Content architecture

Unlike traditional educational software, Linguamix content is formatted and interconnected: different media types are sliced and diced and linked with one another to form one global database. Raw Linguamix media types include text, audio speech, images, logical tags and similarity percentages.

The topmost element is the Linguamix module, which can be seen as a book.

The atomic content entity is the sentence:

content architecture

  • a Linguamix module contains chapters
  • a chapter contains sentences
  • a sentence is a sequence of characters, it may contain spaces, parentheses and punctuation
  • a sentence links to an audio file, within which it has been recorded as speech by a speaker
  • a sentence can optionally link to an image, whether a photo or a drawing
  • a sentence is a segment and may contain child segments
  • a segment, just like a sentence, is a list of consecutive characters, whether one or more words
  • segments are organized hierarchically; parent segments contain child segments
  • a segment can be a child segment, in which case it will link to a parent segment (possibly the entire sentence itself)
  • a segment may also be a parent segment and contain further child segments
  • a segment's speech audio has a beginning and end time measured in milliseconds, which corresponds to its chunk of characters
  • a segment can have any number of logical tags attached to it
  • logical tag are created by the course's author; a tag is just a name and can be in any language
  • tags are not meant to model human languages or societies; their sole purpose is to help teachers structure their content for easier storage and retrieval. Much like a sticker on a paper folder, there is no restriction as to what a tag can mean.
  • a tag can be attached to any type of media: chapters, sentences, segments, audio and pictures
  • tags are stored in a collection and can optionally be organized hierarchically; a parent tag logically encompasses its child tags
  • a tag collection may be shared among several modules and chapters
  • optionally, the relationship between two different tags can be enhanced with a similarity factor expressed in percent

This may seem complicated if all you're doing is playing back a single sentence, but this structure enables you to make complex queries searching through dozens of chapters at a time.

For example, you can ask Linguamix to generate slides using only the sentences that contain colors, or animals, or animals and colors. It is possible to listen to a single word, or on the contrary to listen to a sentence while muting that word. It's also possible to hear a word but not see its text. It's possible to show the question's picture but not the answers' or vice versa. Segments' similarity factors can be used in MCQs to avoid selecting decoy answers too similar to the correct answer - or on the contrary to increase complexity by doing so on purpose.

The richer the Linguamix content database, the more possibilities.