Stop searching

and start solving

Increase productivity by unlocking access
to data. Stardog’s smart search provides
fewer, more relevant results so your team
stays focused on what matters most.
 STOP SEARCHING AND START SOLVING

Having the right information at your providing the right information to customers increases
fingertips is critical — whether you’re a user satisfaction and engagement.
computational biologist evaluating clinical data to
But as search tools have become prominent, user
identify a promising drug target, a customer service
expectations have also increased, dictated by the tech
representative troubleshooting a problem, or a
titans setting consumer expectation for finding the right
manufacturing executive looking for the most readily
information, quickly. Studies show that search sessions
available replacement part. In each scenario, the
decrease with increases in response time as small as
information required to solve these problems is spread
100-400ms. Chatbots and voice assistants like Alexa are
across different systems and lives in a variety of formats.
also changing user behavior — now users are expecting
As a result, people wind up hunting down information
the single, best answer from their AI. Additionally, voice
across various systems — wasting significant time —
search leads to more conversational requests, which can
or sacrificing untapped potential because they lack
confuse even an advanced search engine.
sufficient inputs to make the best decision.
While there are plenty of enterprise search options
Search has emerged as a solution to unlock access
on the market, they’re failing in the face of increasing
to information. Arming employees with the right
complexity and performance demands. Stardog’s smart
information saves time and lets experts do their
search solution reduces time spent searching up to 90%.
job better by reducing blind spots and preventing
duplicate work. For products with built-in search,

“Stardog enables you to browse through the data and all these relationships. It’s
a 10 to 1 savings. It’s not only less overhead, it’s much better job satisfaction and
getting the knowledge in hand that you lacked before.”
Program Data Integration Manager, Exploration
Systems Division, NASA
 STOP SEARCHING AND START SOLVING

SIGNS YOU NEED AN UPGRADE

SIGNS YOU NEED AN UPGRADE

1 You don’t get the answers you need

Irrelevant results happen when the search engine can’t
accurately capture user intent. While search engine The search engine
products have become proficient at fuzzy matches for
misspellings or recognizing synonyms, they still struggle
can’t accurately
to interpret the complex mental associations our brains capture user intent
create between concepts. Bottom line: if the search
engine can’t understand the question you’re asking, it
can’t return the best results.

2 Answers you know should be there are left out

Subject matter experts are often the ones who realize
that results they know are correct are left out! A common
The correct source material
culprit behind this is that the correct source material hasn’t hasn’t been indexed by the
been indexed by the search engine. More and more,
that source material doesn’t just live in databases, but search engine
also lives in images, videos, PDFs, and other so-called
“unstructured” data.

3 It takes too long to find the right answer

You might find the answer you want, but it’s not the top
There is increased
result. Especially with new formats of search applications pressure to return just
like recommendations engines, voice assistants, and
chatbots — essentially any AI app — there is increased
one correct result.
pressure to return just one correct result.
 STOP SEARCHING AND START SOLVING

BUILDING A SMARTER SEARCH ENGINE

You can solve all of these issues by using a knowledge IN PRACTICE
graph to power your search engine. A knowledge Imagine you are hungry for Indian food and you search
graph captures both data and it’s real-world context, “good naan”. What people ask for is good naan, but what
easily representing the layers of connections between they really mean is “where can I go in order to get some
various data. By making real-world context machine- good naan?” And the answer to that is the restaurant’s
understandable, a knowledge graph allows for more human name. A knowledge graph represents naan at various
responses to search queries. restaurants in the following way — showing how the menu
item naan is tied to this particular restaurant, Heritage India.
If you’re unfamiliar with knowledge graphs, you might
not realize that it’s hard to find a tech leader that hasn’t But aside from this restaurant, the knowledge graph also
embedded a knowledge graph into their product to identifies that naan is available from additional restaurants
improve user experience. From Google’s “knowledge on the delivery service Grubhub, and is for sale at Trader
panel” providing key results at the top of the page to Joe’s as well. The query has expanded to include all
Amazon using a knowledge graph to return a single perfect possibly relevant options.
answer to Alexa users, knowledge graphs are proven to
return more relevant results.

