What's new in SuperMemo 2006

Summary

The most important improvements to SuperMemo 2006 as compared with SuperMemo 2004:

  1. The priority queue makes it possible to increase the volume of incremental reading and still improve retention on mission critical material
  2. Incremental reading now uses automatic priority setting to make sure you spend minimum time on pondering priorities
  3. Fast repetition scheduling substantially speeds up SuperMemo and allows of new functionality that was previously not possible (e.g. auto-sort, auto-postpone, etc.)
  4. Repetition auto-sorting ensures that you begin each learning day from your top-priority material
  5. Repetition auto-postpone ensures you never need to worry about the overload of material for review
  6. Incremental reading with pictures is now easy. For the first time, having most of your articles and items illustrated with pictures may become a norm rather than an exception:
  7. Import from Wikipedia takes one keystroke instead of a series of up to 10 operations and no less than 30 seconds in SuperMemo 2004 
  8. Import : Files and Folders allows you to import to SuperMemo the entire folder structure of your file archives (e.g. picture albums) 
  9. Collection chooser provides the fastest way to open collections with a single keystroke (usually the first letter of the collection's name)
  10. Stylesheets in HTML components and templates make it easy to globally change formatting of texts (e.g. fonts, colors, etc.)
  11. SuperMemo use statistics will now help you see more statistics, further back in time, and with more detail
  12. Decompose and Dismember in incremental reading accelerate the conversion of elements that conglomerate information into easy-to-learn well-formulated pieces of knowledge
  13. Hiding SuperMemo in the system tray (Ctrl+Shift+G) makes it possible to keep the alarm timer going while working unobtrusively with other applications. Restarting SuperMemo from the tray is instantaneous
  14. Faster element browser opens thousands of elements instantly (previously, huge collections might take minutes on PCs with less memory)
  15. Inserting registry objects to elements is now simpler. For example, with a click on the browser toolbar, you can choose a picture to illustrate all elements stored in the browser. Elements that already include pictures can have them tiled conveniently   
  16. Quick splitting of large articles in incremental reading with split-marks
  17. Postpone is now less detrimental as the degree of delay depends on element's priority
  18. Unicode titles can now also be displayed in the contents window
  19. Tiling components makes life easier in elements with multiple components (e.g. pictures)
  20. Improved multiple choice tests now allow of all component types and all image formats
  21. Some of the deprecated options have been removed to reduce option clutter and make the program faster and/or more compact
  22. See also: Minor improvements, bugs and annoyances

Priority Queue

Priority queue is the most important new feature of SuperMemo 2006. With the new priority system in SuperMemo 2006, as well as with huge hard drives, large memories and fast computers, only your available time is now a limit on how much material you can handle with SuperMemo. Even your memory is no longer a factor due to the fact that you can store low-priority low-retention material without affecting the mainstream learning process. You may still want to pause before adding excess material to your collection, but your learning overheads will only depend on how accurately you define the priority of such imports.

The priority queue makes it easy to sort the learning material by priority. Each element in SuperMemo 2006 carries a number called a priority. Priorities are relative. This means that newly inserted high-priority elements displace previously inserted elements of lower priority. Reducing the priority of one element, increases the priority of other elements. The entire learning process proceeds along the priority criteria. This means that you will never neglect your top-priority material. Even if your collection holds far more material than you can ever hope to process and learn. You will now always begin your reading with top priority articles or extracts. You will make sure that the retention of your top-priority items is strictly determined by the forgetting index only. On the other hand, you can accept an overflow of lower priority items marked by a higher tolerance for forgetting. Those items will still make a valuable part of your learning. Due to the spacing effect, you will be able to increase the speed of acquiring new material even though, within lower-priority ranks, you will lose some control of which portions of your learning are retained in memory and which might be forgotten. 

For details see: Priority Queue

Automatic priority setting in incremental reading

In incremental reading, extract priorities depend on a number of factors. Primarily, they are derived from the priority of the parenting topic. The higher the priority of the parent, the higher the priority of the extract. However, for very high priorities, extract's priority is likely to be lower than that of the parent, while for low priority articles, extract's priority may be much higher than that of the article. This is because high priority articles are supposed to be read slowly line by line and generate many extracts, while low priority articles are only skimmed or speed-read while fishing for extracts. Length of the extract also affects its priority, with short extracts receiving a higher priority. This is because long extracts are usually created for the sake of later reading, while shorter extracts are more likely to contain meaningful information that is only a step away from generating cloze deletions. Finally, a tiny degree of randomization is added in generating extract priorities in order to prevent clustering of extracts coming from the same article while making repetitions sorted by priority.

