Semaphore Case Studies presented for MarkLogic World 2024

On Tuesday, September 24, 2024, Urmi Majumder and Madeleine Powell presented at the MarkLogic World Conference on the Semantic Maturity Spectrum. Semantic search has long proven to be a powerful tool in creating intelligent search experiences. By leveraging a semantic data model, it can effectively understand the searcher’s intent and the contextual meaning of the terms to improve search accuracy. In this session, we will present case studies for 3 different organizations across 3 different industries (finance, pharmaceuticals, and federal research) that started their semantic search journey at 3 very different maturity levels. For each case study, we will describe the business use case, solution architecture, implementation approach, and outcomes. Finally, we will round out the presentation with a practical guide to getting started with semantic search projects using the organization’s current maturity in the space as a starting point.

1. Semantic Maturity Spectrum
Search with Context
Mature
Semantics
Operational
Semantics
Semantic
Foundations
30% of organizations
10% of organizations
60% of organizations
Where does your organization fall?

2. September 24, 2024
How to Enable Search with Semantics
Presentation for MarkLogic World

3. ENTERPRISE KNOWLEDGE
Outline
EK at a Glance Our Approach to
Semantic Search
Solutions
Case Studies Getting Started
with Semantic
Search

4. Our Approach to
Semantic Search
Solutions

5. ENTERPRISE KNOWLEDGE
Search Capability Progression



Keyword
Search
Navigate by filtering and refining by categories, metadata, and attributes
Customize results based on user profiles and past behavior
Understand intent and context to return results for
synonyms and related concepts
Proactively suggest content based on
interactions and relationships
Natural language dialog
Ability to search content by matching keywords and return relevant results
Intelligent Knowledge Access & Discovery
Personalized
Search
Semantic
Search
AI
Search
(Knowledge Graphs,
Vector Embeddings)
(LLMs, Generative AI,
RAG, Chatbots)
(User Profiles,
Recommendation
Engines, User Analytics)
(Indexing, Ranking,
Query Parsing)
Mature enterprise search provides a
flexible, contextual interaction with
relevant information to empower people
with timely, personalized knowledge.

6. ⬢Understand user intent and the
meaning of user queries

⬢Natural language processing (NLP)
and machine learning to analyze how
concepts are interrelated

⬢Identification of related concepts,
synonyms, and even ambiguous terms

⬢More comprehensive and relevant
results where multiple meanings are
possible
Streamlined User Experience
Users find what they need more
quickly and spend their time on
strategic work.
Enable Action
Users are given information that
allows them to take immediate
action and effectively complete
their work.
Drive Discovery
Users discover information they
didn’t realize they were looking
for which surfaces key insights
and improves the search
experience.
2
1
3
Outcomes
Why Semantic Search?

7. 30% of organizations
10% of organizations
60% of organizations
Semantic Maturity Spectrum
In EK’s experience, most organizations fall within semantic or operational semantics on the
maturity spectrum.
Mature
Semantics
Operational
Semantics
Semantic
Foundations
Case Study #2
Case Study #3
Case Study #1
Proof of concepts/pilots,
working with one dataset or
use case
Production
implementation
across 2-3 data
sets
Production
implementation
handling core
operational capabilities

8. Case Studies

9. Case Study 1: Financial Services Firm

10. 30% of organizations
10% of organizations
60% of organizations
Mature
Semantics
Operational
Semantics
Semantic
Foundations
Semantic Maturity Spectrum
A Financial Services Firm started their semantic search journey with strong enterprise
reference data, but no centralized location to manage that data.
Case Study 1: Financial Services Firm
Proof of concepts/pilots,
working with one dataset or
use case
Production
implementation
across 2-3 data
sets
Production
implementation
handling core
operational capabilities

11. ENTERPRISE KNOWLEDGE
Simplify enterprise risk management ecosystem to align operational and
compliance risk functions and reduce manual work and duplication
across the firm.

Create end-to-end risk reporting and transparency and connect systems
across the ecosystem.
Increase data quality and
consistent user experiences for
end users managing business
risks and regulatory matters.
Project Objective
Specific Objectives
Shift from application-centric to
data-centric program via semantic
solutions and data modernization
program.
Case Study 1: Financial Services Firm

12. ENTERPRISE KNOWLEDGE
Case Study 1: Financial Services Firm
Semaphore Integration Architecture
Taxonomy Consumer View (Application Integration)
Note: Arrow direction is
driven by the system
initiating the call

13. Approach for Central Reference Data
Management
Aggregate
firm-wide
reference
data
Enrich &
connect
Central
consumption
service
Adoption of
taxonomies by
applications
Case Study 1: Financial Services Firm

14. Outcomes for Financial Services Firm
Case Study 1: Financial Services Firm
Data connected across systems using taxonomies to
create a linked data environment
Central search service using standardized metadata
18 Taxonomies Developed 50 Information Sources
Analyzed
11 Consumer Applications Reached