Smart search is an enhanced version of traditional search

options like Elasticsearch, which rely on indexing content
to return results quickly. Stardog’s smart search actually
uses similar indexing capabilities to the search giants but
also leverages the power of the knowledge graph to return
fewer, more relevant results.

Step 1: Ask a better question

Before searching for answers, Stardog makes sure it’s
asking the right question. Humans often don’t ask the right
question, so this is a critical step. Since Stardog captures
real-world context, it’s especially suited to rewriting
search questions to ensure the right data is evaluated. This
ensures that the search application accurately reflects user
intent by interpreting the context of the search terms.
 STOP SEARCHING AND START SOLVING

STEP 2: SEARCH ACROSS SOURCES IN PRACTICE

Once the query is rewritten to accurately reflect the user’s Incorporating data of varying structures is critical for
context, it is evaluated in two phases. First, the basic search applications to reflect user intent. For example,
full-text search is executed. Fundamentally, this is just how do we hone in on the “good” in the “good
like Elasticsearch; fuzzy matches, case inconsistencies, naan” search? Structured data sources — tables and
and stop words are accounted for and the relevant databases — could include the list of restaurants and
results from the full-text index are ranked and returned. their addresses, SKUs at Trader Joe’s, or aggregated
Then, Stardog evaluates the connections between the restaurant review score from Yelp. But a strong review
first stream of results against all the other information score for Heritage India doesn’t necessarily mean they
in the knowledge graph — structured, semi-structured, have good naan, per se. And associating overall score
and other unstructured data — to ensure that the search to a specific dish may even be misleading. This is where
occurs across the full breadth of relevant information. unstructured data comes into play — for example, by
incorporating sentiment analysis of specific written
Virtualization is key for scalable search solutions. Given reviews that mention naan.
that search applications often live on top of knowledge
bases, virtualization cuts down on what would be an
otherwise complex and cumbersome ETL system
migrating data (from dozens or even hundreds of
systems) into a single repository. Working with copies of
data instead of source data both introduces risk (human
error) and degrades data quality (data is not current or
comprehensive).
 STOP SEARCHING AND START SOLVING

STEP 3: REFINE RESULTS IN PRACTICE

At this point, Stardog has ensured that all the possibly The system may assume that the default response is to
relevant results are included in the search and ranked send a searcher of “good naan” to the optimal restaurant
these for relevance. Now, in order to ensure that users based on ranking and distance (in this case, our example
don’t waste time scanning through results, Stardog of Heritage India).
further refines results.
Based on business logic, the knowledge graph quickly
Stardog’s best-in-class Inference Engine is executes a series of If-Then statements to determine the
responsible for intelligently applying business logic best recommendation:
to the underlying data at query time in order to • if its late at night (restaurant is closed, but Trader
provide situationally optimized results. This business Joes is open) → recommends TJs
logic is stored centrally in a data model. Compared
to coding business logic on top of a database • if its raining outside (assuming people wont want to
storing this data, the low-code knowledge graph walk in the rain) → recommends delivery
stores business logic centrally as modular rules.
Stardog was designed from the ground up to support even
Compared to the recommender systems that cant
the most complex logic, no matter how many dependencies
offer a rationale for their results, the Inference Engine
must be evaluated. This performance is critical given the
displays all logic for each result, making explainable
user expectations for quick search results.
AI a reality.
 STOP SEARCHING AND START SOLVING

SMART SEARCH IN PRACTICE For example, an organization with 5,000 Y

users who only use search once per day will
Stardog’s unique combination of enriching
end up making 1.3 million searches in a year.
the indexed search results by refining them
If it takes these users 30 seconds per search
based on the context from the knowledge
to find the right information they are actually
graph search ensures the user will receive N
looking for, it means the company is spending
fewer but more relevant results — this is key
close to 11,000 hours per year just looking
to the users getting their work done more
for the information they need in order to
quickly. The productivity gains from these
perform their job duties. Some organizations
accuracy improvements will quickly be felt Y
have experienced up to 90% time savings
across an organization.
by implementing Stardog; but even if you
assume smart search only reduces this daily
Y search to 20 seconds (a 33% improvement),
N that equates to just over 3,500 hours of
Save up to savings across the company in a year. This

