Artificial intelligence (AI) - Machine Learning-Natural Language Processing

L

loukii

Member
Joined
Feb 18, 2009
Messages
22
Reaction score
0
Location
Johannesburg
Hi Guys if any of you have any experiance in : AI,Expert Systems , Natural Language Processing and Machine learning , please can you share your thoughts ,ideas and research here.

Thanks
 
I have experience in all of the above, but is there anything specific you'd like to discuss here? That's a mighty broad topic...
 
I have experience in Computational Intelligence and machine learning...Let me tell you, it's hard to research and develop a worthwhile commercial solution.
 
@fiaskow: It depends who you're targeting really - I worked at a company about 5 years ago that specialized in probabilistic solutions using Bayesian nets and we (successfully) developed bespoke solutions for the CSIR as well as local government. I'm also developing a neural net solution at my current company. If you can find the right target audience, your golden.
 
Hi Guys ,

thanks for he input. I am currently studying computer science with the end goal to specialise in robotics and part of my studies focuses on AI on a very very intermediate level at this point so I have been doing my on investigation re the topic at hand and find it extremely Fascinating regarind the various theories and stumbling blocks attached with each hypothesis. So to Zoom in to a specific example have a look at this AI chat bot being develeoped http://www.jeeney.com/ she is quite faciniating and the algorithms used to build Jeeneys learning abiltity is'nt too bad but I she isnt really self learning as the input which finally lands up in her database still needs to be cleaned up to make sense. My question is there another AI chat bot/interface which is capable of true machine learning using NLP ? can you recommend an approach i should take to start building the algorithms for an effective AI bot capable of self learning? I do realise this is like asking for directions to the fountain of youth but i hope to learn from your combined experiance , successes and failures in this area. My end goal would be to obviously apply this research to a robotic application.
 
@loukii: Again, that's an extremely broad topic with no definite answers. What is your experience with NLP libraries?
 
Have only researched this area as a key building block , once again quite a few libraries out there whith varying opinions. Whats your thoughts on this?
 
I suggest taking a look at the libraries out there - the one I used previously (can't recall the name) was quite effective at splitting a sentence into its constituent parts, labeling words as parts of speech and identifying the type of sentence (statement, question, etc). That's about all these libraries will do for you. Take a look at how words relate to one another (spatial vs time, etc); specifically prepositions and how the subject relates to the action.
 
Great will investigate thanks for pointing me in the right direction. What are your thoughts on AIML language , seems like an easy way out for chatter Bot developers out there , also no true machine learning ability as well.
 
I don't think it's worth investing much time in AIML tbh - although, I guess it all depends on how much time you have at your disposal :)
 
Start with a good book on pattern recognition.
 
@Keeper: Troll much? :)

@bin3: They're both based on AIML - I don't think it's worthwhile researching that particular avenue as it's not really AI.

@sn3rd: That would be a good starting point for the OP, but, again, that's an extremely broad topic. It sounds like he'd like to get more direction from the community. To that purpose I suggest OP first looks at NLP libraries to get a feel for what's out there and what they can do. Once he's done that, he can take a look at decision trees (built from the output of the NLP). My reason for DT's? They're extremely simple to implement (ID3) and understand.
 
kurtkz said:
@Keeper: Troll much? :)

@bin3: They're both based on AIML - I don't think it's worthwhile researching that particular avenue as it's not really AI.

@sn3rd: That would be a good starting point for the OP, but, again, that's an extremely broad topic. It sounds like he'd like to get more direction from the community. To that purpose I suggest OP first looks at NLP libraries to get a feel for what's out there and what they can do. Once he's done that, he can take a look at decision trees (built from the output of the NLP). My reason for DT's? They're extremely simple to implement (ID3) and understand.
Click to expand...

The thing is that it's really easy to get lost without some background on the topic. A good book will lay the foundation, and will certainly not be a waste.
 
kurtkz said:
@Keeper: Troll much? :)

@bin3: They're both based on AIML - I don't think it's worthwhile researching that particular avenue as it's not really AI.
Click to expand...

Agreed: but both are quite interesting chat bots. In the end, what is AI? I have always classified it into hard AI and soft AI, of which all known implementations at the moment would be soft AI, i.e. intelligent algorithms that mimic some aspect of intelligence, and hard AI would be the Sci Fi / Sky Net type implementations (though not necessarily that malicious)

For robotics, I would think a good starting point would be something in the line of neural networks and fuzzy logic, maybe some genetic algorithms or the like. Just trying to teach a robot to avoid hitting walls is quite an interesting final year project ...

To interface in a 'human' way I think might derail the robotics development quite a bit as you can quite easily spend a few lifetimes trying to get a natural language interface going.

But on that, some interesting robotic competitions as well: will try to find the like, but in a recent competition with the basic instruction of 'find the red apple on the table and bring it back to me' some bots did amazingly well.

Not sure if it was a Naked Scientist podcast or the like, but will try to find it.
 
Hey guys , thought I would share this with you . Go and check out IBM's latest shot at DeepQA - IBM Watson.


Specs:

IBM Watson is comprised of ninety IBM POWER 750 servers, 16 Terabytes of memory, and 4 Terabytes of clustered storage. This is enclosed in ten racks including the servers, networking, shared disk system, and cluster controllers. These ninety POWER 750 servers have four POWER7 processors, each with eight cores. IBM Watson has a total of 2880 POWER7 cores


http://www-03.ibm.com/innovation/us/watson/what-is-watson/index.html
 
I did not read all of youse peoples responses, however for my AI module in computer engineering, the prescribed reading material for neural nets was Widrow's publication on machine learning (google will help you find it), another interesting topic is hopfield nets in pattern recognition. And some basic probability theory knowledge wont hurt...
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X