Notes on Rich's presentation

Notes on Rich’s presentation The chart of the gender gap would be more effective if it also represented the magnitude of people in the field

Alvarado & Dodds, 2010

GENDER CONSTRUCT (perforamtive?) —> Be careful not to impute intrinsic gendering to cultural themes (“war” -> maculine) —> When you show the materials, discuss whether you claim them to be gendered. 

You say it’s not appropriate for kids: Maybe you could instead suggest modifications that seem promising (from a methods perspective)

Maybe a benevolent Hawthornw effect: changes the culture. 

No boys: all girls. (gender works pretty differently at the school)

\section{Improving feature extraction}Table 1 reports the results of normalized and non-normalized feature extraction for each of the provided training sets. Comparing the normalized and non-normalized error classes, I notice that a higher percentage of the errors after normalization are substitutions. This is consistent with normalization as a process which r emoves background sound, w \begin{tabular}{rrrrr} \hlinetraining set &non-norm WER &non-norm SER &norm WER &norm SER \ \hline1 &12.34 &23.47 &9.87 &20.64 \2 &8.63 &18.64 &7.14 &16.03 \3 &2.4 &6.91 &1.99 &5.86 \4 &2.08 &5.95 &1.58 &4.54 \ \hline \end{tabular}