Posted January 15, 2011 12:05 am by with 11 comments

Tweet about this on TwitterShare on LinkedInShare on Google+Share on FacebookBuffer this page

Search engines have historically been the best way to find information on the internet. But what if we didn’t need search engines anymore? What if, all we had to do was verbally tell the computer what we needed and it gave us the best response? All without visiting a web page, navigating through ads or waiting for our browsers to load.

Well, computer scientists at IBM have almost made the above scenario a reality. With their new natural language processing (NLP) super computer named “Watson”, they are breaking new ground that could potentially create a radical shift in information retrieval (IR).

Now, many of you already know that I am fascinated with NLP. In fact one of the first web applications I ever created is highly dependent on NLP. I really think NLP could play a very significant role in the future of IR.

But, it’s important to state that we are a long way away from seeing the type of NLP for public use that is needed to have a successful product. Google already admits that it is using some form of NLP for sentiment analysis, but most are skeptical of how much they use it, and how strong of a ranking factor it is.

So, you might be asking yourself, why the heck is Joe talking about all this?

I think it is extremely important that we as marketers continue to look forward at how information is exchanged. A few years ago if you would have asked most online marketers if the popularity of content on Facebook was important to SEO, most would have laughed you out the room; now it is a reality.

So, my question to you is, if a NLP based information retrieval system one day takes the place of search engines, how will that change the way that marketers promote content online? Instead of SEO, will we then optimize information? And, if so, how?

  • The flaw in NLP is always that the computer has no way of reading the intent of the person asking the question. Jeopardy is simply Q and A, I’m guessing that ChaCha could do a great job in that battle… perhaps better than Watson.

    With search, my question is only an extension of the information I’m seeking. In other words – the actual question isn’t always the question. If I ask, please provide me a list of email providers that work in the enterprise space… I’m only asking a base question and I need to dive in deeper once I see the list. So – it’s an answer – but not the REAL answer.

    In my opinion, NLP really only replaces the text box on Google. The relevance of the responses still has to be filtered based on the person asking the question.

  • good video.

  • In your future vision I think NLP + IR & search engines all still play a role and the front end user interface, however transparent, would still involve back end processes not unlike or further derived from modern search engines, albeit more ‘on-the-fly’ and akin to how web analytics are processed in real time.

  • Maybe it is just my failure, but I completely fail to understand how with or without the use of NLP that one can consider this something other than a “search engine”. The importance of indexed content does not change whether it is retrieved by voice or by text. One of the most important aspects of NLP is the bridge between the two, thus there should be no significant value changes. For people who do not optimize for regular language patterns already, I suppose this would be a concern. Those are also typically the sort who consider meta tags to be SEO, and think that traffic alone is marketing.

    In a feasible iteration of this technology, I would expect that it will use source citations. If this is the case, I will expect my relevant content to still be available and valuable.

  • I see NLP as a means for empowering information retrieval. The next stage of information retrieval, in my view, is context based – retrieval based on relevance to me. So if 2 people search Beethoven, the search engine will know,that one is looking for the composer related results and the other is looking for the movie about the St. Bernard dog related info, all based on the holistic context of that person.
    Imagine a web 3.0 environment that has a universal contextification layer, based on advanced algorithms with NLP elements that objectify subjective contexts. Such a layer will create a scale on which ad-hoc semantic maps of topics and people will be drawn and proximity=relevance defined. In such environments distributiion of content will also be personalized based on explicit and implicit context maps and emower truly personalized relevent distribution.

  • Good post. I share the point of view of Rahav!

  • Nice video. When I look back what the last ten years brought us in software engineering I’m very curious what the next ten years will bring.

    Kinesiology – your chiropractor and your massage therapist may be able to read the subtle changes in your muscle energy to figure out what is going on in your body or even help you release negative emotions (emotional soma relationship) much like a lie detector test works. We are all just energy and the technology is there to read that energy. So why not use it in searches?
    I have a real estate company and a communication company but I have been using kinesiology for years (also licensed massage therapist) in every aspect of my life to unbury and deal with emotions and fears that may be holding me back. I never though it could be used for search engines, but it make sense! I would love to talk to anyone about this online or offline. Robin Speronis, 239-233-9659

  • I look forward to the day, albeit far off, when a search competitor will show up for Google. Unfortunately, the lack of competition has given Google an obviously weakened focus in delivering quality results (see the Marketing Pilgrim post “Google Bowing to Pressure…”). If only everyone in the search space started using Bing, even if just to scare Google… But then we’d probably just end up with 2 lackluster search giants.

  • I am greatly thankfull to you for this nice post. It was very interesting and helped me with my own project I am doing. eAnswerNetwork (I) Pvt.Ltd. is a Company which organize the Search Engine Optimization and Web designing.

  • I also agree with Rahav. To retrieve the exact info. , more detailed input will be needed to be processed. But of course, we, humans will be able to adapt.