I have an idea for a chatbot for a while. It is based on neural networks. It would have a generating and a learning side. First, it would generate some random output. Then it would learn from input response form a user, just to know what to generate in a mirror condition when user enters something on her own in another occasion. It could be based on TensorFlow NN library and it could be trained on Reddit.
SMZ, how does this approach differ from METAQUID?
I have an idea for a chatbot for a while. It is based on neural networks. It would have a generating and a learning side. First, it would generate some random output. Then it would learn from input response form a user, just to know what to generate in a mirror condition when user enters something on her own in another occasion. It could be based on TensorFlow NN library and it could be trained on Reddit.
SMZ, how does this approach differ from METAQUID?
This is how Cleverbot works: https://www.cleverbot.com/ If it gets input it doesn't recognose, it presents it back to a different user and learns how to respond if someone asks the same question again. Unfortunately, this means it gets quickly filled up with nonsense and the botmaster spends most of his time removing unwanted replies.
I have an idea for a chatbot for a while. It is based on neural networks. It would have a generating and a learning side. First, it would generate some random output. Then it would learn from input response form a user, just to know what to generate in a mirror condition when user enters something on her own in another occasion. It could be based on TensorFlow NN library and it could be trained on Reddit.
SMZ, how does this approach differ from METAQUID?
That style of chatbot is cool, ive never actually seen one go exactly like that.
But it has to have a metric of its success, so what would it be. "haha" ?
I want to know too :)
Some possible inputs:
If I need a boat
That has a red bow
And is less than 10,000$
But was owned by an old person
And never used more than 5 times
Which webpage sells exactly this item?
If a rabbit was sitting in a mountain and is now sitting in a cave, why did it change its location?
I have an idea for a chatbot for a while. It is based on neural networks. It would have a generating and a learning side. First, it would generate some random output. Then it would learn from input response form a user, just to know what to generate in a mirror condition when user enters something on her own in another occasion. It could be based on TensorFlow NN library and it could be trained on Reddit.
SMZ, how does this approach differ from METAQUID?
This is how Cleverbot works: https://www.cleverbot.com/ If it gets input it doesn't recognose, it presents it back to a different user and learns how to respond if someone asks the same question again. Unfortunately, this means it gets quickly filled up with nonsense and the botmaster spends most of his time removing unwanted replies.
I wonder how would it perform with neural nets platform. Think of it as OpenAI GPT, but implemented as a chatbot. I think it could show a great deal of intelligence, caring not only about the last response, but having insight into the whole conversation history.
… The interface design and theme is attractive, and engaging. Hanah is good entertainment. ...