Save: Save the current search terms to the last selected menu item.
Expand: enter one or more words or phrases and find the words most commonly associated with it.
Lemmatize: remove words which have the same stem form the list.
|
I am no longer working on Bag O' Dreams. Instead, please try my Bag O' Regex project which has most of the same functionality of Bag O' Dreams. Bag O' Regex is available at: http://kitchen-sync.com/bag/BOR.html. |
Bag O' Dreams is an application for the exploration of dreams. Bag O' Dreams allows you to search dreams for themes and content using information retrieval techniques. You can enter a list of terms and phrases associated with a concept or theme and Bag O' Dreams will automatically identify the dreams that are most likely to be relevant to it. You can also start from a particular dream and find other dreams that closely match it in content or generate a list of common themes contained in the dream. Starting at the level of individual words, you can discover which other terms are commonly associated with it, how often it occurs and display it in the context of your dreams. Starting from a collection of dreams, Bag O' Dreams can automatically partition these into sets of dreams that are thematically related.
Bag O' Dreams is intended for anyone interested in the study of dream content.
Bag O' Dreams is a Java application that requires JavaTM 2 Runtime Environment 1.5 (or higher) to work. (See the Installation section below.)
Important: Bag O'Dreams requires version 1.5 (or higher) of the Java Runtime Environment to work. Download this from Sun at the link above.
The Settings panel is where you add a set of dream documents that you'ld like to study. Settings control how terms are added to the term table. Usually you'll want to exclude words that only occur in a handful of dreams and words that occur in almost all dreams. What remains are the words which carry most of the dreams' content.
to navigate to the folder where you
keep a set of dreams. Click on any dream in the folder and the whole set of dreams is
added to Bag O' Dreams. The dreams must be .txt files with one dream per file. Once
added, you will be able to run any of Bag O' Dreams' reports on the set of dreams.
When you only have one dream from a dreamer and want to quickly scan it for themes, add the dream to an existing set. Then you can run various reports on it - especially the "Interpretation" report. The "host" dream set can be your own dreams or any collection of dreams which you'd like to compare the "guest" dream to. The guest dream button is also helpful when you've already processed a large set of dreams and only want to add a dream which you've just finished recording.
The guest dream button does nothing if you haven't already added a set of dreams.
Newly added dreams appear at the bottom of the document selector list.
You can also add a single dream to an existing set by saving it in the same folder as the set. When you do this Bag O' Dreams will ask you if you want to add it to the current set of dreams.
The Search panel provides many methods for finding text and identifying themes in your dreams. The Search Menus store lists of terms and phrases which identify specific content in your dream set. When you select an item from a search menu, like Driving, a list of related words is entered into the text field. More details here on editing the Search menu lists. Reports which use search terms are described in this list;
A second drop down selector lists reports that give global information about the set of dreams and which don't depend on search terms;
A third lists reports that give global words used in the set of dreams;
A forth one lists the drop-in modules and a few utility functions;
[*These are optional stand-alone applications that only show up if they are put in the "modules" folder.]
The small combo next to the search terms field provides a few commands that effect the search terms.
Save: Save the current search terms to the last selected menu item.
Expand: enter one or more words or phrases and find the words most commonly associated with it.
Lemmatize: remove words which have the same stem form the list.
To create a new dream document click on the "new" button: 
.
To open a dream via the standard file dialog, click on the "open" button on the Search panel: 
.
You can also use the document selector in the middle of the search panel to open a dream for editing or analysis. When you select a dream from this list it opens in its own window. You can edit the dream and run the following reports from the dream's Search menu;
In addition, you can do simple text searches inside the dream with the Find command (Typing "control-f" also works.).
To tag (color highlight) all search terms at once use the Find and Tag command.
This panel provides a way of searching for terms in context where the order of the terms is important. The "Left" and "Right" term fields and their "threshold" fields allow you to distinguish between "dog bites man" and "man bites dog".
The User menu can be customized by editing the file "co_usermenu.txt". Use the format:This example is used to distinguish "feet as body part" from "feet as measure".
To insert a separator between two menu items just include a line with the string "separator" in the file. To leave a text field empty insert an empty line.
Selecting the "Edit" item in the User menu displays the co_usermenu.txt file in dialog text area. After you have edited it you can save your changes be clicking the "Save" button.
Bag O' Dreams uses four methods for finding themes in a set of dreams. These are; k-means, k-fuzzy and r-Cluster. These are collectively referred to as AutoCluster reports.
