One of the key components of semantic search is the end-user behavior. Our actions, the way we provide data, the way we search for data and what we then do with the data we do find are all key components of the complex construct between searchers, search and the information stored on the web.
Until very recently, when we looked for anything we had to use keywords and, the more adept amongst us, used keywords and search operators. Theoretically, the latter practice that instructs a search engine to include specific terms in its search and ignore others is designed to surface exactly the content we are looking for.
When content is surfaced as a statistical probability because Google does not understand what is really on the page (the traditional Boolean search approach) highly structured searches that look a little like this: “David Amerland intext:semantic search AND book” will return a search result that delivers the words “David Amerland” anywhere but which also includes the terms “semantic search” and “book” in the text of the page only.
As you can see from the Boolean search results and the screen capture above the outcome is that by searching like this I can surface material that contains my book and makes a mention of me anywhere and get 159,000 documents come up. Before we get to the implications of a search like this consider now the exact same search carried out using natural-ish language and the search query: “David Amerland's semantic search book”
As you can see from these semantic search query results the number of pages returned is significantly fewer and the order of the results in the first page of Google has now changed to show me content that more specifically mentions my book and where it can be found (this is significant). Of course the search term I have used is only a small improvement from my previous, very targeted, Boolean search.
If I want to go even more natural language in my search I specify an action by typing “Show me David Amerland’s Semantic Search Book”. The semantic search results on Google here are even more telling.
The number of pages returned with this search is the smallest data set yet. In each search I carried out I did two things, I increased the level of accuracy in the search results by defining my intent to Google’s search engine.
Boolean Search is Imprecise
Although my knowledge of how search works and Boolean search operators Google uses should have helped me find exactly what I wanted a search carried out in this way is an imprecise statistical one. I am second guessing that is on any particular page by projecting what I think will best reflect the results I want (in this case I want to find out a little about my book and where I can buy it). The first result on the page (Amazon) certainly does the job and Amazon methodically structures its product pages to answer statistical search queries like mine. The rest of the results on the first page also fulfill that task and in addition surface some reviews and even a PDF sample of the first chapter of my book but they also include a page of quotes taken from my book which statistically fulfills the request of the search but will not help me find out where to buy it.
This method of querying search has also surfaced an enormous number of other pages, asking me now to act as the filter that will validate (or not) the content of each of those pages.
The second search produced significantly fewer results. What is interesting is that the search query “David Amerland’s book” does not clearly define that I want to know where I can buy it. It indicates more of an interest to learn something about it and there the content that actually comes up is way more informational weighted with two video reviews of my book coming up from YouTube.
The third search query drills down to a specific intent: “show me David Amerland’s semantic search book” delivers the smallest data set yet. Interestingly the first three results are identical to my first search (showing my experience at structuring Boolean search queries and understanding of how product pages work) but after that they diverge and in the divergence we see the power of semantic search. Pretty much every link that appears on the first page here (like for instance Martin Shervington’s PlusYourBusiness.com page) contains a link to Amazon where you can buy the book or information on where you can obtain it. This is a really clever (and precise) match up between intent and web pages and it outclassed Boolean search operators.
Enter Google Voice Search for Chrome
The reason the above experiment is important is because search is changing drastically and we need to change with it. Every time we carry out a keyword search or a search using search operators we send a signal to Google about how we search that affects how the search engine can understand intent. In a manner of speaking we are confusing its brain.
Ideally what Google wants is all of us to start using natural language when we search. But as my second search query showed this is hard to do when you’re typing. Condition by keywords still my second query was better but it was also some way from being “natural”. As a result I got results that were still imprecise. What’s more my intent here, while still the same (for me) was way harder to read for Google hence the return of results that were not quite as precise as some of the ones I got with the Boolean, structured search I did first.
The third search I typed was what I would most probably speak into my phone and the results were also more satisfying in that I was able, in almost every case. To find out directly where I could buy a copy of Google Semantic Search. By introducing Google Voice Search App for Chrome Google is beginning to unify the end-user behaviour we employ in mobile search (when we talk to our phones) and desktop search.
This is a really big move. Here are the implications:
- Better search results. As semantic search is scaling we will start to move further and further away from the massive data sets of hundreds of thousands and millions of pages that contained what we were looking for and of which we barely saw anything listed on the first page. With smaller data sets the precision of search results goes up and (as we become accustomed to natural language on desktop) our ability to signal intent will also improve.
- Accelerated semantic development. The building of Google’s semantic index needs our data. The way we search, what we do with the results we find and how we then search next, are all part of a complex picture that helps Google understand the quality of its product and work to improve it.
- Preparation for Google Glass. Google Glass is coming next year. The “OK Glass” command activation interface is going to have way less resistance at the threshold entry barrier if we have already become accustomed to saying “OK Google” each time we want to activate desktop Voice Search.
- Establishing a relationship with Google. This may sound a little odd but the emotional connection between a searcher and Google becomes a lot closer if, as I multi-task at my desk on a busy day, turn round and carry out a voice search using the “OK Google” command on my desktop and, while still doing other tasks, have the answer read back to me by my computer. I start to think of Google then as less of a tool I use and more of a personal assistant. This is of immense value to the Google brand.
- Preparation for even more disruptive developments. Semantic search that produces great results on a screen is only one part of all this. Screenless and keyboardless computers are coming and right now the idea of interacting with one is a little freaky. But consider here that by using Google Voice Search on desktop the way I detailed it in point 4 I have made use of neither keyboard nor screen and I am feeling kinda happy because my productivity has gone up as I am able to do other tasks while still searching. Given enough time with this, it’s only a small step before I move away from my desk, on the move, and my next, wearable device whispers answers to my ear in response to voiced search queries.
Checkout Google’s video of the Google Voice Search App for desktop and tell me you’re not thinking already what I am thinking: The future is here.
What You Missed
Google’s Semantic Analysis of a Spam Website
Semantic Connections in Google Plus and your SEO
Google’s Hummingbird Deepens Semantic Search Results
Semantic Search and What it Means to Businesses
How Semantic Search is Changing End-User Behavior