Information Retrieval: The in's, the out's and everything in between

Information workers waste an inordinate amount of time orchestrating work rather than doing work. Instead of creating new content to drive our businesses, organizations, and missions forward, we spend our time looking for information…it’s terribly inefficient.

-Timothy Morey & Roberto Veronese, strategists at Frog Design

In early 2005, Microsoft took it upon themselves to try and quantify just how productive people were at work. So, to do this they carried out the Microsoft Office Personal Productivity Challenge where over 38,000 employees in 200 countries were quizzed on how productive they perceived themselves as being, whether or not there were any specific obstacles that prevented them from being more productive, and what would enable them to be more productive. The results were incredible. On average, over a typical 45-hour week, the average respondent estimated that they were unproductive for about 17 of those hours, implying that they were unproductive for roughly 38% of the working-week. Why? Well, some of the most common productivity pitfalls that were identified included; unclear objectives, lack of team communication, ineffective meetings, unclear priorities, and procrastination.

Considering that this survey was taken over a decade ago and workplace technologies have improved considerably, you would be correct in assuming that employee productivity would also have improved. Well, not necessarily. According to a more recent study carried out by the financial management services company, Think Money, in 2015 employees wasted an average of 60 hours each month due to similar sorts of reasons to the ones highlighted a decade earlier.

Yes, you could argue that this figure actually represents a drastic fall in the amount of time wasted on a per week basis (15 hours per week), but this low level of productivity is still unsustainable.


But what does all this have to do with kare's Information Retrieval (IR) technology? Well, what is interesting to note is that one of the recurring suggestions for increasing employee productivity that was identified in both studies was to create a means of quickly finding any electronic documents as and when they were needed. This was because it was widely perceived that our way of working and searching for information then, was ‘terribly inefficient’.

It’s clear that there is obviously a demand for a system that enables people to quickly and seamlessly find the documents they are looking for when they need them. That’s exactly why we started shifting our focus towards developing advanced IR technologies that would enable our users to be as productive as possible.

Wait! Before we go any further, please explain to me what Information Retrieval is.

One of the hardest parts about defining IR is that it is quite a broad term that encompasses a variety of different activities. For example, one could argue that the simple process of browsing a cookbook to find the perfect Sunday roast recipe is a form of IR. However, for the sake of this post, and for your personal clarity, we’re going to use Manning, Rahavan and Schutze’s definition of what they believe constitutes IR. “Information retrieval (IR) is finding material (usually documents) of an unstructured nature (usually text) that satisfies an information need from within large collections (usually stored in computers).

Here at kare, we use state-of-the-art IR techniques to accurately find the right bits of information that you seek. What this usually entails is that our technology will automatically find the most relevant documents that satisfy your information needs, as expressed by your initial query.

Furthermore, similar to the way in which we find the most relevant answers to your questions, our technology also intuitively locates the most context-relevant documents that are directly related to any specific query.  

What’s unique about the way in which kare approaches IR is that while many traditional IR systems are based on methods that can’t learn from existing signals (e.g. term frequency-inverse document frequency(tf-idf)), we use models based on deep neural networks that are able to learn latent representations of a resource, based on their relevancy to other resources. In other words, our model is ‘smarter’ than traditional IR approaches such as tf-idf.

This is because most of the conventional IR technologies do not currently possess learning mechanisms (i.e. the ability to learn and improve themselves with time), as they merely adopt static approaches that simply compare texts by projecting them in a vector space. Our deep neural network (DNN) approach, however, processes any given pieces of text and then learns the associations between them, thus making it better at picking up any relationships that are not obvious, and that would not usually be picked up by existing IR technologies. 

Therefore, our technology is capable of getting better and better as more users interact with it, and unlike many feature-based methods, the execution of our approach is almost instant. Furthermore, to make our resource relevancy method more accurate, we apply a more vigorous model to the top-ranked results, making it slightly slower but considerably more accurate. Therefore, we have built a system that is both faster and more accurate than conventional IR technologies.

All this sounds great, but how is it useful to me?

The significant technological advancements that have taken place over the last decade, which have helped facilitate the seamless communication between people in all corners of the globe, has also led to a significant increase in the amount of information we have to process on a daily basis. While some of you reading this will welcome the recent influx of information, others may feel like this ‘information overload’ is too mentally taxing and completely unsustainable. 

Although we may have access to new sources of information, we also tend to lose important information in all the ‘noise’. Very often, the knowledge that we seek is scattered across various locations and not only requires a substantial amount of effort to find it but also expects you to use exact terms or phrases to search for that specific piece of information.

Well, the good news is that the information you seek is still out there and kare's IR technology has just made it considerably easier for you to find it. Our unique IR model can find and surface the information that you require, even if you have slightly misspelled what you are looking for, or did not use the exact same terminology. As a result, this saves our users a lot of time and frustration while also helping them to work more efficiently.


Nice! But how do I experience this technology in action?

Our easily available IR technology makes it extremely easy for you to experience it first-hand. Through our unique Customer Self-Service technology, we’ve made it incredibly easy for you to enhance your existing website. By seamlessly integrating it with our Self Service solution, you will be able to experience our IR technology as it enables your customers to accurately and efficiently answer their own questions regarding your product/service. To book a demo, just click here.

If you would like to find out more about our technology, what we do, or would simply like to chat, then please feel free to contact a member of our friendly support staff at: