What is semantic search?
“Semantic search” is a means of making machines better able to understand a searcher’s intent and match it appropriately with knowledge on the web – making search engines better at providing answers.
Understanding Semantic search means understanding that search engine ranking algorithms are now harder to “game” than ever before, because the ranking signals being used are t – and as a result, the only sensible search strategy should revolve around the production and sharing of high-quality, valuable and unique content.This content should drive user engagement, be as discoverable as possible to engines and should reflect your authority on the subject matter. As author David Amerland states:
…content production, once an SEO necessity that ticked a requirements box, suddenly became the axis around which now revolve all your marketing efforts.
Developments in semantic search are an opportunity in disguise to be truly awesome at what you do best.
Search engines are now better able to understand meaning, having improved their ability to:
- Read text and extract “entities” (subjects, objects) from within that text
- Understand relationships between entities
- Define the nature of entities and publishers using methods such as Trust Rank, Knowledge graph and authority.
- Improve indexing and retrieval capabilities to find more and more helpful answers.
Let’s take a look at the major moving parts of semantic search and what they mean for you.
Entity Extraction: use metadata
This is the ability of a search program to read running text and analyse the subjects being discussed. “Extraction” means that, whilst analysing linguistic patterns, the program has identified matches for specific subjects and objects within the text, and builds a machine-readable understanding of these entities for future reference.
Consider the following example text:
Greek voters have decisively rejected the terms of an international bailout. The final result in the referendum, published by the interior ministry, was 61.3% “No”, against 38.7% who voted “Yes”. Greece’s governing Syriza party had campaigned for a “No”, saying the bailout terms were humiliating. Their opponents warned that this could see Greece ejected from the eurozone, and a summit of eurozone heads of state has now been called for Tuesday. Greek Prime Minister Alexis Tsipras said late on Sunday that Greeks had voted for a “Europe of solidarity and democracy”.
Entity extraction on this text could yield the following information:
Entities / entity type
- Greece / Country
- Syriza Party / Political Party
- Alexis Tsipras / Person
All of which can be stored within a search index for future reference.
Why does this matter? If you’re consistent in your content terminology and references, machines are more easily able to identify common entities within your text, and accurately associate your content with specific subjects which users may search for.
For some serious bonus points, take a look at the very helpful documentation at Schema.org, and learn about how you can apply microformats to your content, making it extra-easy for search engine spiders to understand the entities featured in your content.
Knowledge Graph: give good answers
Knowledge graph is the representation which a search engine builds on any given subject (or entity). It is a high-level mapping of the various entities within a larger entity, and how they relate to one another.
The image above shows the example knowledge graph for a search on “Thomas Jefferson” in Google. On the right side of the SERP, users can see rich, related information within the search results page, all of which is available because Google has built a knowledge graph of the entity “Thomas Jefferson” – associating his date of birth, spouse, and even similar “related searches” based on previous search behaviour related to this entity.
The sum of all of these information inputs (DOB, spouse, etc) yield the knowledge graph – a single knowledge base for the given subject.
Let a user’s search query be forever in your mind when you create and publish content.
Marketers who provide rich and thorough information about a given subject will be feeding the knowledge graph for a given subject and better able to help their audience find the answers they seek. Take some time to think about the type of questions you can provide answers to for the people you want to communicate with, and have a plan to publish content which answers these questions.
To get you started, consider the following definitions, which capture the various types of queries a search user may have:
- Factual: e.g. “What is the population of Japan.”
- Problem solving: e.g. “How to jetwash a patio”
- Lists: e.g. “Online CRM software providers”
- Location Information: e.g. “Delicious curry houses in Farringdon”
Trust Rank: spark conversations
Trust Rank is Google’s means of scoring the trustworthiness of a given content publisher. Google’s search algorithms look at a myriad of factors to determine this rank, but among the more influential signals are:
- The number of user comments on a publisher site
- The number of replies to publisher / commenters
- The number of times content from a publisher is shared
- The number of links to/from other content
- The average time spent on pages from users arriving via search engines
Publishers with a higher Trust Rank are more likely to appear in search results and will be better able to reach a wide audience.
To help search engines understand your trustworthiness, post content which gets people talking. Ask difficult questions, which get your audience talking, and whenever possible, share content which they can use. Be bold – put a stake in the ground for what you believe in, and generate intelligent debate among your audience.
Sparking interesting conversations online and encouraging others to comment on your content is good for your audience and good for search engines – it sends a clear message that your content is worth reading.
Authority: build your networks
One of the harder-to-decipher assessments performed by search engines, authority is essentially a function of:
- The number of other sources linking to you
- The availability of verifiable information about you as a person, publisher or company (e.g. your address, affiliations and other data)
The more able you are to confirm this information to search engines, the easier it is for search engines to infer your authority within a given domain.
To help demonstrate your authority, be an online networker. Form meaningful relationships with online publishers in your sphere of influence and interest, and share ideas with them. Link to content from their sites, and encourage them to link back to you. This will help justify your relevance and authority, and drive your online reputation for a given subject.
Check your technical boxes
Finally, there are some technical considerations which qualify as no-brainers, but are so often overlooked.
- Short summaries: for every content page you publish, have you provided a page <title> and <h1> heading which clearly describes your content in <70 characters? (70 characters are available within Google’s results page).
- Meta description: Make sure that you use a tag and set the description attribute to give a succinct, <150 character description of each piece of content which you publish. Done correctly, your meta description can entice readers to view your content by giving a clear indication on what they can expect to see on your page.
- Give good images: be sure that your images used are high-quality, optimised for the web and make sure that you use the alt attribute to tell search engines what content your image contains, as well as to help any visually impaired readers who may rely on screen readers to describe the content on your site.
To succeed in the age of semantic search:
- Be relevant, current and true to your brand position- publish content which will generate debate and interaction with your users, as this will increase your relevance in the eyes of google.
- Use everything you know about your users to think carefully about what answers they will be seeking, then publish and share these answers.
- Tick all the technical boxes: meta description, title, images and get your bonus points by reading up on schema.org.