Unlocking Value from Unstructured Data

TopNaturalLanguage
ProcessingApplications
inBusiness
UNLOCKING VALUE FROM
UNSTRUCTURED DATA

TopNaturalLanguageProcessingApplicationsinBusiness 1
For years, enterprises have been making good use of their
structured data (tables, spreadsheets, etc.). However,
the larger part of enterprise data, nearly 80 percent, is
unstructured and has been much less accessible.
From emails, text documents, research and legal reports
to voice recordings, videos, social media posts and
more, unstructured data is a huge body of information.
But, compared to structured data, it has been much more
challenging to leverage.
Traditional search has done a great job helping users
discover and derive insights from some of this data. But
enterprises need to go beyond search to maximize the use of
unstructured data as a resource for enhanced analytics and
decision making.
This is where natural language processing (NLP), a field of
artificial intelligence (AI) that’s used to handle the processing
and analysis of large volumes of unstructured data, can be a
real game changer.
While AI describes a broad range of technologies, NLP is
one of three AI-driven capabilities that enterprises can
readily harness to create business value and competitive
advantage:
1. Internet of Things (IoT): applying technologies, such
as real-time analytics, machine learning (ML), and smart
sensors, to manage and analyze machine-generated
structured data
2. Computer Vision: using digital imaging technologies,
ML, and pattern recognition to interpret image and video
content
3. Document Understanding: combining NLP and ML to
gain insights into human-generated, natural language
unstructured data.
As it powers document understanding applications that
unlock value from unstructured data, NLP has become an
essential enabler of the AI evolution in today’s enterprises.
This white paper discusses the emergence of NLP as a
key insight discovery technique, followed by examples
of impactful NLP applications that Accenture has helped
clients implement.
Natural
Language
Processing
Market Size
Estimate
Source: www.marketsandmarkets.com/PressReleases/natural-language-processing-nlp.asp

EmergenceofNLPapplications
intheenterprise
As AI technologies like machine learning (ML), deep learning,
and cognitive search have entered the mainstream, we’ve
seen a rapid evolution in NLP applications. This has been
driven by two key factors. First, having grown accustomed to
virtual assistants like Siri and Alexa, as well as sophisticated
search like Google, in their daily lives, users expect the same
experience in the workplace. Second, today’s NLP no longer
relies on manual, rules-based approaches: integrating ML
with NLP allows for far greater automation, scalability, and
accuracy.
Enterprises can leverage NLP applications in two ways
– query understanding and content understanding – to
improve both user experience and insight discovery:
• Provide better, more targeted responses by
understanding the user’s questions and intent
• Identify out-of-scope requests and present intelligent
alternatives
• Extract business entities from text documents to identify
employees, customers, products, procedures, etc.
• Identify and understand the meaning of natural language
content – documents, reports, emails, etc. – to provide
natural language answers
NLP enables improved understanding of user queries
and enterprise content. This ensures that every user is
connected to the most relevant, helpful resources which
would otherwise remain hidden within vast quantities of
data.
Today’s NLP applications – spanning industries and ranging
from chatbots and virtual assistants to intelligent business
solutions – are supported by a robust NLP technology
ecosystem.
Understanding the user’s
natural language query inputs
(text or speech)
Understanding content
(mining insights from vast
amounts of unstructured data)

NLPtechnologyecosystem
Rising demand for NLP-powered applications is fueled
by a growing range of technologies, from open source
frameworks and evolving vendors to cloud-based APIs.
Commonly-used NLP technologies include:
• Python NLTK
• Python SpacY
• Python GenSim
• UIMA
• GATE
• Apache OpenNLP
• Quid (market and competitive intelligence)
• Twiggle (e-commerce)
• Google Natural Language API
• Amazon Comprehend
• Microsoft LUIS (Language Understanding
Intelligence Service)
• Pool Party
• SmartLogic
• Cotical.io
• SAS NLP offering
• IBM Watson
While NLP is a relatively new enterprise technology, it’s
being enhanced every day. The ML algorithms supporting
NLP are improving all the time, with industry giants like
Google, Microsoft, and Amazon all making strides to
improve accuracy.
With this growing ecosystem of NLP solutions, enterprises
have increasing flexibility to select appropriate approaches
and toolsets for their specific use cases. Accenture
leverages this robust ecosystem and our own technology
assets to orchestrate the different components of NLP
applications – making them easily maintainable and
scalable.
We’ve helped many clients implement NLP applications.
Some of the most impactful are highlighted in the next
section.
Open source frameworks
Industry specific platforms
Cloud API providers
Evolving vendors
Other:

