Artificial Intelligence applications in search engines
Artificial intelligence is breaking into all areas of knowledge and many of the activities we carry out day by day.
The ones susceptible to be touched by the transformation are the repetitive, manual tasks that do not require great capacity but that can be moderately simple optimized.
There are others that, although automatic, are having a greater impulse through artificial intelligence to deliver improved results quickly, improving the customer experience and learning on the fly.
Web search
Something that is already common for most of us are the search engines like Google, which is the most popular in Mexico, although Bing and Yahoo are still far behind.
For these businesses, which live to give an excellent search service for what we are looking for, it is necessary to improve all the time and respond to what users require in the most precise way and do it quickly.
It is here where Google and later Bing have been improving their algorithms for some time, which, despite being complex sets of automatic rules, were fixed and the learning they could have was limited.
In the section that may have more impact is in the new queries, for which there is no search history, and those account for approximately 15% each year of the total search volume, to identify intentions and objectives within the queries and incorporate that new knowledge into the base it already has.
Since 2015, Google has enhanced its algorithm with RankBrain, a piece that applies machine learning to determine which results will be most interesting for a given search.
For its part Bing has launched in 2017 Intelligent Search that gives faster responses and takes into account more information to interact in a simple way with the search engine.
Going beyond the signals that are typically taken into account to interpret the user’s intention as the location of the device, the type of device (cell phone, tablet, computer), time of day, his search history, personalization options, previous queries and the words used.
One of the improvements that can potentially have the use of machine learning is that you can learn and lead to personalized responses, not just consider the most popular results, about which everyone is talking at the moment.
Taking into account the whole set of signals and how the user has behaved before can predict the best response and if it is wrong, learn from the error to correct itself.
History
For this artificial intelligence model and any other to work, it is necessary to have a history of the actions that we expect to predict and that have many variables (location, history, time, keywords used) in order to enrich the recommendations and have a greater possibility of combinations that derive in a response with a high probability of being the right one for the user.
One advantage of Internet search engines is that they have large amounts of data to train their models.
Context
A fundamental part of the search engines is to understand the context of the queries.
For example, if we look for it, Google will throw us tickets related to the movie that (it) was very popular. But I could be looking for the pronoun or maybe for information technology.
Although it is difficult to have context for only one word, it takes into account what users look for and what the user has particularly looked for in the past.
While if we do the same question in Bing, the result is similar, showing up IT the movie results primarly.
And if what we want is information about the demonstrative pronoun that, we would have to write it specifically to obtain results, being the question less popular.
Therefore the context of the words within a query is very important to understand exactly what the user is referring to and give him what he needs.
Intention
The key concept of search results is to understand and anticipate the motivation of users when writing their query to be able to give them what they expect to find.
Yes for example I’m looking for “World cup México”. On July 4th 2018 it would seem that the answer is simple, I mean how it went to Mexico during the World Cup in Russia.
Provides the results from that day’s match.
But if we think about the user, what is he really trying to find out?
- Of the participation of México in this world cup?
- Of the new world cup to be held in México by 2026?
- Of the World Cup of México 86 or México 70?
- The players of each edition?
- Historical statistics?
And today, August 2nd 2018, the Google results are as follows.
Bing responded more to this intention on the July query, with information on world futures and history of past world events in Mexico.
But to the same query today August 2, 2018, the results are focused on the just finished Russia world cup.
To answer it in a traditional way it would only take into account the signals like the quality of the content of the sites, or the number of reference links that the site has won.
As in the example, using RankBrain as a factor to decide the answers, it was taken into account that in July the users were more interested in knowing about the World Cup in Russia in which Mexico has just been eliminated, and in August that same interest persist but in a more general way.
Although these results may vary if I were in Russia, it should give me information about the following games or accommodations and availability of tickets to the matches or places to see them.
The idea of implementing RankBrain and Intelligent Search as part of the algorithm that decides the search results is to predict in the most accurate way the intention that the user has when making the query.
Application of Artificial Intelligence to define the best results
One of the great changes that it generates is the end of keyword saturation practice (keyword stuffing) and focus only on these words and their distribution on the site, which had been obsolete for some time now but with the application of the machine learning techniques and natural language comprehension will be less useful and we will have to change the way of thinking, which was previously adapted to how the search engine algorithm worked to do it the way we express ourselves.
Both verbally and in writing, in mobile search engines or on desks, everything that will affect the results that for the same user could be different according to the context and how it does it.
With this it will be possible for the search engine to understand the meaning and real value of the content of a site and its information, as well as the sentiment of the content to determine a position within the results.
Search by images
Another application that has artificial intelligence in search engines is the ability to search through images.
You can identify a product, what it is made of, where to buy it, people and places just by taking a picture.
Or just like in the movie Anon from Netflix, we can through our eyes combine AI and augmented reality to search everything using images.
It will be very useful to make more fluid the process of search for information or better yet, of products.
It is something that is still in process but will soon be available and will continue to improve day by day as it is used.
To learn more about artificial intelligence and chatbots, talk with my chatbot in messenger: https://www.messenger.com/t/GaboJimenez0
