Okay, so this might seem a somewhat redundant news post (no work has gone on for oooh, about 3 years), but Actor is no longer active. Since I last worked on, I've converted to Linux, stopped using Windows, had my VB setup largely broken and kindof had other stuff to do (like earning a living).
But there's a small chance that one day I might rewrite and develop it again on Linux (though I think there are already a couple of implementations on Linux of this kind of thing around on Sourceforge). But I wouldn't advise you to hold your breath :-)... read more
This is a minor point release of Actor, an Eliza type program which learns solely from what users type (say) to it. It still doesn't make much sense (it helps to picture it as a demented, drunk Yoda if you're speaking to it) but this version does see the introduction of 'contextual exclusion', whereby long common words (e.g. 'thought', 'example') are not regarded as contextual candidates for the program to store.... read more
This project is being actively updated, despite no CVS stats being shown (the statistics system doesn't seem to be working properly atm). Currently, I'm working on improving the conversation contextualisation and hope to have a new version out within the month.
If anybody does have any good suggestions (excluding using a neural network, unless you can tell me in detail how to construct a working one for this application :-) please give me an e-mail.... read more
I've finally got round to learning (well, learning the basics, at least) how to use CVS under Windows. So if you want the very latest (possibly broken) source, you should be able to download it from the CVS tree and compile it if you have a copy of VB.
If anybody tries this, could you please let me know. I think everything should work correctly (since I've actually manually edited the VB files to make all filename references relative rather than absolute).... read more
This is an ongoing project to try and make a program which comes as close as possible to passing the Turing test. In other words, the eventual aim is to try and produce a program which you couldn't tell apart from a human over the Internet until you'd spoken, ooh, at least 3 sentences to it.
This might sound easy but really it isn't, especially as the architecture eschews the 'take a randomish word from the user's input and insert it into some generic sentence such as 'How do you feel about <word>?' in favour a much more node based approach.... read more