This is a method for finding themes in a set of dreams without user-determined sets of search strings. This makes it a "blind" search method. K-means uses a variation on a technique described by Dhillon and Modha (i.e. the spherical k-means algorithm) to find the themes.
The algorithm starts by randomly partitioning the dreams into some number of sets (usually 10 to 40). The quality of each set is determined and a concept vector created. (The concept vector is a numerical representation of all of the words and their frequency of occurrence in the set of dreams.) Once this step is done each dream is compared to each of the sets of dreams and moved to the one to which it is closest. Next, the sum of the set's quality is calculated and the steps are repeated until this sum stops increasing.
To use automatic cluster detection follow these instructions
After a long time you'll see a table of results consisting of lists of dreams, their "best terms", "characteristic terms" and a measure of the quality of each cluster generated. Clusters of 5 to 20 dreams seem to be the most coherent.
For my own dreams I get some intersecting results. My flying dreams include the terms "fly, float and glide" as expected, but also "ball, body, fence, tuck, ability, inch, land". I am sometimes flying over a ball game of some sort, or over a fence. I also tuck into a ball. In many cases I mention the position of my body while flying and reflect on how unusual my ability to fly is. Often my flight is only an inch above the ground and I often wonder where I will land.
Other themes that often show up are "being on or near a beach", "swimming" (distinct form just walking near a beach), being at a "train station/trying to catch a train", "trees", "climbing" and a second slightly more narrowly defined variation where I climb on a hillside that is loose or crumbling. I've also found "my old laboratory workplace", "coin operated machines/finding money", "apartment living/secret rooms" and "boy in danger of falling" themes. A very clear "reading" cluster also show up, though I have very few dreams about reading. Maybe this is because my reading dreams are only about reading.
The k-means algorithm only allows a dream to be assigned to a single cluster - the one that it is "closest" to. Since many dreams contain more than one theme I thought it might be useful to reassign each dream to the cluster where it scored second best. The Fuzzy k-means report lists each dream in two clusters; the dreams found by non-fuzzy k-means and "secondary" dreams which score higher in other clusters but also might have been included within the primary cluster. The first dreams in the secondary cluster often score higher than the last dreams of the primary cluster!
| Tip: Since AutoCluster reports take a long time to run consider saving them as HTML files in the "settings" folder. When Bag O' Dreams starts up it loads any HTML files that it finds there into the drop-down selector on the About Panel. If you've reloaded (added) the same set of dreams you can start where you left off in your exploration of the report's results. |
The k-means algorithm assigns every dreams to a cluster. In some cases a dream might be so different that excluding it from all clusters would improve the coherence of the results. You can exclude poorly fitting dreams from k-means and k- fuzzy results by entering a "closeness" threshold value (in the r-threshold field). A good starting value is 20 (range > 0 to < 100). When you run k-means or k-fuzzy, dreams that aren't close to any cluster are excluded. These dreams are displayed in a list at the bottom of the report titled "Outliers".
The "r-Cluster" report is found in the AutoCluster panel. This report clusters dreams based on relatedness of terms. Dreams are clustered together if their relatedness is greater than or equal to the value entered in the "r-threshold" field. This is the only parameter used by the report and should usually be set between 10 and 30. A low threshold results in several very large clusters (near the end of the report). A high threshold tend to result in a set of smaller clusters. When the "r-Cluster" button is pressed a relatedness matrix of all of the dreams is generated. Next a single dream is randomly selected and all of the other dreams with relatedness scores at or above the threshold are added to form a cluster. The process is repeated until every dream has been added to a cluster. (Note that this process gives greater weight to the first dreams selected so that by the end a large number of dreams end up by themselves in single-dream clusters. Even so, these dreams tend to have fewer close neighbors than dreams in larger clusters.) Set minimum cluster size to 3 or 4 to avoid seeing the very small clusters.
The phrase "spotlight dream" comes from Calvin S. Hall. He used it to mean a dream who's meaning was obvious and helps to illuminate the meaning of other dreams in a set. In this context spotlight dreams are simply the highest scoring dream in each cluster. These dreams ought to give a good idea of the main themes in a set of dreams.
The "Spotlight" dreams link appears near the bottom of the AutoCluster report.