Enterprise
NLPapplications
inaction

Chatbotsandvirtualassistants
Chatbots are becoming ubiquitous. All around us, Siri,
Alexa, Cortana, and Google Home are incorporating natural
language conversations between humans and AI into
everyday interactions. As consumers get accustomed to
these virtual assistants, they expect the same experience in
the workplace. The emphasis is on how well computers talk
to humans, interact with them, understand text and so on.
The challenge for enterprises? Providing their people with
sophisticated natural language understanding applications
that meets their expectations – and improving information
discovery and collaboration.
CHALLENGE
Sources:
1. Accenture Research: Chatbots are Here to Stay
2. IDC FutureScape: Worldwide IT Industry 2018 Predictions
3. Gartner: Conversational AI to Shake Up Your Technical and Business Worlds
4. Gartner: Market Guide for Conversational Platforms
1
1
1 3
2
4

Chatbotsandvirtualassistants(contd.)
Powered by NLP, enterprise chatbots and virtual assistants
can enable conversations between humans and computers
in everyday business processes. Bringing deeper natural
language understanding, they enhance search as well as
providing an entirely new way for employees, customers
and partners to interact with enterprise data and work more
productively. For example, an enterprise chatbot that can
hold a dialog with the user to acquire targeted information
(dates, times, order details, etc.) and/or integrate with
business systems to complete common office tasks quickly
(reservations, orders, etc.).
By combining content processing, ML, NLP, and voice
recognition, enterprises can transform simple
search-based intranets or support portals into AI-powered
virtual assistants. Whether it’s a question about technical
support on a consumer electronics website, pricing and
product details on a partner or e-commerce website,
investment products on an investor portal, or management
advice on a hotel franchisee site, AI-powered chatbots and
virtual assistants can play key roles in process automation
and user experience improvement.
• Improve business processes and reduce support costs
by enabling AI-driven self-service virtual assistants for
customers and partners
• Enhance search and knowledge-seeking experiences for
customers, partners and employees
• Improve brand reputation through more efficient
processes and more engaging customer experiences
SOLUTION BENEFITS

Intelligentdocumentanalysis
With overwhelming volumes of incoming documents
and correspondence every day, large organizations often
struggle to properly analyze and derive useful insights from
this content. Whether it’s legal documents, medical records,
policies, or contracts, without a centralized and automated
analysis approach, it may be difficult for organizations
to effectively understand and make use of the document
content to support decision making and operational
efficiencies.
Intelligent document analysis uses AI techniques including
NLP, entity extraction, semantic understanding, and ML to
analyze content, extract meaning, and reliably aid process
automation and decision making. These applications can
identify specific items of information in documents – like
date, order number, and policy number – so they can be
categorized and analyzed.
Core functionalities of this solution can include:
• Optical character recognition (OCR) – converts different
types of documents, such as scanned paper documents,
PDF files, or images into editable and searchable data
• Text analysis – analyzes documents to identify specific
language or terms and extract linguistic meaning
• Deterministic classification – uses a pattern-based
classifier to look for sequences of terms that indicate a
specific sort of document
• Machine learning – trains a ML model with example
datasets to predict document type or extract and classify
text (e.g. learning aircraft component names)
CHALLENGE
SOLUTION

Intelligentdocumentanalysis(contd.)
BENEFITS
• Improve compliance and risk management
• Drive internal operational efficiencies
• Enhance business processes

Documentsearchandmatch
Enterprises store vast quantities of documents, so it’s critical
to be able to identify and match documents of related
themes – not just keywords – quickly and accurately.
In the recruiting and staffing industry, for example, recruiters
are in a race to find the right candidate before competitors
identify the same person in their candidate databases. Given
the large amounts of candidates that most recruiters deal
with, it’s easy to lose track of the best matches.
There are many candidate search and match solutions
delivered as standalone products or embedded into
applicant tracking systems which can provide convenience,
but they’re limited to basic resume parsing and matching
corresponding metadata to a job posting. As a result, most
of these embedded search systems fail to significantly
improve fill rates or deliver a better experience to recruiters.
The ideal solution is an application that can indicate people’s
fit for jobs by accurately and automatically comparing job
descriptions to people’s resumes, CVs or other documents.
The solution we’ve developed is bidirectional and works like
this: a job description (the whole document) is submitted as
a search request and the comparison system automatically
returns a short list of the best-qualified candidates from a
database of CVs. Alternatively, a job seeker (or professional
recruiter) submits a CV and the system returns a list of the
most appropriate, currently-available vacancies.
It’s an approach that combines advanced search techniques,
analytics, NLP, and ML to provide statistical and linguistic
capabilities for understanding applicant profiles and
identifying the best candidates. Predictive analytics uses
previously successful candidates’ placement histories to
find additional candidates ahead of a new job posting –
giving recruiters a head start against the competition.
CHALLENGE SOLUTION

ML leverages feedback from recruiters to improve results
over time.
This solution also adds value to human resources teams in
global enterprises, enabling them to identify the right people
from hundreds of thousands of geographically-scattered
experts. For example, a project manager can submit a
project requirement document as a search request or ask a
question like “Who are our NLP experts in North America?”
The system will return a list of relevant experts for the project
by automatically analyzing and identifying matches from
employee databases.
Documentsearchandmatch(contd.)
• Reduce the time to fill a job position or to identify experts
for specific skillsets within an organization
• Scale to millions of jobs postings and resumes (in
recruiting) or multiple employee and project databases
(in an enterprise “Find the Expert” application)
• Increase revenue and reduce costs – candidates’
matches can be found within the recruiter’s database
rather than via paid job boards
BENEFITS
Delivering results across multiple use cases
Document search and match can be applied to various
types of enterprise documents. Examples include
analyzing legal documents for variations of risky
contractual languages or, in the financial services
industry, cross-analyzing loan or mortgage documents
with borrowers’ profiles.

Datastorageoptimizationanalytics
Corporations routinely maintain petabytes of content in
expensive, on-premise storage, so it’s no surprise that
CIOs want to reduce costs by deleting or migrating stale
or obsolete content. Standard storage management
software applications can count the bytes, but they can’t
compute the business value of content or the risk of losing
data. Organizations need an approach for automating the
migration of content to lower-cost storage, while providing
file-level traceability and real-time tracking of ROI.
To reduce storage, organizations must answer these FAQs:
• When was the content created? Last accessed?
• Who created the content? Are they still employed?
• Which business unit owns the content?
• Which products are associated with it?
• How much does it cost to store?
• Is it covered by a data retention policy?
• Can we move to low-cost cloud storage?
With a data storage optimization analytics solution, IT
departments, CIOs, and content owners get the visibility
they need to identify content eligible for deletion or
migration to lower-cost storage.
First, organizations should use enterprise data connectors
and search engines to help them understand the enterprise
context for their data. Connectors help join content source
metadata with other corporate sources, with source
metadata and context indexed into a search engine and
made accessible for ad hoc search and visualization.
Business rules should also be developed by stakeholders
to help drive data movement to optimize storage utilization
and achieve program ROI. Once the business rules are
laid out, organizations can identify files eligible for lower-
cost storage, archiving, or deletion, if obsolete. Where
possible, this should be an automated process that provides
traceability.
With NLP and ML, documents and their content can be more
accurately read and identified. This drives storage cost
savings through rapid detection of duplicate or
near-duplicate content, as well as providing a 360-degree
view of enterprise data.
CHALLENGE SOLUTION

Datastorageoptimizationanalytics (contd.)
Example business rules
• Retain active files on primary storage
• Archive content inactive for two or more years
• Archive content authored by contractors and terminated
• Provide actionable visibility into content in high-cost
storage
• Identify content provenance, ownership, and frequency
of use
• Apply domain-specific business rules to identify eligible
content and migrate it to lower-priced storage
• Enable cloud-readiness
• Track savings (potentially millions per year) and compute
ROI in real-time
BENEFITS

Sentimentanalysis
Sentiment analysis, also known as opinion mining, uses AI
to automate the process of identifying opinions about a
specific subject from a piece of content.
As mentioned earlier, it’s estimated that 80 percent of an
organization’s data is unstructured in the form of emails,
chats, articles, documents, web content, and social media.
Manual analysis of this information for customer, product, or
employee sentiment would be practically impossible. With
AI-driven sentiment analysis, an organization can detect
people’s opinions or feelings about a topic and uncover
actionable insights that would otherwise be unobtainable.
In its simplest form, sentiment analysis categorizes a
sentiment as positive or negative. It could also quantify the
sentiment (e.g. -1 to +1) or categorize it at a more granular
level (e.g. very negative, negative, neutral, positive, very
positive). By combining NLP technologies, text analytics,
linguistic analysis, and ML, sentiment analysis applications
can cope with the complexities of language.
CHALLENGE SOLUTION

Sentimentanalysis (contd.)
Example use case:
Voice of the customer – sentiment analysis is widely applied
to voice of the customer materials like reviews and survey
responses, online and social media, and healthcare materials
for applications that range from marketing to customer
service to clinical medicine. By using sentiment analysis
across this data, organizations can better understand
customers (and possibly change how they engage with
them), predict demand, and improve overall company
performance.
• Provide marketing and competitive intelligence
• Enhance product development
• Improve customer retention
• Analyze the impact of an event (e.g. a product launch or
redesign)
BENEFITS

