As an organization, you can’t survive without an efficient enterprise search function. When you are a law firm, offer payroll services, or are a consultancy firm, information, knowledge, customer files and past projects are the foundation of your business. But in that overwhelming number of documents, you find it hard to retrieve the information you are looking for. So far you weren’t able to set up the folder structure that makes you find information more easily. And now you are considering tagging all your information. But that costs a lot and is simply not efficient, as we explained in our previous blogpost.
One of our customers wanted to experience the difference in efficiency between two enterprise search engines. And so they set up a comparative test between a tag-based search function and Alexandria.Works.
Imagine a large Flemish city where residents can ask the administration questions about directives and regulations, decrees taken by the city council. Imagine having a chatbot that gives relevant answers without the help of any municipal employee. Imagine the savings, knowing that today, city employees answer these questions on the phone.
The test: one dataset, one set of questions, two enterprise search mechanisms
In order to find the best solution for the above-mentioned challenge, the city administration set up a comparative test between 2 suppliers of enterprise search solutions. As the basis for the test they provided a corpus of 30,000 pages in 5,500 files, in total 1.5 gigabytes in size.
Supplier 1, enterprise search mechanism 1: Supplier 1 claims that supervised learning and tagging make large amounts of information more searchable. Their approach is to make experts determine relevant tags that are catalogued in a hierarchical structure. Afterwards, they tag the documents themselves, either automatically or not. A process that can take up to several months.
Supplier 2, enterprise search mechanism 2: Alexandria.Works proposed a tool that, autonomously, analyses the texts, determines relationships between words and gives them meaning. This analysis is done unsupervised, without the intervention of any of the customer’s employees. The computer and a mathematical model do all the work. On their own. In thirty minutes.
To test the two search engines the customer asked a user panel to come up with a set of 100 typical questions asked by citizens on the phone. The test consisted of a comparison of the two systems above, with the same data set (the communal decrees), and the same questions.
And the winner is …
Despite months of preparation, the tag-based solution scored less well on the questions: 82% correct and 18% wrong.
Alexandria.Works on the other hand scored 94% correct and 6% wrong. That is three times fewer wrong answers. And of course, the biggest gain lies in the unsupervised learning, saving the human labor cost and gaining time.
But there is more. Alexandria.Works is language-independent. Whatever language your documents are written in, the system will always find relevant meanings. And when your organization and your collective knowledge expand, you can simply add new documents in real-time. There is no need to re-index your database and, again, no need for human intervention.
A flexible solution for every knowledge question
This Proof of Concept proves that Alexandria.Works is an expert in the field of knowledge-based searches and offers the best enterprise search engine. And whether it’s data from a law firm or the European Commission, a large city or payroll organization: indexing is fast, language-independent and very flexible. Due to the low threshold, the tool can is invaluable in various domains. As long as searching for knowledge is at the base of the activity.
There is definitely a trend in making more and more use of chatbots to help customers – see also “Could chatbots help you with more efficient (and responsive) Customer Service?” Alexandria.Works can help to make your chatbot more intelligent.
Want to know more? Intrigued by the tool? Let’s meet for coffee and we’ll explain it to you in detail. Contact us now.