Semantic search vs keyword search: what has changed in the way SEO is done

Semantic Search vs keyword search


This expression ‘semantic’ is often heard when people discuss latest SEO practices. It is said that modern search engines are capable of semantic search. If you are new to blogging/ web technology, SEO is search engine optimization or actions that help web pages in being shown high on search results.

To de-jargonize 'semantic', let's see the simple dictionary meaning of this word. This word means ‘relating to meanings of words and phrases’. So, big search engines like Google and Bing do not look at a word or phrase blindly but try to find the true meaning of the word/ phrase used in search query. That's it: semantic search = search based on understanding the meaning of search terms.

Though semantic web was conceived around 2003, it was not until 2013 that the concept was used effectively by search engines. In that year, Google came with its Hummingbird algorithm update and changed the way search was carried out by the search biggies. Till then, search engines depended heavily on keywords that searchers typed on their browsers, and matched these keywords with web pages in their databases. SEO guys stuffed keywords in web pages and easily got them on top of search results. SEO fraudsters even stuffed unrelated keywords within useless content and made it shine in search results. So, the techniques of semantic search have improved search results and made them less prone to manipulation.

For serving the best results, search engines now try to give the searcher the most relevant results, not depending on exact matching of keywords. To succeed in this, search engines must find:
  • The real intention behind the search query and
  • Web pages where the best content regarding the query is located. 
Search engines are increasingly using machine learning to guess the search intention better. Machine learning is when the software 'learns' based on examples and earlier data and keeps improving its own capabilities. The search software are constantly getting better also in processing the way humans speak. So, it is logical to expect that search engines use these new technologies in finding the real meaning of queries as well as the real content in web pages.

semantic search seeks meaning

   

What is the real meaning or semantics of a search query?


So as to find the real intent behind a search query, search engines use many signals, including
  • Recent history of searches made from that browser.
  • Probable meaning of the search phrase going by the main keyword and qualifying/ additional expressions.
  • Whether variations or synonyms of the words in the query occur in the same article/ website.
  • What others have been searching for, using the same or similar search queries.
  • Location of the searcher. 
  • Formatting (e.g. captitalization, use of  'the' before a word).
  • Etc. etc. 
To visualize how a modern search engine works to find the intent of the searcher, let us take a simple example:

If I write ‘net’ on the Google search box and press 'enter', Google will try to find what type of ‘net’ I am looking for.

I actually typed ‘net’ on Google and found that it threw results relating to NET [the National Eligibility Test (NET) in which many people in my region right now are interested and must have searched for]; then came websites of some prominent bands offering net banking; then the Wikipedia page on Net; then .Net, Netflix, etc.

If I type 'How can I crack net?', it gives results only relating to NET.

Did you notice that Google did not give any result on different types of nets such as fishnet or mosquito net. Then I typed ‘net catch’ and, as expected, got all results on fishing nets. Google's machines know that when someone types 'catch' and 'net' together in the search box, he is wanting to know about nets used for catching fish etc. but when he types 'crack', he is looking for ways to pass the NET exam.

If I keep going to the fish net websites, in a few days my top search results for 'net' are likely to relate to fish net sites. And yet, because globally and in my location, people would search for NET or internet related websites, it is also likely that websites on fishing net might show up down on search pages.

Search queries have also become complex. Earlier, we used to type specific search terms - because it did not help writing a long sentence to search something. But now people get more specific with search because they know that search engines give more targeted results when we explain the search. In addition, more people now use virtual assistants instead of typing out things. Due to this, the use of voice search is rising - and voice search is all about making queries the way we speak to a friend. See these examples:
  • Earlier: we typed tomato or tomato puree preparation or tomato soup puree (results came according to what you typed: in the first case, about tomato as a plant or crop; second one on recipes; third one comparing the two).
  • Now: we type how to make tomato soup ( will give recipes in search results) or tomato puree (will likely give recipes and also local shop selling tomato puree) or where to get tomato puree (most results will be about online stores and nearby shops selling tomato puree) or tomato soup or puree (mixed articles, some with comparison between the two), or tomato (the results will depend upon whether you earlier searched for soup or plants, whether you have been making food orders from that browser, whether you have been searching for botanical names, etc).
  • Voice search: You will ask, Tell me where should I buy tomato puree? (very specific query as if asking a person, and very specific search results: will have a list of shops with location, open timings, even different packs and prices). 
  • If you say or type, Why the soup is not thick when I use tomato puree from McC, the modern search engine will have guessed your intent and will try to show web pages in which there is advice about using puree correctly to get thick soup, and not tomato plant etc.

Search index: this is where Google searches when you make a search query 


When you type a search query, the search engines does not sprint from your browser to the entire w.w.w. to locate the thing you are looking for. Rather, it has a huge index of web pages in which each web page is tagged with one or more phrases that searchers are likely to use.