The "Index" report is found in the AutoCluster panel. After you run at least one AutoCluster report (k-mean, fuzzy-k or r-cluster) an "index" is generated. The index is the intersection of each cluster's characteristic terms and the top scoring dream (spotlight dream) of the cluster. The report shows the list of spotlight dreams with their index terms. The idea is to make the characteristic terms more concrete by filtering them through an actual dream document. As you run more AutoCluster reports more terms and dreams are added.
Dream dictionaries are notorious for producing nonsensical, silly or just plain obvious interpretations for various elements occurring in dreams. Even those who believe that dreams have deeper hidden meanings tend to scoff at the use of dream dictionaries because they claim that dream symbolism is individualized to the dreamer. Those who are convinced that dreams are random nonsense don't feel there is a need to explain every detail of a dream because dreams aren't messages at all. While I agree that dreams aren't coded messages (e.g. from the unconscious) I think that dreams are none the less meaningful in the following ways:
Anthony Stevens points out that some of the nonsensicalness of dream dictionaries
stems from translations which obscure puns that are only apparent in the original language.
"... to dream of uncovering one's behind means one is about to lose one's parents. This is incomprehensible unless one knows that the Egyptian word for buttocks closely resembles the word for orphan." |
The claim that (apparent) symbolism found in dreams is all made up in the process of a "dream expert's" interpretation, like a "cold reading", or that holding a particular "theory of dreams" predisposes the dreamer to have or recall dreams that match the expectations of the theory has some merit. But in spite of this I'll continue to identify some dream elements as symbolic even though I don't think there's a agent working behind-the-scenes, trying to send me a message or trying to obscure or censor the "true" content of my dreams.
The usual assumptions made by a dreamer, while dreaming, are that they are awake, embodied and situated in a place and time. These assumptions make up a context for the dream which acts as a filter on the incorporation of internal and external stimuli (or the lack of them!). Under these conditions, the call of a cardinal (a very red member of the finch family) which eventually wakes me is symbolized as blocks of red colored text periodically scrolling by on a computer screen when the ongoing context of the dream was "searching for a bug in the code of a computer program".
"I am looking at some java code, trying to find a bug. It scrolls by in a computer window (or just a paper print out?) Every so often I come to some code comments that are printed in red. I slowly wake up to hear a cardinal singing: its bursts of song are spaced at the same interval as the red code comments."
In another dream the ongoing context of being at a workplace was interrupted by an episode of sleep apnea. This was symbolized as "squeezing past an annoyed man in a narrow corridor". Hartman gives an example of a woman with a sore throat who dreamed (symbolically) of a staircase with pink scratched walls (ongoing context not mentioned). Though anecdotal in nature, these cases show that elements of a stimulus are only incorporated in a dream if they match its ongoing, assumed context. Other elements that don't match context never reach the level of conscious awareness.
In the case of the cardinal dream, there's no reason to think that "my unconscious*" acted as a censor to my becoming fully aware of the fact that a cardinal was singing (*i.e. my unconscious as an agent the "per-screens" all sensory data and only lets me be aware of some of it - to preserve sleep, as a Freudian might claim). Instead, I'd prefer to describe this dream as a result of automatic "feature detectors" working to find associations for the bird song. These "feature detectors" would have normally been able to recruit and activate memory and attention to the fact that a cardinal was singing, however, the ongoing dream and my conviction in it's reality, needed to be overcome first. In the mean time, only "red" and "periodic" make it to conscious awareness (because these features could be fit to ongoing context).
A dream dictionary can't know the ongoing context of the dream. In this way the objections that the meaning of a dream symbol is too individual to be simply looked up on a list is seen to be well founded. It is unlikely that I would have guessed that a cardinal's song was the source for the periodic appearance of red text in my dream if I hadn't woke up and heard the bird singing. I'm comfortable calling "red/periodic" symbolic of a cardinal singing even though there's no behind-the-scenes agent who's trying prevent me becoming aware of the bird singing, disguising and censoring it.
