Enterprise Search Solutions — Ontology, Knowledge Graph & Semantic Search
Introduction
Today, enterprises are accumulating data faster than ever before. According to the report from Seagate & IDC (International Data Corporation) in 2020, enterprise data collection is projected to increase at an annual growth rate of 42.2 percent. The report expected the total enterprise data volume to be 2.02 petabyte (1 petabyte = 1000 terabyte). That is a lot of data. With such huge amount of data, ensuring that employees can search through data to find exactly what they need has become a critical topic. This is where enterprise search solutions come into the picture. Enterprise search solutions can connect to multiple data sources in the organisation like SharePoint, Outlook, Relational Databases and allow employees to search across all these data sources.
Evolution of Search
Before we dive into the capabilities of enterprise search, let’s have a look at the timeline below on how search engines have evolved over the years from the days as early as 1970.
Enterprise Search Capabilities
Typical enterprise search solution as following capabilities:
- ability to connect and crawl data from multiple data sources (both cloud & on-premise)
- a variety of search capabilities (keyword, facet, boolean, multi-lingual, semantic)
- recommendations and smart suggestions (auto-complete, spell-correction)
- ability to build a domain specific knowledge graph and ontology
- integration with Intelligent Document Processing (IDP) solutions
- operational capabilities: scalibility, high availability, disaster recovery
- security & authentication — integration with existing authentication systems like Active Directory
- analytics capabilities — insights into what employees are searching, dashboards and metrics
Looking at enterprise search solutions in the market, we will come across a lot on terms like ontology, knowledge graph and semantic search. So what are they and how do they play an important role in enterprise search solutions?
Ontology
Ontology comes from 2 Greek words: “Onto” which means existence or being and “Logos” which means science or study. In essence, ontology is a study of existence/entities. Ontology encompasses the following components: individuals, classes, properties and inter-relationships between the entities/concepts. Using these components, we can capture the entities and conceptualize the abstract view of entities/data that we have and how they relate to each other. For example: a company has multiple employees → an employee works for a company → a company has multiple projects → an employee works on a project.
Knowledge Graph
Knowledge graph is a way to represent a collection of entities with their values declared using the ontology and mapping the relationships between those entities. The entities are represented by nodes and the relationship between those entities are represent by the edges between the nodes which forms a graphical representation and hence, a knowledge graph. Below is an example of a knowledge graph:
Semantic Search
Semantic search is a highly efficient search and retrieval technique where a search engine not only finds a keyword match but also finds the contextual match between the search query and the content. It leverages the data to reveal the connections between concepts, entities and relationships.
For example: the word “jaguar” could mean an animal or a car depending on the context that it is being used. If I searched buying jaguar on Google search, Google knows that I am more likely looking to buy a Jaguar car and not an animal Jaguar.
“Semantic search retrieves information based on what you mean, instead of what you type”
So it is clear that understanding of the context is really critical in the semantic search. But how does the system knows the context in an enterprise search solution? This is where domain specific ontologies and a knowledge graph comes into play. We can create these ontologies and the knowledge graph with the help of Subject Matter Experts in the domain the solutions is expected to be used. By having a mapping of relationships between different entities, the system can now extract the context of both the search query and the content it has, to perform a semantic search.
Why semantic search matters in Enterprise Search Solutions?
In Managed Print and Document Services for Controlling Today’s and Tomorrow’s Information Costs published by IDC in 2011, it is stated as “IDC surveys find that the time spent searching for information averages 8.8 hours per week.” McKinsey also reported in 2012 that employees spend 1.8 hours every day searching and gathering information. In today’s world of information overload, it is crucial for employees to find the information they need as fast as possible to do their job well.
Semantic search helps employees access information faster. It is powerful. It is simple and it is an experience every single worker should have in their workplace.
Thank you for reading this far, I hope you have enjoyed reading this article. To get unlimited access to quality content on Medium, join as a Medium member!