The Search for Meaning

lavishgradeΛογισμικό & κατασκευή λογ/κού

25 Νοε 2013 (πριν από 3 χρόνια και 11 μήνες)

89 εμφανίσεις

© 2008 IBM Corporation

The Search for Meaning



Enterprise Search Summit, May 2008

Mike Moran

IBM Distinguished Engineer



© 2008 IBM Corporation

2

Separate worlds are coming together

Content Management

Text Search and Analytics

Business Intelligence



© 2008 IBM Corporation

3

Classification and Content Discovery Vision

Records

Management

File

Shares

Basic

Content

Services

Email

Existing ECM

Repositories

Unclassified

(or Lost)

Documents


Content
-
based

Solutions

Business

Process

Disorganized

Content

Email
Management

Doc Class: Marketing

Doc Class: Legal

Doc Class: Finance

Doc Class: R&D

. . .

IBM

Classification

Module

AutoClassify

Extract Metadata

Self
-
Learning

Taxonomy Proposer


Analyzed Content



OmniFind Analytics Edition

















Visualization /

Interactive Mining


Timeline Analysis


Topic Extraction


2
-
D Analysis


Trend Analysis


Distribution Analysis

Organized Content

in IBM ECM repositories



© 2008 IBM Corporation

4

Business Insight Case Study: Manufacturing


An automotive company mines technician notes and
warranty claims to preemptively address quality problems

PART 1: Fuel Pump

PART 2: Fuel Filter

PART 3: Wiring Harness

PART 4: Wiring Harness Cover

PROBLEM 1: Corrosion


PART 3: Wiring Harness

ACTION 1: Replace


PART 1: Fuel Pump


PART 2: Fuel Filter

ACTION 2: Remove


PART 4: Wiring Harness Cover

Semantic Search and Drilldown Interface



© 2008 IBM Corporation

5

Business Insight Case Study: Financial


A financial services company analyzes and visualizes
emails for compliance monitoring



© 2008 IBM Corporation

6

Business Insight Case Study: Pharmaceutical


A pharmaceutical manufacturer mines public and
corporate repositories of patents and scientific literature
to set strategy for highest
-
value drug development



© 2008 IBM Corporation

7

Business Insight Case Study: Healthcare


A healthcare insurer is optimizing coverage models by
mining and correlating treatments, outcomes, and costs
from unstructured records



© 2008 IBM Corporation

8

Business Insight Case Study: Telecom


A major mobile phone carrier optimizes customer
retention, improves offerings, and reduces support
costs by mining customer contact transcripts

Customer

Employees

FAQs

FAQ

VOC

Other
sources

Analysis

System

Business System

Voice

Email

Web



© 2008 IBM Corporation

9

Business Insight Case Study: Law Enforcement


A major law enforcement agency analyzes patterns in
police reports and other content to solve crimes, predict
crime trends, and optimize police deployment



© 2008 IBM Corporation

10

Keyword search for
“Neon” (the car) returns
irrelevant results

The Basics: Keyword Search



© 2008 IBM Corporation

11

Behind the Scenes: Text Analytics

A

was

driven

Neon

by

Driven By

Arg2:Person

Arg1:Car

VP

PP

NP

Parts of Speech

Annotator

Named Entity

Annotator

Relationship

Annotator

Timothy

Higgins

Car

Person



© 2008 IBM Corporation

12

Semantic search for “arabic male”

Keyword search for
“arabic male” returns
actual Arabic male
names, including
variants and aliases



© 2008 IBM Corporation

13

Semantic search for “Hispanic”

Keyword search for
“Hispanic” returns
actual Hispanic
names and places



© 2008 IBM Corporation

14

The Next Level: Content Understanding

Start by searching on a
topic

Relevant
concepts
automatically
detected

Tonality in document results is
automatically identified



no need to read!



© 2008 IBM Corporation

15

Behind the Scenes: Sentiment Analysis

Sentiment annotators perform two main functions:

1.
Recognize words that are “positive” and/or “negative”


bargain
,
successful
, ...,
bad
,
pollution,...