An unintentionally ironic entry from "10000 Dreams Interpreted" by Gustavus Hindman Miller
might serve as a admonition against relying on dream dictionaries:
"To dream that you are referring to a dictionary, signifies you will depend too much upon the opinion and suggestions of others for the clear management of your own affairs, which could be done with proper dispatch if your own will was given play." |
So why a dream dictionary? Well, for one thing, some people believe that dreams have deep, hidden meanings and that elements that occur in their dreams are symbols. They couldn't care less about my opinions on dreams! For these users it's fairly easy to write their own list of symbols and definitions. (Use the little drop-down menu to add/edit or delete dictionary entries.) The second reason is that the Interpretation report uses the same file as the dictionary to provide suggested meanings for themes found in a dream. These can include both traditional, metaphorical interpretations as well as "brain based" explanations for common dream phenomena. Since a dream can contain more than one of these themes I though it might be interesting to see if the list of "meanings" provided by different themes intersects in a dream. Maybe a "car out of control" dream is just a variety of a "flying" dream with all of the motion confined to the horizontal. Maybe the difference between the two varieties of dreams is a tacit awareness of the sleeping body's horizontal position - with this (and "fictive motion/REM motor cortex activation") you get "flying", without it you get "car out of control".
The Dictionary has a simple built in editor (controlled by the drop-down with the tiny icons). There are a few limitations with this;
Here's an online version of the dream dictionary that's included with Bag O' Dreams.
You might also find the Cambridge dictionary of idiomatic expressions useful.
Hyperdictionary also provides this excellent dictionary/thesaurus:
|
|
|
For a frankly nonsensical dictionary see Gustavus Hindman Miller's 10000 Dreams Interpreted. This is also available in a format for use in Bag O' Dreams. To use it click on this link and save it as a text file, replace the interpretations.txt file in your settings folder with the new file (after you make a back-up copy) and restart Bag O' Dreams.
"He said something like 'blah, blah, blah'."If I'm not sure of an event or element I put it in parenthesis and add a question mark at the end:
"We drove in my car(?) but got to the airport too late."A lot of popular dream interpretation emphasizes emotions in dreams and asks the dreamer to give emotional impressions for almost every element in a dream. This seems a bit forced to me, especially as events in dreams are sometimes experienced without a lot of emotion. Adding emotion to a dream when there really wasn't any distorts the dream to fit the theory.
The cluster, relatedness and theme searches used by Bag O' Dreams work best when you have a fairly large collection of dreams. Detailed dream reports are needed in order for these methods to reveal any unexpected relationships between dreams and dream elements.
Dreams fade quickly. If possible try to go over the dream before you get out of bed. Record the dream as soon as possible. Write (on paper) as many details as you can remember. Get the overall sequence of events in order:
Starts out at my parents' house,Fill in details:
Joe is there, strange red dog,
My brother, my father, my sister, my daughter,
waiting for someone (my mother?) to return,
I go out and drive to the beach,
two women in a car,
follow them to a pier,
a child in a small boat (I'm very concerned he'll capsize)
etc.
Joe (friend from work) sat to my right at long table in dining room. It was twilight.
He said something like "when do we eat?",
There was a small dog with red fur that sat curled up under the table. I was afraid that I'd put my foot down on its tail.
We were waiting for my mother (I think?) to return so we could start the meal.
The occasion wasn't really specified - might have been my son's birthday(?)
Nothing much seemed to be happening for a long while, just feeling a little bored
I finally decided that I'd go out and try to find her. Its very dark now. I get in my car.
etc.
Select a title: something unique about the dream or a phrase from the dream. This will make it easier to remember the dream later.
Select a date: I usually use the date that I recorded the dream, assuming that the dream occurred after midnight. (Use the "Date stamp" command in the File menu.)
Save the dream in your dreams folder and you are ready to add it to Bag O' Dreams.
I had the web site of Schneider and Domhoff (reference 1) in mind when adding this feature. If you generate a random sampling of dreams from this site's search page, it is supplied as a long HTML document that can be saved as text. Each dream in the text document starts with a string like the following: "# 001 (01-01-1964)". The "Split" command will nicely chop up this document into it's individual dreams.
Bag O' Dreams also saves the date if it is found in the first line of a dream. This is used for searching by a date range. The format is month-day-year e.g.: 06-27-2002.
You can "bootstrap" the reports by using the results from one report as input for another.
An example would be to do a 'Cluster (tw)' search using the Driving search menu item;
"drive car ride passenger automobile truck traffic highway steer accelerate van suv".This should return a cluster of dreams that have an above average amount of driving. It also returns a list of "Best Terms" (words that occur in all or most of the dreams of the cluster). The Best Terms will usually not be identical to the original search terms;
"drive road hill path"Using them in a second Cluster search often results in a more representative cluster of driving dreams. You can then run reports such as Modes, Interpretation, Quick Profile or Most Significant Terms from the cluster's window to see which themes and features are over or under represented in the driving dreams.
