Enterprise search is 10x complex than Google search!

Empowering an enterprise with the convenience of a Google-like search is the holy grail in enterprises. A simple search box that understands the question in natural language and gets answers seems to be the need. While we are proud of what we have accomplished at MachEye, customers' expectations of search technology can vary greatly, and the technology needs to address incredibly complex expectations. While Google focuses primarily on relevance and performance, we must also pay attention to the presentation aspects. Without the right presentation, the answers are neither intuitive nor insightful, and even look bug-like. And not everyone appreciates or even understands this. So I thought of explaining this with just 2 perspectives (out of the many more that we have come across).

1. Giving the answer, and not just leading to it

I was looking for “hiking trails in Yosemite” in Google, and after scrolling through the initial 3 ads, 1 featured snippet, and 1 Q&A section, I finally got to the list of results that included “Top 5 trails”, “Best trails”, and 51,60,000 other results. What I wanted was information on all the paths available so that I could choose the best one for me. I don't see any trail paths on a map in any of the answers; instead, I just see point locations. These points may lead to an answer, but they are not the answers themselves.

Apart from not seeing the expected responses, most of us would have had to read through a big list of results to get the one that was most beneficial. And I'm sure none of you would have considered reporting it to Google. There is an expectation match that is set to spend a reasonable amount of time exploring each answer, which begins with a partial response.

While users don't mind exploring Google search results for the right answer, their expectation changes when it comes to enterprise search.

In an enterprise search, most users would expect one and only one answer, the right answer, to emerge. This one expectation results in an extremely complex technology. Interestingly enough, it's actually harder to explain this feature engineering to someone than building it!

2. Presenting the answer

The search box may look simple but under its hood lies an intricate network of various technologies like NLP, ambiguity corrections, ranking models, prioritization models, NLG, machine learning models and algorithms. Each of these technologies must be carefully calibrated and orchestrated to work in such a way that you can get most accurate and relevant results. There are minute nuances that we as humans understand by default but to teach them to machines is a whole different ball game.

Let’s take a scenario where you ask an enterprise search for “working days in 2022”. You would expect a number to be returned, say 250 (which excludes the 105 weekends and 11 holidays). Now if you ask for “holidays in 2022”, what would you expect – a count or a list of dates of the holidays? That is the tricky part.

While 250 is an acceptable answer for “working days in 2022", why isn't the number 11 acceptable for “holidays in 2022”? It’s not a bug but an expectation mismatch.

Is 11 a wrong answer? Definitely not. Will it be a different answer that what’s expected?

Let me present a different perspective.

For the question of "working days in 2022", what if the system returns an answer as "Monday, Tuesday, Wednesday, Thursday and Friday". Is it a wrong answer? It is an expectation mismatch.

This nuance to display count of days when you are asking for days is something that needs to be taught to the search engine. And it’s not easy. It is indeed a very hard problem to solve.

MachEye’s Intelligent Search is an enterprise search based on a complex and well-researched logic. I'm creating this piece to pique your interest and give you a better understanding of what we've produced, which is 10x more complex than Google. There are many more aspects at play when it comes to enterprise data and answering business questions, including data quality, query completeness, metrics, business drivers, security, governance, and role-based access regulations, to mention a few. I'm not going to go into detail about them because if I do, I'll end up writing a book.

If you are curious to learn more about what’s under the hood of MachEye’s Intelligent Search, please give me a buzz or please request for demo here.

MachEye's AI-powered enterprise search offers intelligent search, AI-powered insights, why analysis with an easy-to-consume interactive presentations for every business user.