bigger databases!

• `stamp collecting':
• collect a VERY LARGE AMOUNT of speech data
• remove any parts that are SIMILAR to some other part
• voice classification:
• intentional: mood/emotion/politeness
• unintentional: tiredness/health/age