Priority of cloze deletions is similar to that of the parenting extract. To prevent the clustering of items derived from the same extract, as well as to avoid spacing effect and interference, priorities and the first intervals of cloze deletions are dispersed randomly around the presumed optimum value.

Fast repetition scheduling

Some of the repetition scheduling procedures in the previous versions of SuperMemo for Windows have been inherited from SuperMemo 1.0 for DOS (1987)! Those procedures were optimized for times where disk space was tiny and computer memory was limited to 640 KB (i.e. a few thousand times less than an average PC today). In SuperMemo 2006, large structures related to the repetition schedule are kept in memory and periodically saved to the disk to ensure negligible sensitivity to errors on power failure. This makes it possible to increase the speed of some rescheduling procedures by up to 200 times (esp. Postpone, Mercy, Reschedule, Sort repetitions, and repetition schedule verification at Repair collection, etc.). Rescheduling operations now proceed at 10,000 elements per second on an average PC. Most importantly, new functionality became possible only with fast repetition rescheduling. In particular, auto-sort and auto-postpone options that now, by default, begin your each learning day would not be possible in SuperMemo 2004.

Import from Wikipedia

Edit : Add to category : Wikipedia (Ctrl+Shift+W) makes it easy to import articles from Wikipedia in incremental reading. All Wikipedia articles opened in Internet Explorer will be imported upon confirmation. If no articles can be found, SuperMemo will open the article whose title matches the text currently selected in SuperMemo. If no text is selected, SuperMemo will ask for the title of article to import. If no such article exists in Wikipedia, SuperMemo will initiate a search for the term of interest in other Wikipedia articles. Upon import, each article is filtered to remove junk tags irrelevant in learning (e.g. navigation options, search boxes, alerts, funding requests, redirections, section edits, tables of contents, etc.). All articles become tagged with Wikipedia references, the title, the date, and the link to the original article. Those references propagate into all extracts and clozes in incremental reading. Only table filtering and image imports are left at your discretion.

Import : Files and Folders

Import : Files and Folders allows you to import to SuperMemo the entire folder structure of your file archives (e.g. picture albums, collections of archive articles, miscellaneous to-do files, etc.). You can clean up your file archives with a single keystroke and process them incrementally with tiny allocations of time each day. Instead of keeping your family picture archive in mothballs, you can now come back to selected memories in a systematic manner. You will be surprised how much you can learn from old pictures and how addictive the whole process can be. You can finally clean up the archive of articles you have downloaded from the web for later reading. Incremental reading will provide a systematic and rational way of matching your reading ambitions to the available time. You can apply incremental processing to any imaginable to-do file archive by importing the archive to SuperMemo.

Collection chooser

Collection chooser helps you quickly access your most frequently used collections. Instead of navigating through folders with File : Open collection, or searching for the right collection number of the pick-list on the File menu, you can simply press Ctrl+O and the letter that designates your collection (e.g. K for Knowledge as in the picture below). You decide which collections will show up in the chooser by adding them with the button Choose another. You can still use the pick-list for recently used collections and File : Open collection for less frequently used collections that are absent from the chooser or pick-list. Collection chooser is also used in various contexts where you need to pick a collection. For example, if you want to transfer an element to another collection, SuperMemo will open the collection chooser to help you pick the destination collection. 

Improved image downloads

Temporary Internet Explorer cache is used to instantly retrieve files without the need to download. For a default cache size, this provides instant access to 95% of image files. This statistic may not apply to articles with multiple images immediately after opening the article in SuperMemo (i.e. at a time when HTML component is still downloading the images via http). Multiple images can be tiled to illustrate a single article.

