Science Buddies: "Ask an Expert"

Hi !!! the science project idea of "Computer Sleuth: Identification by Text Analysis" has fired my imagination. I am thinking to make a project on this to identify authors of ancient Sanskrit literature. I asked my father he suggested that using an artificial intelligence technique like Artificial Neural Network (ANN) would be more appropriate than suggested Principal Component Analysis or statistical methods.
Can I use ANN ? Will it be correct ?
I am 13 year old girl.

I asked the same question earlier but I am not sure why it was not posted here.

Thank you for patiently reading.

Niharikka

Niharikka,

Both ANN and PCA are numerical methods, so you would have to convert your inputs and outputs into numbers, usually between 0 and 1. Have you used these before? Do you know how to write programs, and if so, in what language? There are quite a few open source programs for both ANN and PCA, but they require you to write small programs to use them. It may be that the methods described in https://www.sciencebuddies.org/science- ... p022.shtml can be adapted to use ANNs or PCA. It seems a bit advanced, though.

Also both methods work best when you have lots of training data, especially ANNs. PCA is probably easier to use and apply than ANNs. ANNs require more data and can stagnate before reaching a good solution, in which case you either have to do some clever massaging, or try again (sometimes repeatedly).

How familiar are you with these methods? Have you identified a specific piece of software to use? Do you have your example and test texts in digital form?