90%
on time spent
N
time savings equates to hiring an extra 1.5 full
time employees per year; time better spent
investigating potential cures for a disease,
designing a better customer experience to
searching increase retention, or using the quality data to
further improve the supply chain.
 STOP SEARCHING AND START SOLVING

Fewer, more
Springer Nature is one of the world’s relevant results
largest publishers; they have mountains of documents Dr. Yafong Fan, Science Librarian at the
containing scientific information. Their Database and University of Science and Technology of
Solutions Group is charged with productizing this China, explains one way SpringerMaterials
knowledge for researchers who are investigating supports her work.
scientific unknowns — their search requirements
are robust. By rebuilding their portfolio of search- “One of the students was exploring TiO2 flexible
based products with Stardog, platforms including dye-sensitized solar cells, in particular, its
SpringerMaterials are able to arm these researchers with electron transport properties. The student came
fewer, more relevant results. for a tutorial at the beginning of the research
Benefits of transitioning to Stardog smart search: project. They were at a loss, having searched
for information on major search engines, which
• On-the-fly interpretation of search context yielded millions of results. None provided the
and users’ intent
information they needed to start the project.
• Adaptive ranking of results based on search
context They also tried several major specialized databases
but again didn’t find the information they needed.
• Entity-based search and explore capabilities
What made the difference was a simple search
• Reverse search to find substances based on on SpringerMaterials. We typed the term “titanium
property values dioxide” along with “transport properties” in the
• Smart light-weight inference on substance search box. The number of results was far less than
and property identifiers using a general search engine. Crucially, the results
gave the student the information they had been
• Plug and play integration of external
knowledge models looking for. In fact, almost every result from this one
search was relevant to the project. The student
• Dynamically created landing pages went away able to begin work, when before they
hit a dead end.”

Read the Case Study

 STOP SEARCHING AND START SOLVING

At Banque Nationale du Canada the majority of As a pioneer in research-based healthcare and

internal information isn’t stored in tables. Instead, this a massive pharmaceutical and biotech company,
information is stored in documents, chat logs, and Roche maintains dozens of laboratories worldwide.
blog posts that are easy for anyone to produce, but These labs are distributed both in research focus and
very difficult to sift through in order to answer urgent geographically, and, to meet their particular research
questions. BNC staff spent too much time trying to needs, each lab maintains its own Laboratory
find the information they needed to do their jobs. Information Management System (LIMS) to store and
This limited how many customers they could serve manage its data and work processes. Because each
and how well they could serve them. To improve LIMS had different data definitions and relational
internal productivity and customer satisfaction, BNC schemas, it was all but impossible to answer broad
built a smart search app to take the information questions across the enterprise like “What were all of
locked in text-based data silos and get it in the hands the chromotographic results from the past year for a
of their employees. particular molecule?” Roche needed to structure its
knowledge in order to answer the questions coming
Their solution, Mon Hub Intelligent (MHI), is a single from the business side.
application that connects the data from legacy data
systems across the bank. Unifying the data using Roche Smart Search will give all lab personnel and
virtualization allows them to leaving the data in place. analysts a global view of the knowledge held within
Using cognitive search functionality, the MHI chatbot the dozens of LIMS across Roche and its subsidiary
translates the users’ natural language search and Genentech. This new application will begin by
applies the context of the question to the keyword connecting the LIMS on a subset of attributes that
query. Now, bank employees no longer need to are critical for lab functions. Understanding that
wade through multiple data systems to try to find the more data only breeds more questions and makes
information or hope that the first answer they find is, people hungrier for insight, their knowledge graph
in fact, the correct answer. Instead, they have access is constantly iterated on – adding more data and
to a ‘personal assistant’ that surfaces the information creating new relationships. Now, thanks to the
they need from the applications where the most flexibility of Stardog’s platform, the Roche team can
collaboration happens. integrate the data needed to answer a particular
question within days.
ENRICH YOUR SEARCH APPLICATION
WITH STARDOG
Improve productivity by arming employees
with the information they need to make critical
decisions and save countless hours by slashing
time spent searching through irrelevant results.
Get started today.

Contact us: sales@stardog.com

Learn more: stardog.com