So, when I searched for 'net' on Google, it thought that I was perhaps looking for NET test or net banking or .NET or NetFlix, and then served to me the best possible results from its indexes for NET, net banking, etc., especially those suiting to my location. But when I searched for 'crack net', it knew that I wanted to see web pages that advise about passing the NET test. If I used 'cracking NET', that would be one more signal that I was looking for NET and not a net of other types.

Google's index has many thousand web pages even for a highly specific query. These add up to millions of web pages for a generic query. For 'net' with the meaning of NET test, its index has 155 million web entities. So, out of these many web pages, Google must then intelligently choose the best ten results for the first search page, then the next ten results and so on.

One last word on indexing here. Google has not perfected search semantics yet, so it shows millions of results in response to a search query. However, the only top ones are important because the searcher hardly goes beyond the first few results. Therefore, Google takes extra care in choosing the top results, and only web pages with very good content and from established websites/ blogs are usually shown as the top results. Besides, when people click on search results, that gives signals for machine learning - so the results become more relevant/ focused in future.
   

Semantic search and modern SEO


As I shared above, search for 'NET examination' or 'cracking net' on Google ten years back would have shown those web pages on top, which had these words placed in the title, description and allwhere in the body. But Google today has a complex 'algorithm' or formula that judges each web page for:
  • signals of relevance (e.g. whether the web page discusses NET test in totality, with guidance on how to apply for test, old question papers, tips to solve the examination, etc?) 
  • signals of authority (e.g. whether the web page is well written, whether top websites in testing field link to that page, whether candidates speak highly of that web page, etc?)
These two can still be tweaked or manipulated but not to the extent that happened ten years back. That gives us valuable insight on what type of Search Engine Optimization or SEO is effective now on blogs/ websites.

For bringing a web page on top of search results, authority and relevance matter ned to be supported with SEO techniques. Yes, on page and off page SEO techniques work, but they are effective only when the content is valuable and relevant.

In the right type of SEO, such actions are taken that improve the visibility of web pages but do not fool search engines into erroneously believing that a web page is of high relevance and authority when it is not. Experts, including Google's own staff and its documentation on the web, suggest that the best SEO is the sum total of actions that are NOT directed towards search engines but are meant to help people in getting the right information or advice or help they seek through search. 
    

How to improve Google search ranking in times of semantic search?


Let us recall some points that we noted above:

(i) people are increasingly making more specific queries, 
(ii) Google and other search engines use tools of semantic search to find real intent behind a search,
(iii) search engines also use semantics to index web pages for relevant keywords and 
(iv) search engines rank high the content that helps the searcher find the information/ advice/ help he needs.

These give us enough insight to know what type of easy SEO actions we need to take on our blog/ website. Even if it is difficult to compete with the top Google results, it is not impossible either, if we write great content and apply right amount and quality of SEO. Let me convert this knowlege into practical action points:
  • Write content that suits one of these intents: buying, information, advice, help in solving a problem.
  • Write valuable content, not copy-pasted one. Develop authority on your subject.
  • Put keywords in your post's title, description and body. However, do not stuff blog's homepage, pages and posts with specific keywords. Rather, sprinkle all types of expressions around the topic - as it happens when we explain something to our friend/ colleague/ child. 
  • Use different expressions, phrases and synonyms of the main keyword for which you want the blog to be optimized.
  • Don't optimize the blog just for very broad keywords such as 'Indian food' or 'Chinese customs', and also for very narrow ones. Think of a range of broad, narrow and medium sized expressions that people would use in natural language while talking about that subject.
  • Try to give answers to questions that people ask about your subject.
  • Explain concepts.
  • Write detailed articles, at least once in a while, on main topics relating to your subject.
  • Do a bit of extra SEO: place the keyword (for which you want to rank high on Google) in title and search description, but do not use the same expression more than once in these two places. Especially in description, wtite natural sentences; if you feel like, use an additional similar expression. Use heading tags in text, alt attributes on images. Connect with reputed blogs/ websites/ social media entities so that they talk about you and link your blog on their websites. On these posts I have explained some very simple SEO techniques that work: 
  • Link your blog with web entities that relate to your subject: that will make Google understand what your blog actually relates to. If you review toys, your blog has links to toy manufacturers and you also have knowlege articles on toys, Google will index your web pages and mark them relevant to toys. If you sell toys and people make positive comments on social media about your shop and toys, your blog will come up high in local searches even if you have less content on the blog as compared to others' websites.
  • If you are into selling a product or service, you might need the help of an SEO expert because some additional tweaking is needed, which becomes a bit technical. Choose the SEO firm or expert carefully; if they claim to put you on the top of Google search pages and maintain your position on top, they are bluffing. That is what Google also cautions against. 
If you are new to search engine optimization, you might like to visit this detailed article answering many questions on SEO that come in the mind of bloggers.

No comments:

Post a Comment