Some example reports are shown here.
To run an Interpretation report on a single dream, select "Interpretation" from the dream's Search menu.
The Modes report is available from the search menu of individual dream documents and clusters.
Once the program is running you may wish to customize the menus in the Search and Co-occurrence panels.
The Search menus can be edited by changing the menu.txt file in the settings folder. You can use the "Save terms" command (in the Search panel, right bottom drop-down selector) to edit the currently selected search strings for a menu item. This updates the search strings but will not add new items. To add new menu items use the MenuEdit utility ("MenuEdit.jar" in the modules folder). Changes made by MenuEdit aren't available until you restart Bag O' Dreams.
If you want to edit the "menus.txt" file by hand use the following format;
<Menu title>Example:
menu item
search terms and phrases
<Typical>
Climbing
climb "almost vertical" "nearly vertical" "road climbs" cliff hillside steep
Avoid spaces at the end of lines. Don't use parentheses, quotes or other punctuation characters in menu titles. Do not use line returns in the search terms text area.
Some useful search stringsHere's some useful strings of terms and phrases to use in various searches.
|
A few examples;
my (mother|father|mom|dad)
meaning: any sequence starting with "my ", followed by one of "mother" or "father" or "mom" or "dad".
matches: "my mother", "my father", "my mom", "my dad"
fl(y|oat){1}(ing|s){0,1}
meaning: any sequence starting with "fl", followed by exactly one of "y" or "oat", followed by zero or one of "ing" or "s".
matches: "fly", "flys", "float", "floats", "flying" and "floating"
[^a-c]at
meaning: any sequence NOT starting with "a", "b" or "c" and ending with "at".
matches: "rat" and "hat" but doesn't match "cat" or "bat"
[mf][oa]ther(.in.law)?
meaning: any sequence starting with "m" or "f", followed by "o" or "a", followed by "ther", followed by an optional ".in.law", where the "." character can be any character.
matches: "mother", "mother-in-law", "father", "father in law" (no hyphens) and even "fatherXinXlaw"
\bdog\b
meaning: the sequence "dog", but only if found on a word boundary.
matches: "dog" but not "doggerel" or "hotdog"
(To match the last two use \bdog\B and \Bdog\b.)
Because regular expression searches can return anything (the whole text of a dream, parts of words, all of the punctuation characters in a dream) there's no simple way to score the dreams for how well they matched the search pattern. (You could count the number of matches returned for each dream and then weight them by the rarity of each match. But this would only work for "sensible" results.) Therefore no scores are given to the individual dreams in clustered versions of Regex reports.
Here are a few useful regular expressions;
metamorphosis
\b((switch|change|turn)s{0,1} (out|into|to))|((bec(a|o)me)s{0,1}\b (a|an|the))\b|\b(now (is|a|an))\b
flying
\b(I|I'm|I was) fl(ew|y|oat){1}(ing){0,1}\b
i do
\b(I|I'm) .{2,34}[.?!]("|')?
he says/she says
\b(she|he) says.{2,51}[.?!]("|')?
repeated word
b(w+) 1b
Relatedness of dreams is calculated from the number of occurrences of terms in a dream and their overall occurrence in all dreams. The list of a dream's term frequencies is treated as a vector. The cosine of the angle between two dream's vectors is determined and is used as a measure of similarity. Documents with a lot of terms in common "point in the same direction" in term space. The cosine of 0 degrees is 1.0 (perfect agreement). The cosine of 90 degrees is 0.0 (no terms in common). This number is multiplied by 100 to make it consistent with other measures. In practice, a value of greater than 20 indicates a shared "topic" between two dreams.
If you set the threshold to zero then the Term related report will automatically adjust it to the median of the score so that low scoring dreams are excluded.
Cluster searches take a set of terms and phrases and search for dreams that contain them in higher than average numbers. Rare terms and phrases boost the score of dreams containing them.
If you set the threshold to zero the cluster searches will automatically adjust it to the mean score for the cluster so that low scoring dreams are excluded.
q is a measure of cluster quality which shows up in Cluster and Term related search results. q is calculated from the sum of squares of the mean term frequency vector divided the sum of the individual dreams' term vectors:
q = 100 * (||mean||/ sum of ||vi||)
q gives some sense of the "single-mindedness" of a set of dreams. If the set is small the q will be higher. You can get some idea of the quality of your result set by finding the q of a random set of dreams of the same size as yours. (Use the Random Cluster command.)
