Comments on: Word Searching as a Tool in the Study of Dreams, or, Dream Research in the Era of Big Data https://bulkeley.org/word-searching-tool-study-dreams/ Dream Research & Education Sat, 23 Jun 2012 20:33:55 +0000 hourly 1 https://wordpress.org/?v=6.4.3 By: Paul Duffy https://bulkeley.org/word-searching-tool-study-dreams/#comment-377 Sat, 23 Jun 2012 20:33:55 +0000 http://bulkeley.org/?p=2176#comment-377 Good presentation – especially on what could be seen as the weakness of the approach. A whole lot of work is done (easily, by computer) and you end up with a dozen or more facts about the dreamer, and 2 or 3 of these may turn out to be misses. People are never satisfied unless you get everything correct!

I’ve been working with co-occurrence of features (word lists or categories), trying to guess at very basic information about the dreamer; gender, age, sexual orientation. For the later I match sexual words and then expand by +/- 5 or 6 words around the matches. This gives me a ‘neighborhood’ text that I then search for two expressions that match generic character gender (he|him|his|man|men|boy|guy…) and (she|her|woman|women|…). This works very well for sexuality.

I also generalize the method by using two expressions for positively and negatively valenced words in the neighborhoods of a third, target expression. For instance, in the hvdc norms an expression for unknown males has overall negative valence for the male dreamers, but positive valance for the females. The valences are reversed when the target is unknown females.

I ran the same search with male vs female dreams from your Sleep and Dreams Database. The unknown males target showed negative valence for male dreamers, positive for females, but the unknown females target was positively valenced for both male and female dreamers in that set..

]]> By: Paul Duffy https://bulkeley.org/word-searching-tool-study-dreams/#comment-376 Sat, 23 Jun 2012 20:33:10 +0000 http://bulkeley.org/?p=2176#comment-376 Good presentation.- especially on what could be seen as the weakness of the approach. A whole lot of work is done (easily, by computer) and you end up with a dozen or more facts about the dreamer, and 2 or 3 of these may turn out to be misses. People are never satisfied unless you get everything correct!

I’ve been working with co-occurrence of features (word lists or categories), trying to guess at very basic information about the dreamer; gender, age, sexual orientation. For the later I match sexual words and then expand by +/- 5 or 6 words around the matches. This gives me a ‘neighborhood’ text that I then search for two expressions that match generic character gender (he|him|his|man|men|boy|guy…) and (she|her|woman|women|…). This works very well for sexuality.

I also generalize the method by using two expressions for positively and negatively valenced words in the neighborhoods of a third, target expression. For instance, in the hvdc norms an expression for unknown males has overall negative valence for the male dreamers, but positive valance for the females. The valences are reversed when the target is unknown females.

I ran the same search with male vs female dreams from your Sleep and Dreams Database. The unknown males target showed negative valence for male dreamers, positive for females, but the unknown females target was positively valenced for both male and female dreamers in that set..

]]>