Style quirks

Style, usage and other quirks of written expression might be useful in profiling a dreamer.
I'm currently building at a collection of usage features that I'm are identifiable using regular expressions.

Past vs present tense
Use of contractions
Emoticons
Repeated letters like "Ooooooh"
Repeated punctuation like "!!!???"
All capitalized words
Use of Latin: "etc, i.e., et all"
And a few grammatical errors

A number of these boil down to formal versus informal usage. Some quirks are, I think, rules that the dreamer internalized that could reflect social background, age, education, etc.

Use of present tense tends to increase with age.
Someone who uses two hyphens for a dash or puts two spaces at the end of a sentence, probably learned to write on a typewriter.
Use of emoticons, all uppercase or repeated sequences of letters or punctuation point to familiarity with the internet.

I'm thinking about a two-way scale for formality-informality/modern-old fashioned. A negative evaluation of these scales suggests terms like stuffy, fussy, sloppy, lazy, maturity, childishness, conformity, anarchy, etc.

Most of the expressions are paired so the scales could be the ratio of each match. Then sum these for pairs of expressions of the same scale.

Fictive content in dreams

Fictive elements are dream content which don’t correspond with literal waking experience. When making predictions about a dreamer, under the assumption of the continuity hypothesis, fictive content needs to be excluded. An obvious example of fictive content is flying: the dreamer reports of floating or flying in the air without mechanical assistance. There are a few other typical dream themes which we can probably identify as fictive. More difficult to identify are content that isn’t obviously “fantastic”.

One place where I think I’ve had some success is in identifying fictive family members. In the previous post I used an expression that matches most of a dreamer’s family members:

(my (\w+ )?(mother|father|sister|brother|daughter|son|wife|husband|...))

The sub-expression “(\w+ )?” optionally matches a word followed by a space. It helps to match text like; “my eldest daughter”, “my older brother”, “my ex wife”.

How do you tell if a report of a family member is fictive? In the case of a young dreamer, reports of “my (son|daughter|husband|wife)” could be excluded, except that we might be wrong about the dreamer’s age. We could also look for quotes around the family name, or phrases like “supposed to be” preceding the match. Another idea comes from the observation that fictive family member are relatively rare compared to real family members. A dreamer who mentions “my son” 3 times but mentions other family members at a much higher rate, probably does not have a son. In the Predictions search I use a “fictive threshold” for reported family members.

fictive_threshold = max(brother_count+sister_count, son_count+daughter_count,         mother_count+father_count)/15;

If the count for a family member match is lower than this threshold, or less than 3, it is considered fictive.

Of course, a dreamer’s life situation is subject to change. The dreamer may marry or have a child - after recording years of dreams. In that case real spouses and children may be marked as fictive.

Although this fictive threshold trick seems to work for family members, I’ve yet to find another category where I can apply a similar technique.

Predicting something about a dreamer

Elements

The Elements search is is run automatically after you enter a dream set via the “Select set” button. 

For this demonstration I choose the “Lawrence” set available at Kelly Bulkeley’s Sleep and Dream Database (http://sleepanddreamdatabase.org:5000/dream/).

The result is a periodic table display of the 100 element categories. The search compares the matches for each category against the average results for 50 sets of dreams - referred to as the baseline results. Categories that aren't very different from the baseline are greyed out. Categories that are significantly more frequent are outlined in red. Significantly less frequent categories are outlined in blue.

If the icons representing each element aren't obvious you can hold your mouse over them to get a preview of the matches. Clicking on an icon runs a search for the category.

From the higher than baseline results we can guess that Lawrence is involved in the arts, is British (the grey icon of the man with a top hat) and that he drinks a bit (the blue cocktail glass icon) and that he works in an office setting (cubical icon).

British-isms matched in the Lawrence set

• centre
• maths
• my flat
• orientated
• their flat
• to hospital
• theatre

Also, the top match in the Places search (map icon at the bottom of column 5) is “London”.

The star icon (upper left) represents famous names who Lawrence dreams of or mentions.

• Neil Young
• Nick Cave
• Francis Bacon
• Jimi Hendrix
• Elton John
• Debbie Harry
• Charles Darwin
• Stephen Fry
• Elvis Presley
• George W Bush

Under-reported in Lawrence’s set are Vision (eye) and Color (color wheel), female character identifiers (pink silhouette). Also seen are low Exercise, Thinking, Talking, Dogs, Reading, Body references and a very low reference to Self. 

Gender 

(Refer to the first column of the Elements result.)

How do we know that Lawrence is male (other than his male name)? The Elements search uses two methods for guessing the dreamer’s gender.Look for mentions of male versus female self references. These include body parts, clothing, etc. A little less obvious are matches for declarations of gender such as “I was a man”. These wouldn't be worth mentioning unless they contradict waking reality. So if a dreamer says “I was a man” we can assume the dreamer is a woman. 

The first 2 icons (pink lips and mustache) show results for these categories. In Lawrence’s set there are no matches for either, so we can’t rely on this method to determine Lawrence’s gender.

The second method for guessing gender relies on results from the baseline sets. I did searches for the most commonly occurring words in male versus female sets. Any word matches which were significantly different between the male and female sets noted. The difference between the male and female average percent matches is used to set a threshold for that word’s “vote” for male versus female. The size of difference is used as a weight to multiply each vote. The sum of the weighted vote is the second gender guess. (A similar procedure was used with paired young versus old dreamers to guess the age of an unknown dreamer.)

The second gender test guesses that Lawrence is male, and also that he’s in his twenties.

Sexual orientation

I do a simple search for matches for sex which includes the words immediately surrounding the matches. I then search these matches for male versus female character matches. The results guesses that Lawrence’s sexual preference is for females.

Family

A fairly straightforward search for “my (mother|father|sister|brother|duaghter|son|...)” gives most of the information for the family search. There are a few exceptions. “My mother” is rather formal usage. A dreamer is more likely to say “mom” or ‘dad”. If you want to capture the names of family members or other information and expression that optionally matches capitalized words (names) near the Family matches is used. The search must also check for fictive family members that only occur in dreams. 

Lawrence's result shows a daughter but no wife. He could be widowed or simply not married.

A search for “(my (lover|partner|significant other|girlfriend|fiance|ex.?wife|late wife))” returns 52 matches for “my partner” and nothing else, so Lawrence is unmarried but in a relationship.

Lawrence has a brother (or brothers) but no sisters. Also note the Lawrence’s father is mentioned almost two times as often as his mother, which may indicate his mother is deceased.

Note that Lawrence never mention the names of his partner, daughter or brother. This is likely to protect their privacy.

Lawrence prefers the company of males

The low result for female character identifiers is interesting. Looking it the Names results shows that the top 10 names are all males. Lawrence lives in a ‘man’s world’. 

• Ralph
• Terry
• Sal
• Toby
• Reece
• Ron
• Tyler
• Tim
• Karl
• Juan
• Claire
• Kate
• Len
• Catherine
• Saul
• Trisha
• Kevin
• Adrian

Conclusion

We can gather some basic information about a dreamer from their dreams. In the case of anonymous dream sets we can't check that all of our guesses are correct. 

REve (js)

Profile a dreamer from a set of dreams.
REve (js) is a web app I made that tries to guess something about a dreamer from their dream log.

You paste in a set of dreams and then run searches for various content categories to get predictions about the dreamer's waking life experience. This blog is a place for me to post more detailed descriptions of what I'm working on and the kind of results or predictions that REve can make.
 

http://gdriv.es/reve