Will an IBM Watson-Wolfram|Alpha collaboration advance Natural language processing ?

Watson is fundamentally different from Wolfram. Wolfram Alpha is in a way similar to AIs lime Siri. It uses curated, structured data to answer questions.Watson technologies are built around the idea of using natural language and deep learning to build knowledge graphs automatically, then process queries in natural language in a manner that would allow AIs to answer them based on the built knowledge graph. As Richard Benjamins noted below, our approach doesn't care much for a deep understanding of the data, although the line can get a bit blurry. How do you define "understanding"? If understanding is the ability to explain things and to reason based on their properties, then some of the Watson Solutions can do exactly that.I think very soon, we will get to a point where there will be no difference between symbolic and non-symbolic AIs.  Clever deep learning algorithms make it possible to - at least with a specific domain in mind - build software capable of learning common sense not only throug specifically stated data, but also based on behavior of users, thus eliminating the huge issue of imolicit information  has been a major challenge for symbolic AIs in the past.As for your original question: I'm not sure what either of the parties would gain from a partnership. IBM is a company with a long history of natural language processing (from mid 70s), a huge amount of experts in the Watson group and cooperations with leading scientists in universities across the world. So technology-wise, there are few things that Wolfram  Research would have to offer. The other way round might be the same. Stephen Wolfram always had a very clear idea of what he was going for and never really let the rest of the market influence his design decisions.

Few people here noted correctly that IBM Watson and Wolfram|Alpha are quite different approaches. I would recommend reading an article that compares these two amazing achievements, written by Stephen Wolfram, the founder of Wolfram|Alpha:http://blog.stephenwolfram.com/2011/01/jeopardy-ibm-and-wolframalpha/In that article you can find Stephen's quote: "perhaps with something like IBM’s Jeopardy approach it’ll be possible to get a good supply of probabilistic candidate data answers—that can themselves be used as fodder for the whole Wolfram|Alpha computational knowledge engine system.". I personally think that the real tendency for algorithms and data is integration. It is sometimes noted in http://www.pnas.org/content/97/23/12926.full. Many things are achieved by mixing up various approaches, algorithms, knowledge domains, data, and new things emerge on intersections of interdisciplinary sciences and technology innovation stems from technology blending. For more details read the article where this nice infographics is from:

In the early days of AI there was a distinction between symbolic AI and non-symbolic AI. In other words: logical reasoning, Knowledge-based reasoning versus statistics, neural networks. It seems Wolfram Alpha is like symbolic AI ("understanding" the domain) and Watson is more like a more non-symbolic approach. I do think, however, that today the boundary between those two approaches is blurring, and the non-symbolic approach is more successful (in applications) than the symbolic one.