SuperMemo: downloading images (Blue Mountains article from Wikipedia)

Improved web import

The following import modes are available for web pages opened in Internet Explorer:

Example: imagine you are studying the landscapes of the Lost Coast and you open a few articles on the subject in Internet Explorer. Here are some of your choices:

Disclaimer: not all pages and not all images can be processed using SuperMemo web import. Some pages and images generated programmatically may need to be imported manually via copy&paste. This problem affects 1-2% of typical pages and/or images.

Image zooming, trimming, compressing and cutting

Several new operations on images help you save space and better illustrate your articles in incremental reading:

Tiling components

This option makes it possible to choose components from the component list for tiling. By default, all components in editing and dragging modes are checked for tiling. The tiling covers the maximum rectangle currently covered by the most extremely placed components on the list. To change the tiling area, move the extreme components to new locations. Tiling is used automatically by default in several contexts, e.g. pasting images to an element.

Exemplary topic with images auto-tiled at image download:

Picture templates

It is now easier to use picture templates. If you take any template, add a picture to it, and then name the new template by adding 'Picture' or 'P' to its name, you can have SuperMemo automatically use that template in elements where pictures are used. For example, if your category has a default item template called 'Geography', you can create a picture template, name it Geography Picture or GeographyP (without a space). You will then be prompted to use this template in cases such as: moving an element with a picture to any category where Geography is the default template, pasting picture to elements with the Geography template applied, generating cloze deletions in categories where Geography is the default item template, etc. 

Inserting registry objects was made easier

SuperMemo 2006 simplifies adding components by making it easy to insert selected registry members into elements. To insert a picture to many elements you previously had to use templates or tediously repeat Links : <Registry> : Accept in many elements. Insert in registries now makes it possible to open the image in the registry, leave the registry in the background, move between elements, and add pictures to elements with a single click on the Insert button in the registry. A new option, Template : Insert picture in the browser (on the subset operations menu) makes it easy to illustrate the entire subset of elements with a selected picture. Finally, Insert in registries simplifies organizing components in the element by shifting or tiling components to make room for new inserts.

Operation Decompose in incremental reading

The operation Decompose makes it possible to generate multiple topics or multiple items from enumerative elements. 

For example, the following cloze violates the minimum information principle: 

Q: The levels of serotonin, acetylcholine, noradrenalin, and somatostatin all [...] in Alzheimer's
A: decrease

This cloze can easily be edited to the following form:

Q: The level of {serotonin/acetylcholine/noradrenalin/somatostatin} [...] in Alzheimer's
A: decreases

Upon applying Decompose, it will be be converted to 4 cloze deletions such as:

Q: The level of serotonin [...] in Alzheimer's
A: decreases

Q: The level of acetylcholine [...] in Alzheimer's
A: decreases,
etc.

The option Decompose can be found on the component menu as Reading : Decompose.

Operation Dismember in incremental reading

The operation Dismember makes it possible to generate multiple elements from multi-component elements.

For example, if you receive mail with multiple picture attachments and import this mail to SuperMemo, you can use Dismember to produce separate elements for holding each picture. Selected components (e.g. mail body) can be chosen to propagate to child elements for retaining the context. This way, individual attachments or pictures can enter the review process as separate elements.

The option Dismember can be found on the element menu as Edit : Dismember.

Stylesheet registry

Stylesheet registry can hold a set of stylesheets that can be used to format HTML texts in individual components or in templates used throughout the collection. Stylesheets can also be used to customize the appearance of extracts, cloze deletions, headers and references in incremental reading. A mini-editor for styles was included for those who do not know stylesheet specifications (CSS).

SuperMemo use statistics

SuperMemo will now display use statistics without cutting off old data. You will be able to view some of the data back as far as from the moment you started using SuperMemo 2002 (e.g. Use : Element count : Items). The graph clutter can be solved by sweeping off portions of the graph with the mouse (press down the left mouse button and move the unwanted portion off the graph out of the graph area).   

SuperMemo will also use timeline statistics to recover some of the past data from the moment you started using SuperMemo 2004 (e.g. Use : Work done : Use time).

Finally, SuperMemo 2006 introduces new use statistics that will be collected as of the moment of the upgrade to SuperMemo 2006 (e.g. Use : Efficiency : Retention).

