How machine learning improves search engines
by Clemens Niekler, on Oct 27, 2019 10:59:29 AM
How often do you search for something on Google and how many times do you end up with a solution to your particular problem? To find the fitting answer(s) to the individual query, it takes a lot of steps behind the scenes.
During the time you have read the article just until here, Google processed around 580.000 search queries. Every day, the search engine delivers results to more than 3,5 billion questions and topics. But how do search engines work and how can they be improved?
How Search Engines work
Whenever you search for an answer to your problem or question, you get thousands of results shown within a fraction of a second. Nevertheless, multiple processes are going on in the background of search engines to find the right answer to your query.
First of all, a web crawler (also called web spider or web bot) crawls almost all the web pages to find out what the particular pages are about. Crawling the web allows for the information to be quickly accessible when needed resulting in minimal waiting time for the users of search engines. The bots are organising the vast amount of pages and topics available like a shop assistant would categorise CDs into genres.
Search engine index
Next up is the step of so-called indexing. Once the URLs are organised, they are sent back to the search engine and stored in data centres. Google, for example, owns several big data centres around the globe to process the massive amounts of data. Along with the URLs, data includes keywords, currentness, content and previous user engagement to deliver the best results to the search engine users.
At this point, all the data is made available to the searcher. The question now is: In what order should all the fitting results be shown? To tackle this problem, every search engine uses its own algorithm. Google doesn’t only use one algorithm, but multiple ones to present you the best results possible to your specific query. These algorithms undertake a process of live tests and thousands of Search Quality Raters who follow precise guidelines.
The criteria for the order of the results include various factors like query words, the relevance of the pages, their freshness and your location and settings at the time of the query. For instance, Google prioritises recent news to a topic over dictionary definitions. These factors allow for continually improving results. When you are searching for something in English, the search engine will list the outcomes in English as well. On the other hand, when famous people are googled the first results to pop up might be breaking news about them instead of their Wikipedia page.
Search engines and Machine Learning
By using all of these steps, we created BNA’s own search engine for architects called Archy. We helped our customer to improve it so that the users get results for architectural projects in the Netherlands. Not only the preceding steps are implemented.
Machine Learning is continuously improving the results Archy, search engine, delivers. Built with tools offered by Google Cloud Platform, the search engine becomes smarter with use and keywords like “school” or “bridge” provide more and more precise answers.
Machine Learning is one of the biggest keys for search engines to improve and succeed. Google uses ML for several areas reaching from pattern detection to image search. Even though human quality raters are working for Google, technology has helped the biggest search engine in the world to eliminate low-quality pages.
Furthermore, Machine Learning allows for the customisation of search suggestions and results. To clarify, when I searched for “football” first and then in a later search type “Amsterdam” Google suggests “Amsterdam football teams” first.
Moreover, Machine Learning may help the search engine to deliver more precise and more customised answers to the user’s query. It analyses behaviour and search requests without human help. Using Artificial Intelligence, therefore, allows employees to focus on other tasks machines can’t fulfil yet.
Machine Learning reveals a significant potential for search engines like Google. It is not only improving the ranking of search results but also technologies like voice search. In the future, it is also imaginable that the user won’t have to search for an answer anymore and receives push notifications of frequently asked queries.
Crystalloids helps companies improve their customer experiences and build marketing technology. Founded in 2006 in the Netherlands, Crystalloids builds crystal-clear solutions that turn customer data into information and knowledge into wisdom. As a leading Google Cloud Partner, Crystalloids combines experience in software development, data science, and marketing making them one of a kind IT company. Using the Agile approach Crystalloids ensures that use cases show immediate value to their clients and make their job focus more on decision making and less on programming.
For more information, please visit www.crystalloids.com or like us on LinkedIn.