Semantic Search and AI
You may be hearing more and more about semantic search with the role of voice search capabilities increasing within Google and Bing. However, semantic search has existed in search engines for years so why now do Google want us to understand semantics? And what does it mean for SEO? Mediaworks explains more:
What is Semantic Search?
Semantics in search context means the words or phrases entered and the logic associated to them. Simply, Google’s aim is to understand what you are entering and provide the most reasonable / contextual answer to support your query. Take the query – “where is the best sandwich?”, will produce a different range of answers if you are asking someone in the middle of New York to the person sitting next to you in the office. The question “where is the best sandwich?” uses exactly the same words but the context isn’t the same. Google’s use of semantic search intends to act in the same way providing answers based on contexts such as location, intent and natural language.
What are Google doing to develop semantic search?
Search engines are always going to be challenged by the ever-developing world of semantics. However, Google’s algorithm update in 2013, Hummingbird has went a long way to understanding the user’s search query, search patterns, device type, time of day and many more contexts.
Google are constantly looking for further ways to understand semantics and have recently announce a pair of interactive experiments that uses artificial intelligence to allows users to understand semantics and natural language processing. The demo tools Talk to Books and Semantris are set up for the AI to consider what the user types as an opening statement and then looks across the pool of responses to select the response best fitted to the statement. The objective from Google is to use billions of lines of dialogue to teach an AI how real human conversations flow.
Talk to Books
Talk to Books is a demo tool that lets you type in a query or a statement and the machine learning-trained algorithm will search through the thousands of books in the model to find the response best suited to your query. The experimental tool acts differently to the standard Google search by highlighting the answer in bold instead of having to click through to a link to find the answer.
Talk to Books does show some limitations in the model however, with the queries only effective if asking a question that is factual such as “Why is water blue?”, however when asking it a more complex query that has geopolitical context or topics of modern cultural and historical importance such as “When did Joan of Arc die?”, the tool struggles to perform and lists several irrelevant answers. Overall, the tool offers great insight to what the future of AI may look in Google.
The second of the experiments released by Google, called Semantris is an interactive word association game that scores the words on the screen based on the answer you enter. For example, if you are given “bike” as the word at the top and the user types in “tyre”, Semantris will then rank the 10 words and give you points based on how well it thinks the semantic relationship between “tyre” and “bike”. Google advises when playing Semantris to try using slang, technical terms, pop culture references, synonyms, antonyms, and even full sentences.
Both of these AI experimental tools are used for Google to gather data expected to be used to inform their search technology going forward. They are also great tools for users to interact with AI models in a real-world scenario.
How you can incorporate semantic search?
One area where semantic search is highly useful in search is providing answers to questions from users. Google and Bing use this content as a Featured Snippet as it understands the context of the question and provides you with the best answer.
Take for example; “How to remove AirBnB review”
The answer now only tells the user how to remove the review, through the use of Structured Data. However, the additional benefit is the use of the “People Also Ask” feature based on the semantics of the question.
Similarly, a search for “What food to eat to lose weight” goes one step further with Semantic Search and produces a mixture of Structured Content but also an answer carousel as seen in the screenshot below:
To achieve results like this, you have to create pages that contains structured, ordered listing content whilst also provide a question and an answer. This is the backbone that powers voice search, for example when a user searches for a local business they will ask for information such as operating hours or company address, so structuring your data so the search engine can classify this information is crucial for your SEO.
This is the time to get ahead of your competitors and start optimising your site by structuring of your data, contact Mediaworks today to talk with our experts on +44 (0) 330 108 4263 or email email@example.com.