar results;
however, it would just takes them much longer. On the
other hand, the greater the volume of data and
complexity involved, the more difficult it becomes for
humans to consistently and accurately review and analyze
the data. Complex data with many variables renders the
average lawyer's analytic capabilities inefficient and
ineffective.
The
use of ontologies, taxonomies, and dictionaries
Relationships such as synonyms, hyponyms, (words that
are more specific than a more general word i.e. table or
chair are hyponyms of furniture) and hypernyms (words
that are more general than a given word i.e. musical
instrument is hypernym to guitar) need to be leveraged
to perform accurate analysis. Other aspects of relevance
are anaphora resolution and stemming. Again, you cannot
perform accurate analysis without being able to identify
the subjects, predicates, and objects. If codes were to
be used to replace the text analysis, the codes will
have to be assigned by a human after carefully reading
the text.
Analyzing natural language along with structured codes
Finding
relationships among unstructured concepts is not
possible without analyzing the context, which is the
text surrounding the concept. If text was abstracted to
codes, which many text analytic tools do, the context
will be lost. Hence, discovery using NLP is exhaustive,
i.e. nothing is missed. Importantly, one of the
characteristics of OLAP is that its dynamic nature
allows the user to try combinations and permutations
among many of the concepts. That feature provides the
user of NLP the opportunity to look at multiple
possibilities to see which works best for them without
repeatedly having to rebuild the extract. The result is
faster, deeper analysis and greater insight into the
data.
Using functionally complex ideas for analysis
Functionally complex ideas may be assembled, revised and
combined in English to include multidimensional “if
then” propositions as well as the elimination of
unneeded “noise” terms to focus on important
relationships intended by the analyst. Concepts
“flagged” by the Exeact analysis are more apt to be on-
target and in synch with analyst intent with less noise
or “false positives” included in the analytical “catch.”
This is in contrast to other technologies that force
unstructured data into a structured format or use
abstract statistical analysis to produce a one-
dimensional template that may identify desired data, but
mask the target with noise and unwanted clutter.
Moreover, with other technologies the analytical task is
more difficult due to loss of touch with the original
data. This is not the case with our e-Case Dictionary.
The
reference to the original document is always maintained
This
means at any time users can immediately drill-through to
the actual document to find the actual text and see
it in context. This is very important to users at
almost every level in an organization and especially to
lawyers reviewing documents and developing lists of
search terms and trying to assess where or not there is
a smoking gun.
Natural Language Processing Text Analytic Engine