What existed in the tech world in the 1990s, is hardly there in its original form today. But what about online search? Has it been reinvented? Most of us still use Google in the same way we used decades back? Since 1998, the algorithms are updated continually. Instead of using newer techniques, the major search engine giants have only been refining their existing features. They aren’t doing away with removing content quality as a defining factor, but they are just updating their algorithm to be adept at determining content quality.

But now things are gradually changing. Technology disruption is making inroads into online search. Deep learning, is revamping popular internet services like Facebook, Twitter and Instagram at a fast rate. In the last couple of years, even online search has been re-invented by Deep Learning, Machine Learning and Artificial Intelligence.

Deep neural networks, networks of hardware and software that approximate the web of neurons in the human brain are some of the factors at play that are re-inventing the online search. These neural nets can learn all kinds of useful functions by studying the vast amount of digital data. This includes functions such as identifying photos, commands given to smartphone as well as responding to search queries on the Internet. It has turned out that at times they can do a task much faster, better and larger than human beings.

Google’s Rank Brain: The Harbinger of a new era

Let us take the example of Google, the world’s most popular search engine. Google’s search engine has always been algorithm-driven such that it automatically generates a response for each of the search query. But these algorithms have their own constraints. They are automated based on a fixed set of rules designed by Google engineers. Unlike Deep Learning, these algorithms cannot learn and update themselves.

Google launches a deep learning system called RankBrain to help generate responses to search queries. RankBrain works by carefully by not just focussing on the keywords and past queries but it also tries to judge the mindset, behaviour and intent of the user searching queries. Rank Brain guesses from the word or phrase entered to find words or phrases that might have a similar meaning. Then these results are filtered accordingly.

It has been observed that in certain instances, its query handling mechanism it can handle queries better than hand-coded algorithmic rules which are at the mercy of human engineers. Artificial Intelligence is truly the future of internet search, not just Google search.

Data handled in a completely different way and it changes the way the keywords and searches have been handled so far. The search terms are classified into word vectors or distributed representations. So words or phrases that are closest to each in terms of linguistic similarity are grouped together.

Rank Brain actually trying to make sense of what people mean as they key in a search and record and adapt the results according to a certain understanding. Giving the users results to their searches in a short time and saving them from browsing through innumerable web pages is sure to give a unique experience to the users.

Increasing applications of AI & ML in internet search

It is not just web search engines that are leveraging Machine Learning to offer better results and enhanced user experience, even enterprise search vendors are now using AI techniques and machine learning and into their search engines. These engines take cues from the search behaviour of the users to improve search quality. Two of the enterprise search engines that are presently using disruptive technologies like machine learning and artificial intelligence techniques to enhance internal search results are Sinequa, a consulting firm and Highspot, a software company.

Sinequa ES Version 10 offers a platform that includes cognitive search and analytics. It uses advanced natural language processing (NLP) and machine learning algorithms to derive insights from data, whether structured or unstructured. On the other hand, Highspot uses Machine Learning to understand how users interact with each other and how their behaviour impacts the search operation.

Final Thoughts

Internet search backed by machine learning is still evolving. Machine learning isn’t perfect in itself, however, the more humans interact with it, the more accurate and smarter it gets. Technology is taking internet search function to a point where it can sneak into the minds of the users and deliver the results are more close to their intent. There will be minimal human intervention in the search engines in the times to come. After the initial query, machine learning will identify behavioural patterns and customise the results.

Links

Search engine journal