Insiderthreatdetection
Costly data breaches may originate from external culprits
or employees. But in many cases, an insider (employee,
former employee, contractor, business associate, etc.) who
has authorized access to valuable data can intentionally
or inadvertently cause even greater damage. And as data
volumes and accessibility grow, organizations – government
and commercial, large and small – need to guard their data
against these incidents.
Insider threats could involve fraud, the theft of confidential
or commercially valuable information, the theft of
intellectual property, the sabotage of computer systems,
or the disclosure of information damaging to a company’s
brand or reputation.
However, when organizations have large volumes of both
structured data (documents, spreadsheets, transaction
records) and unstructured data (social media, emails, voice
recordings, notes), it becomes very difficult to conduct
accurate data classification, monitoring, and analysis.
Integrating search, analytics, and NLP can help to solve
this challenge.
ON THE RISE – Insider threats account for nearly 75
percent of security breach incidents.
According to the Ponemon Institute’s report, “2018 Cost of
Insider Threats: Global Organizations,” the average cost of
an insider threat annually is $8.76 million.
With an effective insider threat detection solution in place,
organizations can avoid the lengthy legal battles – as well
as monetary and reputational losses – resulting from insider
trading, non-compliance, leaks of trade secrets, data
breaches, and government intelligence leaks.
This is where integrating NLP-based insider threat
applications can help determine if there is any illegal or
nefarious intent within communications and detect threat
patterns for rapid risk mitigation.
CHALLENGE
SOLUTION
Source:
1. https://securityintelligence.com/news/insider-threats-account-for-nearly-75-percent-of-security-
breach-incidents/
2. https://www.observeit.com/ponemon-report-cost-of-insider-threats/

Insiderthreatdetection (contd.)
Solution features:
• Search engines can scale to billions of records
• Threat investigation using data available via the search
engine
• Integration with third-party archiving solutions and
analytics dashboards
• Greater flexibility and customization for in-depth analysis
and reporting
• A 360-degree view across organizational data from
multiple sources
• Mitigate risk by detecting patterns and identifying red
flags
• Keep a better pulse on the organization by customizing a
“risk” score for any employee
• Faster time to incident and response with comprehensive
enterprise visibility
• Deeper insights with a full complement of analytics
frameworks for all threat detection workloads
BENEFITS

NLPinyourenterprise
Analyzing structured data alone is no longer enough.
Sophisticated business analyses, predictions, and decision
making all need more. The use cases we’ve provided
show how unstructured content can be used to unlock
tremendous new insights. With a well-implemented NLP
solution in place, organizations can enable a deeper
understanding of unstructured content, providing enhanced
BI and analytics.
As your organization starts designing and building your NLP
applications, it’s essential to ensure that your IT staff and/or
implementation partners have the bandwidth and expertise
required to conduct a thorough assessment for aligning NLP
technologies with your business objectives. Whether your
organization is only in the initial phase of evaluating various
NLP technologies or has already identified the preferred
solution for implementation, we can help – from evaluating
and selecting the best-suited solution to implementing,
tuning, and managing the application.
Looking for support and want to leverage Accenture’s
NLP expertise? Contact us to discuss your goals and start
defining a strategy and implementation roadmap for your
NLP applications.

About Search & Content Analytics
Search&ContentAnalytics,formerlySearchTechnologies,ispartofAccenture
Applied Intelligence. We live in a data-driven world. But not everyone is
making the most of their data. 80 percent of all data is unstructured – imagine
the hidden insights trapped within unstructured enterprise content such as
voice, images and emails. At Search & Content Analytics, our mission is to help
enterprises unlock the full value within their unstructured and structured data.
We combine innovative technologies such as machine learning and natural
language processing with search and big data analytics to transform the way
people work. Whether it’s improving intranet and website search, monitoring
internal communications to detect insider threats, helping recruiters match
jobs to résumés, analyzing oil wellhead reports, or exploring molecular data,
we bring comprehensive search and analytics services to clients across
industries. Clients include organizations in e-commerce, media, healthcare,
financial services, recruiting, manufacturing, and the government sector.
What knowledge and insights are trapped in your data? Let us help you find
better answers.
Visit our webpage to learn more about our capabilities and client work.
About Accenture
Accenture is a leading global professional services company, providing a broad
range of services and solutions in strategy, consulting, digital, technology
and operations. Combining unmatched experience and specialized skills
across more than 40 industries and all business functions – underpinned by
the world’s largest delivery network – Accenture works at the intersection of
business and technology to help clients improve their performance and create
sustainable value for their stakeholders. With more than 469,000 people
serving clients in more than 120 countries, Accenture drives innovation to
improve the way the world works and lives.
Visit us at www.accenture.com.
Copyright © 2019 Accenture.
All rights reserved.
Accenture, its logo, and High Performance
Delivered are trademarks of Accenture

Unlocking Value from Unstructured Data

More Related Content

What's hot (20)

Similar to Unlocking Value from Unstructured Data (20)

More from Accenture Insurance (20)

Recently uploaded (20)

Unlocking Value from Unstructured Data