Quick splitting of large articles

To speed up splitting large articles, you can now insert split marks and execute Reading : Split : Split Article (on the component menu). The article will be cut into smaller portions separated by split marks. To insert a split mark you can:

Degree of Postpone depends on the element's priority

In earlier versions of SuperMemo, you could define a degree of delay when postponing a subset of elements. Now the delay is modified depending on the priority of an element. For example, if the delay factor is 1.1 (i.e. all intervals are to be increased by 10%) then the modified delay is spread in twice the range of the delay factor: from 1.0 to 1.2 (1.0 for highest priority elements and 1.2 for lowest priority elements). After the modified delay is computed, minimum and maximum delay interval limits are imposed. Finally, the increment in the interval is dispersed randomly around the proposed value. The dispersion follows a normal distribution with cut off points between 50% and 150% of the proposed value. For example, if the delay factor of 1.1 produced an interval increment of 20 days, the actual increment will fall in between 10 and 30 days. As delay increases faster than the drop in priority, average delays should be more significant except where limited by maximum delay interval.

Improved multiple choice tests

Multiple choice tests can now use all component types (e.g. HTML, RTF, etc.). All image formats available in SuperMemo can be used in tests.

Reduced option clutter

SuperMemo 2006 sheds vast amounts of code used to implement options which became obsolete or deprecated. This helped reduce program's size and complexity as well as to increase its speed in a number of operations. Please read the list of options removed before you decide to upgrade: 

Minor improvements, bugs and annoyances

The complete list of changes to SuperMemo 2006 is 989 positions long. Here are some minor improvements that could make a major difference to some users:

Improvements

  1. Customizable web search for a phrase that is currently selected in SuperMemo (through dictionaries, encyclopedias, databases, blogs, or sites and services of your choice, etc.)
  2. Find button on the element toolbar helps searching through longer texts. In HTML texts, all occurrences of the found string are highlighted. F3 can be used to jump between highlighted matches
  3. Ctrl+Shift+E to send the current element via e-mail
  4. Shift+click for editing pictures
  5. Q&A export may now include HTML tags
  6. Postpone makes it possible to postpone all elements except a set number even if they do not meet postpone criteria
  7. Copy, Paste and Move for elements in the knowledge tree (Contents) make it possible to use keyboard for rearranging the tree (instead of the mouse and drag&drop operations)
  8. Workload graph can be used to visually inspect the number of elements scheduled for review in future months and years (the display can be dragged with the mouse to scroll as far ahead as the most remotely scheduled element) 
  9. New shortcut, Shift+Alt+X (same as Alt+X for Remember extract with Shift pressed), makes it easy to set the priority of a new topic at the moment of extracting it in incremental reading 
  10. Repetition history stores the precise time of a repetition (for circadian research and future optimization of learning time in SuperMemo)
  11. Q&A import supports ANSI, UTF-8, little endian UTF-16, and big endian UTF-17
  12. Alt+R for a fast rename of the registry member associated with the current component
  13. Image components show hints with image registry member name, file name and path, file size, date, etc.
  14. Improved memory manager (courtesy of Delphi 2006) speeds up operations requiring vast allocations of memory
  15. Possibility to shift a repetition to "later today" (Ctrl+Shift+J)
  16. Picture forwards may be compressed before sending
  17. Picture forwards may optionally include only the zoomed portion of the picture
  18. Element and picture forwards via e-mail leave forward data in the comments field (e.g. to prevent forwarding the same information twice)
  19. Image : Use as : Windows desktop
  20. Image : Use as : About box
  21. Image : Use as : SuperMemo background
  22. Delete components makes it possible to uncheck components that should not be deleted
  23. Sorting elements in the order of occurrence in the knowledge tree in the browser
  24. Child : Descendants in the browser to find extracts and clozes produced by a set of articles
  25. Item and topic overload statistics (Analysis : Use : Overload)
  26. Reducing size of HTML files and automatically converting some of HTML to plain text
  27. When importing pictures, repeated imports of the same picture from the same URL will keep only one image in the registry
  28. Item repetitions can be advanced in a way similar to topic advancement. The difference is that absolute intervals will be shortened as opposed to topics where the interval is shortened relative to the day of execution
  29. Search and Replace makes it possible to do global replace without using the text registry. In certain cases, depending on the size of the registry and number of replace hits, using the text registry may still be a better choice due to speed
  30. Skipping text statistics in subset statistics can speed up the operation 20-50 times
  31. Tools : Plan allows of keeping Monthly and Annual statistics of daily activities exportable to Excel. Accessory statistics can also be compiled (e.g. jogging time, calories consumed, visits to the cinema, injuries, etc.) 
  32. WMA support
  33. Resizable text input dialog box for longer texts
  34. Ctrl+Shift+click can be used to reset reference hyperlinks
  35. Faster display of data in daily and monthly workload windows
  36. Browser toolbar can be used to quickly generate a child browser with all selected or all unselected elements of the current browser
  37. Ctrl+Shift+B in registries displays a browser with elements that use only the selected registry members (previously all members of the current registry subset were used)
  38. Split in Plan uses the time that elapsed from the beginning of the slot as the split default (previously, the slots would be cut in half by default)
  39. Author, Date and Source references in incremental reading can be chained in the same way as titles (e.g. New Scientist : Neuron)
  40. Collection chooser does not allow of opening incomplete collections to prevent confusion in cases where the user moves the kno file without the accompanying folders 

