In Fernando Pereira’s latest take on Powerset, he eludes to a similar nonmonotonic function for natural language search. Conversing efficiently in natural language requires that you can predict, to some degree of accuracy, what will be understood by your conversation partner. Systems that get uncannily close to human understanding yet fail in cases that are only predictable to the system designers may be frustrating to use. How will the tradeoff between increasing utility and increasing inscrutability pan out? Where is Powerset going to be on this hypothetical graph?

As Fernando says, only testing on the final system will tell.

Before watching that the phrase “global warming propaganda” would have smacked of blinkered Republican rhetoric. As is pointed out in the film, there is a great global warming juggernaut brimming with funding for scientists that is very hard to slow down now. Even Bush conceded! (Wouldn’t it be funny if his stance on man-made global warming redeemed him in the end.) I think this is a very important film to legitimize the other side of the debate. I wonder is it having much effect?

I don’t know about you, but having a previously certain belief shattered by what looks like good science gives me a really good feeling. This science stuff is pretty awesome.

After more reading around, it seems the documentary was a bit misleading, including theories which have been discussed in the literature and found to have problems, and not including some theories that have been put forward to explain the data presented. Check out this detailed response. Nevertheless it’s good to have some counterpoint to Al Gore. Now maybe we can meet in the middle?

I’m not going to go into details but user-friendly it most certainly is not! We all know this could be done so much better. Where are the web 2.0 startups doing 3-click tax returns?

I recommend Wikipedia to explain it in plain English.

Have you ever wondered why programming languages that have good regular expression support (Perl, Javascript, Ruby, etc.) nevertheless omit syntax for negation or conjunction within regexes? I’m sure there must be a good reason for this. Has it been found that people just don’t intuitively get what negation does to a regex? Are language designers unwilling to complicate their syntax? Are they avoiding the processing it takes to complement or intersect finite state machines? They do compile to finite state machines, right?

Let me give you an example of where negation would be useful. Say I’m scanning through some text for some sub-sections which have been explicitly quoted. If the delimiting characters are angle brackets then I can write /<([^>]*)>/. The bit in the middle [^>] matches any character which is not a ‘>‘. This is a limited form of negation on a single character. Taken as a whole the expression says: Find me the strings of non-’>‘ characters which are surrounded by a ‘<' on the left and a '>' on the right. So far so good.

But what if the sub-sections I'm interested in are bounded by multi-character strings? '<<<' and '>>>' for argument's sake. In analogy to the example above I want to be able to write /<<<(?^.*>>>.*)>>>/. I've invented some syntax for negation here: (?^R) means match anything that the regex R doesn't match. The R in my example (.*>>>.*) matches any string that contains the sequence '>>>' so the expression as a whole says: Find me the strings that do not contain '>>>' that are bordered by '<<<' on the left and '>>>' on the right. Now, in this example, the same can be accomplished with a "lazy star" (if your programming language supports that) which just makes the regex non-greedy within the bounding strings: /<<<(.*?)>>>/. But I would much rather state the patterns I'm looking for declaratively than change the ambiguity-resolution mechanism of the processor. Why can't I?

Any finite-state toolkit with regex support has syntax for negation and conjunction. Do you know of any programming languages that do? If not, why not?