In 30 years from now, what kind of natural language processing (NLP) technologies will be most important? And why?

30 years -- that's an eternity in the world of technology. Let's talk about a shorter time interval. Being 2013 as I write, What will happen by 2028? By 2028, NLP systems will be taken for granted in all of the following: providing accurate real time translations generating synopsis for novels and other forms of long writing grading college level essays and providing insightful feedback deducing conversational context from overhearing a conversation and interrupting said conversation with useful information (e.g. whether in presidential debates, or in the home) passing the Turing Test with flying colors understanding human mood and effectively empathizing, thereby serving as a primitive therapist How could all this be possible? 1. Computers will be immensely faster I write this answer in Jan 2013; from a very conservative application of Moore's law alone, it is reasonable to project that we will effectively more than 100 times the current compute power. (Areas of breakthroughs mediating this hardware improvement may include optical computing, 3D IC, and severely limited forms of quantum computing.) 2. Crowd sourcing and machine learning Both fields have been around for a while -- but they are are only starting to be applied industrially everywhere. Since natural language is such a basic task for humans, NLP data collection for applied machine learning will be readily feasible. (For example, I am sure that the answers on Quora will be analyzed closely by machines, in much more detail than how they currently are.) startups like Coursera/edX/Udacity will also lead to a variety of exciting new opportunities in this area. In my answer, I have only considered a small number of current events that will advance NLP research; I feel that my answer has been extremely conservative on the time frame -- we live in exciting times.

I believe that the most prominent technology at that time will be a stronger framework of AI. In 30 years from now, a new computational theory may be discovered. This theory will be a new fundamental which replaces Turing Machine and tackles the NP-hard problem class. Many AI/NLP problems will become solvable instead of only being approximated, unleashing AI/NLP scientists and computational linguists from traditional Markov Random Field models. This makes me believe that AI will still be the most important in the future.

I believe that within 30 years, machines will routinely pass the Turing test, which involves machine understanding, reasoning, language generation, and the ability to carry on a conversation.  Machine translation (among human languages) will also have been solved.  All those technologies will remain important.  The new thing they will enable in the 30-year-from-now timeframe is, let's call it, NLP bionics, essentially embedded assistive language technologies, in particular, language analysis, information enrichment, and translation. Think "The Six Million Dollar Man" with the ability to converse in Italian, Thai, and 400 other languages in addition to (just) the ability to run 60 MPH and lift a car.

I am curious if we'll see close and fruitful cooperation with neuroscience. Computational linguistics would very probably benefit a lot from understanding how human brain works with language -- there is so much that must be going on but we know so little! I don't see many breakthroughs here, turns out it is rather difficult to study the brain, but 30 years is a long time, so we could get somewhere by then.

Read any book on good sales techniques, cold calling, sales closing etc. This will be the area of Natural Language Processing interests in three decades time.

In 30 years, I will need an automatic translator to tell me what the heck my grandchildren is talking about. I'm going to be behind the curve on everything, so any NLP technology that can take not just context but the background and knowledge of the listener into account, would be of enormous value.

Within 30 years... pragma-polyphonic NLP helps us, together with Google´s omnipresence, filter the polyphonic layers of massive multispeaker discourse clusters in a way that not only enables us to expand the limits set up by such things as intentionality, but also enables us to easily understand the pragmatic meaning of these clusters. The other side of this technology enables some of us to have a simultaneous multi-presence in these clusters and speak (using sock puppets) with thousands of dissonant, but pragma-polyphonically ordered, voices, in a way that will make Arnold Schoenberg turn in his grave.