15. Case Study 2: Pharmaceutical Company

16. 30% of organizations
10% of organizations
60% of organizations
Mature
Semantics
Operational
Semantics
Semantic
Foundations
Proof of concepts/pilots,
working with one dataset or
use case
Production
implementation
across 2-3 data
sets
Production
implementation
handling core
operational capabilities
Semantic Maturity Spectrum
A Pharmaceutical Company started their semantic search journey with robust content and
data, but no enrichments to provide structure and or meaning.
Case Study 2: Pharmaceutical Company

17. ENTERPRISE KNOWLEDGE
Provide a unified view of existing product development data and information
across sources for analytical and compliance purposes.
Enhance search experience
through metadata enrichment .

Enable faster and easier creation
of compliance reporting.
Project Objective
Specific Objectives
Facilitate more efficient drug
product development.

Improve data standardization and
metadata quality.
Case Study 2: Pharmaceutical Company

18. ENTERPRISE KNOWLEDGE
Semaphore
Integration
Case Study 2: Pharmaceutical Company

19. Approach for Metadata
Enhancement MVP
Use case
definition &
search experience
design
Semantic
model
development
Metadata
extraction &
enrichment of
documents
Search integration
Case Study 2: Pharmaceutical Company

20. Outcomes for Pharmaceutical Company
Search enhancements through faceted search and
search with context
Case Study 2: Pharmaceutical Company
Extraction and utilization of critical information stored
within industry-specific document types
7 Taxonomies Developed 27 Information Sources
Analyzed
1 Consumer Application Reached

21. Case Study 3: Scientific Research
Laboratory

22. 30% of organizations
10% of organizations
60% of organizations
Proof of concepts/pilots,
working with one dataset or
use case
Production
implementation
across 2-3 data
sets
Production
implementation
handling core
operational capabilities
Mature
Semantics
Operational
Semantics
Semantic
Foundations
Semantic Maturity Spectrum
A Scientific Research Laboratory started their semantic search journey with a few unrefined
taxonomies and concept models but did not have a formal strategy for structuring their
unstructured data.
Case Study 3: Scientific Research Laboratory

23. ENTERPRISE KNOWLEDGE
EK supports the research laboratory in their multi-year effort to migrate content
and data from a legacy repository to a new Document Management System
(DMS). Semaphore serves as the centralized location for metadata
management .
Auto-classify
documents that are
submitted for the
new DMS.
Project Objective
Specific Objectives
Enhance APIs to
support integration
of additional
sources with
Semaphore.
Case Study 3: Scientific Research Laboratory
Support
development of a
custom front-end
resource built on a
knowledge graph.

24. 1
3
2
Phase 3
Goal: Leverage centralized
metadata
●Knowledge Graphs
●Pilot Search System
●Migrate to closed side
Phase 1
Goal: Get gold sources for
metadata values
●Built initial dictionaries
●Developed Integration API to
connect systems
●Identified topical taxonomies
Phase 2
Goal: Standardize, apply, and
centralize metadata
●Support system roll out
●Developed
auto-classification
pipelines
Building the Foundation
for Semantic Search
Case Study 3: Scientific Research Laboratory

25. Approach for Semaphore as a Central
Reference Metadata Management Platform
Refine existing
taxonomies
Integrate with
prioritized
sources
Establish rule nets
for classification
Build
classification
pipelines
Case Study 3: Scientific Research Laboratory
Goal: Launch a new Content Management System updated with standardized
metadata from Semaphore to increase findability of documents.

26. High-Level Integration
Architecture
Case Study 3: Scientific Research Laboratory
Information
Request
GUIDs &
Change Details
Retrieve/Update
Concepts & Tasks
Change
Notifications
Content / Data
Management System
API Middleware

27. High-Level
Classification
Architecture
Case Study 3: Scientific Research Laboratory

28. Outcomes for Scientific
Research Laboratory
Metadata applied from Semaphore models to over 90,000
unstructured documents during system migration
Case Study 3: Scientific Research Laboratory
Ability to extract structured information from unstructured
content (for retrieval & AI applications)
Creation of a graph composed of standardized metadata
from various sources to enable semantic search across
prioritized applications

29. How to Get Started
with Semantic Search
Projects

30. Getting
Started
Mature
Semantics
Operational
Semantics
Semantic
Foundations
30% of organizations
10% of organizations
60% of organizations
Proof of concepts/pilots,
working with one dataset or
use case
Production
implementation
across 2-3 data
sets
Production
implementation
handling core
operational capabilities
❏Identify pain points in current search experience
❏Define use case and supporting search features to alleviate pain points
❏Define conceptual model for use case and underlying metadata needed to
instantiate model
❏Implement model and metadata extraction pipeline
❏Leverage metadata in improved search experience