Bugs and Annoyances

  1. Mshtml errors in incremental reading are now nearly extinct once Internet Explorer 7.0 is installed
  2. Pictures, subscripts and superscripts are marked in registry names to prevent confusion between text-less HTML and mathematical texts that differ only in subscript/superscript formatting (Full HTML attribute must be on) 
  3. Minimize button was absent from the toolbar dock
  4. Pet annoyance of non-English keyboard users, Trim Shortcuts, is no longer needed
  5. All image conversions are now possible (BMP, JPG, GIF, PNG). Previously, image compression and format choice were not reversible
  6. Changing midnight shift would not reload the outstanding queue
  7. Binary components would not transfer between collections
  8. Edit : Save file with binary components would not remember the last used path
  9. Postpone no longer annoys with Reload browser? inquiry. As the browser is now much faster, it reloads instantly without asking
  10. Progress box would hide behind the Postpone dialog when switching between applications
  11. Alarms are now raised by displaying a non-modal dialog box
  12. Zero-sized remote images could hang the download procedure
  13. Toolbars would cover Plan when switching to other applications
  14. Some animated GIFs could hang SuperMemo
  15. Not all PNG images could be displayed
  16. Registry Unicode sorting changed from UTF-8 to UTF-16 with fewer sorting surprises
  17. Image download dialog box could not be resized and would not size label font for better readability
  18. Q&A import would impose the import template on newly created items (instead of just applying it)
  19. Pictures with non-ASCII codes in their names would not import 
  20. Search for Clinton with match words would not find Clinton's
  21. Annoying progress bar no longer shows on filtering or importing longer texts (these operations are now fast enough to be dealt with just an hourglass)
  22. Interval distributions would display range check errors for Topics + Items > 65536
  23. Collections would not open in read-only mode on a DVD-R
  24. Feedback questionnaire is now saved between sessions and may be sent later 
  25. In video extracts, playback start and end positions would be occluded by video templates 
  26. Sounds would play via Auto-Play at question time even if Play At was set to answer time only
  27. De-HTML-ize would not cut preceding tags at the beginning of texts
  28. Case conversions would ignore HTML tags
  29. FAQ dialog would require an empty FAQ html file exist before storing FAQs
  30. FAQ dialog could not be sized
  31. Removing date picker prevents tinkering with the date in SuperMemo (although it is still possible in Windows)
  32. Mail attachments would use numeric file name instead of the actual registry name
  33. Solving problems with import of mail from non-English Outlook
  34. Mail import goes to a single, dated folder
  35. Element titles could hold long texts that would produce false search hits
  36. References are no longer cached. If you edit their texts, the edits will show up in new extracts
  37. You can now paste sounds
  38. HTML templates in the template registry could produce "Error retrieving background color"
  39. Align button in the browser would not restore the browser window if it was maximized
  40. Many dialogs were enlarged to allow for a larger font
  41. Intervals beyond 40 years from setting up a collection would not be accepted
  42. Very long entries in Plan would be cut without warning
  43. Alt+Backspace in Plan would bring up the element window
  44. Exports from Plan would get confused by HTML markup if it was included in the schedule