This week I ran the example classes and am having an easier time doing so. Eclipse is making more sense and so is the programatic structure of our Java applications. This week I tested the strength of bayesian analysis by training our program with poetry from Bukowski and Edward de Vere.

As I understand it, the analysis doesn't discern 'good' from 'bad' rather only knows 'one' form 'another' based on what it is taught. Right? (somebody please clarify this for me if I've got it wrong). Hopefully we'll talk more about the statistical nature of how the code performs its analysis in class today.

Once the application was trained with Bukowski poems and poems from Edward de Vere (Edward de Vere ina n 18th century poet who I've never heard of but though 18th century poetry would be about the furthest thing from Bukowski), I fed the app poems from both authors. The app knew who's each text belong to each time. This type of analysis could potentially be useful for determining where text originated, and the authenticity of authorship.

Possible ares for improving how the analysis takes place include the ability to analyze structurer. It would be useful to know how many lines were there in each paragraph? What was the average word count per sentence? Quantitate measurements used for analysis would get a t the qualitative components of the text.

The modified example code for this week can be found here: BayesianSpam.zip. Also, Dan asked that we post our code from the first few weeks spent crying. So here it is: inPut_outPut.zip, paella.zip, regular_expression.zip.

You can read more about bayesian analysis on the A-Z blog.