See the Dhillon and Modha reference for a more detailed explanation of cluster quality.
An interesting result is to find all dreams without characters. To find these dreams I used the following search string in an anti-cluster:
them who his him her we they kid child children baby female she students people somebody someone stranger woman women girl girlfriend lady mom dad male he man men boy boyfriend gentleman guy "group of people" "my mother" "my sister" "my daughter" "my father" "my brother" "my son"This returns a very small cluster of very short dreams which feature a zoomed-in point of view, static imagery and little emotion.
Best terms are those terms that occur in all or most of the dreams in a cluster. Terms that occur in all of the dreams are shown red. Those that occur in greater than half of the dreams (but not all) are shown in orange. Terms that occur in one third to one half of the dreams are shown in blue.
Example of best terms from a cluster of 20 dreams:
| 20: train | 12:
station | 11: subwai | 10: platform |
7: line push singl slightli track
Best terms links appear at the bottom of any report that returns a cluster of dreams;
All cluster searches show the raw frequencies of each term in the cluster. This is the sum of the occurrence of a term in all of the cluster's dreams. Only the 40 most frequent terms are shown.
Example frequencies from a cluster of 31 dreams:
| 136: fly | 18: float | 17: beach | 16:
fenc | 14: ian | 13: yard ball | 12: pond meet inch | 11: roof | 10:
wooden women wet tub tree polic path jetti | 9: tall squar sidewalk
moment leg hole hit catch bodi | 8: watch stone push marsh land
island hot hair foot dune | 7: test talk
Characteristic terms are those terms that score highest in a particular cluster. Unlike best terms, each characteristic term can only occur in a single cluster. Because of this characteristic terms can only be shown in clusters produced by AutoCluster which partitions a set of dream into mutually exclusive sets. Characteristic terms are displayed in a link in purple text. If you click on the characteristic terms they are entered into the cluster's search field so you can do other searches on them.
Example characteristic terms from a cluster of 31 dreams:
fly float yard ball pond inch roof wooden wet tub jetti tall squar sidewalk
moment leg hit catch bodi stone
The modules folder allows you to access several stand-alone applications from Bag O' Dreams. Modules must be Java applications in jar format. Modules show up near the end of the second drop-down list of reports on the Search panel
The thesaurus is a stand-alone application which uses Grady Ward's Moby thesaurus file and a "sounds like" lookup algorithm from William Brogden. It's another java application. For installation and usage see the READ ME file in the distribution. Download it here thesaurus.zip.
If you install the Thesaurus it becomes available on the Dictionary panel.
14, ENTER, 12, *, loginstead of;
14, *, 12 ,= , log.
Calculate h-score for differences in occurrence of a category between two groups of dreamers. (See Schneider and Domhoff http://psych.ucsc.edu/dreams/Info/statistics.html for explanation and example.)
Note that items can belong to more than one category: "April" can refer to the month or a woman's name.
"...no book or even computer software could ever list the cast of thousands who have appeared in your dreams..."
-Gayle Delaney, In Your Dreams, p. 163.
The Names report generates a list of possible character names found in a set of dreams. To narrow down the results to people likely to be important to the dreamer, set the threshold field to 1.0 and run a Frequency report on the list of names. Characters who are important to a dreamer usually appear with a frequency greater than 1 per 10000 words. To find out who these characters are run a Pairs report on them after adding phrases such as; "my sister", "my brother", "my daughter", "my son", etc.
float tree branch air fly higher surfac lake abil low easili swim altitud ball straight fenc proveA few of these terms have to do with swimming and a Cluster search returns dreams about flying and swimming. To exclude swimming-related dreams I used "fly float -swim" and got these search terms;
float air fly higher abil low twenti altitud height roofWhen I did a Cluster search with the above terms I got much better results. Note that this doesn't exclude all dreams with swimming because I sometimes fly over water or fly and swim in the same dream.
java -Xmx128m -jar bag.jar
If you are using Windows, you can save the command line in a bat file (e.g.: "BOD.bat"). You can launch Bag O' Dreams by double-clicking on the bat file. This is especially useful if you find a bug because running from bat file will open a system console where error messages are displayed.
You can also add less than all of the dreams in the set. Use the drop down next to the Add button (on the Settings panel) to select a suitable number of dreams.