One word can have different meaning:

fine
(adjective) vs.
fine

(noun)


Based on standard + app
-
specific word lists


2.
Recognize syntax in order determine the sentiment of a paragraph


A
bargain



positive


(Is) A
bad

bargain



negative


(Is)
not

a
bad

bargain



positive


Underpinned by sophisticated understanding of linguistic rules



© 2008 IBM Corporation

16

The Next Level: Insight from Text Analytics



Advanced
natural language
extraction

annotators


Integrates
structured data

with
analyzed unstructured content


Rich
mining UI

to analyze, correlate,
and search unstructured and
structured data


Targeted solutions

for Customer
Care and Quality Insight use cases


Additional solutions

in development
for eDiscovery and more



© 2008 IBM Corporation

17



For decades, DoCoMo’s priority has been
a customer
-
oriented business strategy,
offering products and services based on
real voice of customers (VoC)

The Corporate Strategy :



Japan‘s premier mobile communications company



focuses on customer
-
oriented management policy


Increased use of voice of customers


All employees can access, share
and utilize VoC for their business


Changes to the business from VoC
analysis.


Set initial parameters of mobile
phones based on VoC


Started new Premium Club points
program based on VoC


Better rate in model and service
upgrades for loyal customers


Many others…

Staff

Customer
Service

Business

System

Analysis

System

Analyzer



Collecting VoC



Sharing VoC



Accessing VoC



Utilizing VoC

VoC

Analysis of



major VoC



trends in VoC



remarkable VoC

for specific
products/services

Call Center

The Results:

The Solution:

Over 2 million
meaningful
records of VoC
per a year

Enabled by IBM OmniFind Analytics Edition

Because of its tradition of customer
-
oriented business, DoCoMo was an early adopter and has been a "design partner" for OmniFind

Analytics Edition’s evolution. DoCoMo’s strategy enrichment has been strengthened through deployment of OmniFind Analytics E
dit
ion.



© 2008 IBM Corporation

18

Business Intelligence faces challenges with
unstructured information


Typical BI tools can only
report on and analyze
structured data


Critical knowledge and
details required for deeper
analysis are buried in
unstructured data


Supporting documents not
replicated in the warehouse
often include critical
background



© 2008 IBM Corporation

19

Add Context: BI can figure out the
“what”

Product has over
79% return rate



© 2008 IBM Corporation

20

Add Context: Search shows
why

Structured
Warranty data

Search results view:



© 2008 IBM Corporation

21

Add Context: Search shows
why

Documents from

Marketing Repository

Documents from
Warranty Repository

Content

Manager

QuickR

Documents from Mfg
Repository

FileNet

Search results view:



© 2008 IBM Corporation

22

Add Context: Search shows
why

Search results view:

Hinge falls apart

Inadequate requirements

Improper rivet chosen



© 2008 IBM Corporation

23

Create insight: What if we had no OLAP cube?

Visualization
indicates areas
needing
attention. Simply
click to activate a
query.

Drill all the way
down into original
text.



© 2008 IBM Corporation

24

Create insight: Search as an alternative to BI


When full BI is overkill, modern enterprise search
capabilities can offer an alternative for basic ad
-
hoc reporting


Or as a way to decide what data
requires

a cube

Search for:

Top prospects in the East with
sales > $1M



© 2008 IBM Corporation

25

IBM continues to add to its convergence story


Industry
-
leading text
search and analytics

(Gartner Leader Quadrant)




Plus FileNet Content
Management



Plus Cognos Business
Intelligence



© 2008 IBM Corporation

26

OmniFind Analytics Edition Wins Prestigious SSPA Award!


OmniFind Analytics Edition wins the
Innovator in Voice
of the Customer

award from the SSPA!






Winning case study on how a client uses OmniFind
Analytics Edition to analyze customer calls and
contacts to:


Provide business users with rapid insight into customer
concerns and issues


Detect and manage potential customer churn


Identify new marketing opportunities


Determine claims processing bottlenecks causing payout
delays, overpayments, extra claim reviews


Example of
Content
-
Centric BI

delivering new ROI on
unstructured enterprise content

“This award recognizes an SSPA partner that enables
member companies to
better solicit, capture and
analyze customer feedback
, ideas and experiences,
and
present the findings in an actionable way

to the
support organization, as well as across the enterprise.”



© 2008 IBM Corporation

27

Stop by the IBM booth to see OmniFind in action

Go to the IBM
booth

Subscribe to
my free
Biznology
e
-
mail
newsletter

Enter to win a
free book





ibm.com/software/data/enterprise
-
search

“Buy this book, read it, and then
read it again.”


--
Chris Sherman, Search Engine Watch






Named one of 2007’s 10 “Best
Business Books”


--

Richard Pachter Miami Herald