Ontology Builder - Learn the Language of Your Business

Capturing the language of your business is one of the most POWERFUL things you can do to improve your enterprise search. With Noonean's Ontology Builder we make it easy to do that.

The Ontology Builder (formerly TopSearch Optimizer) automatically monitors the queries that users are entering and automatically creates an Ontology of the language of the Business. By understanding the specific language of a business we can improve search.

The Ontology Builder helps companies drive users to the right search documents and improves the search experience for 50-60% of the most common queries. It lets you tune your queries easily through a simple web interface and improve your hit% as well as boost the relevancy for your most important queries.


We've found many large companies abandon their TAXONOMY efforts. It's a lot of work and is often arbitrary, returns too many matches in search, and is often just a simple tag technology.  Noonean Ontologies are different.  They are all based on user pattern of use. A human merely has to "curate" out of an already defined list. Being based on real use is important. More importantly is it integrates into our whole query solution including advanced EnterpriseNLP and combined provides leading precision in enterprise search.  Being able to rank is just as important as being able to match and this is where taxonomies fall apart. The noonean approach combines AI and precision to correctly rank and bring the right documents to the top of your search. You wont spend your time endlessly managing ontologies with Noonean. 

What about tools that use Tuplet Extraction? (OpenIE, PoolParty) ? Knowledge Graphs ?

These tools are a simplified form of morphological grammar capturing only the main concept of a sentence. E.g.  John -> Walked -> School.  While useful for forming fact-bases and graphs, they are not a good tool for enterprise search as they distill things down to too simple a relationship resulting in over-extended queries which match too many things.  Knowledge graphs are just another bloated taxonomy usually intended to be used in a graph database. In the end the approach is simplistic and not much better than matching tags.  Conceptual indexing, a Noonean technology which will be available in our SULU release, is a much more advanced technique.

How does it work?

It works by capturing query statics on every query that a user executes. Then we filter them out searching for new terms and potential Corporate entities and apply advanced AI heuristics to dynamically create new Ontology entries related to key search terms.  Then a human takes over and confirms things with a managable set. This is much easier than trying to write complex rules. Once the ontology is set up, it works automatically helping to enhance user queries using a technique called "query expansion"

The system can also be used to examine "No-Click" queries - that is queries where the user did not pick an answer. Usually there's a correction needed and the Ontology Builder helps you find these.

What does Forrester say about Cognitive Search? 

Forrester writes that cognitive search should:
"Include integrated tools for usage analytics, tuning, and app dev. Use cases for cognitive search are not one-size-fits-all. They must be customized based on the kind of data that the use case requires to ingest, periodically tuned by administrators to boost relevancy, and embedded in applications used by employees and/or customers."

Unlike a synonym which replaces a query, the Ontology Builder uses query expansion to enhance the query.  Then sites can be configured with boosting to determine how much an effect the optimizer has on the user query. Importantly, the original user query is still maintained not replaced.

Later, when several months of data has been created, the need to re-organize terms and move data, update data, and delete entries becomes more important. The Ontology Builder supports all of these maintenance operations making it easy to maintain.

The Ontology Builder uses query frequency/month to both recognize corporate entities and enhance queries.

It is a game changing tool that has a dramatic impact on the precision of your user's queries and their satisfaction.

When you purchase Ontology Builder with EnterpriseNLP it extends the functionality of EnterpriseNLP with recognition of your business entities. This eliminates the problem of phrase detection in queries and the complexity and slow processing of phrase (or shingled) queries versus simple queries.

Lessons from MIT's TextFooler:

The researchers at MIT created a tool to try to stump popular NLP engines. 

TextFooler works in two parts: altering a given text, and then using that text to test two different language tasks to see if the system can successfully trick machine-learning models.The system first identifies the most important words that will influence the target model’s prediction, and then selects the synonyms that fit contextually. This is all while maintaining grammar and the original meaning to look “human” enough, until the prediction is altered.
In one example, TextFooler’s input and output were: “The characters, cast in impossibly contrived situations, are totally estranged from reality.”“The characters, cast in impossibly engineered circumstances, are fully estranged from reality.”In total, TextFooler successfully attacked three target models, including “BERT,” the popular open-source NLP model. It fooled the target models with an accuracy of over 90 percent to under 20 percent, by changing only 10 percent of the words in a given text.

Noonean survives "TextFooler" which really is a proxy for what we call a "near Miss". This is one of the core strengths of using Ontology Builder. These "near Misses" occur all the time. The CNN based models are rigid.  Think of Ontology Builder as being a flexibility point to herd your user's queries in the right direction. 

Let's take an example in depth from real life.  

         Corporate Entity:  529 college plan   ------- Child Of ------   College Savings Plans

         Search Terms:
               university plan 
               dorm housing plan
               college savings
               519 college plan  (example of a common typo) 

Now when user searches for "college savings" they will get to the "529 College Plan" site.

Let's consider a financial case:

            Corporate Entity:  Tax Free ETF   ------- Child Of ------   ETF

             Search Terms:
                      untaxed etf
                      tax advantage etf
                      untaxed fund
                      tax advantage fund

Now when the user searches for a "untaxed fund" they will get to the site that has the corporate language "Tax Free ETF"

How far the query expands can be configured so a user might be able to chose no expansion, expand with entity, and expand with terms of related entity.   

Contact Noonean today to learn more about how the Ontology Builder can optimize your business and your customer's experience today!