Families

Families are, in simple words, “units of meaning”, or semantic nodes. A family is associated with a real-life concept.

Consider the English word table. What does it mean? Several things.

It may mean this:

Table, as in "furniture"

or this:

Table, as in "set of data"

Simply put, a family = one of the meanings. Of course, a meaning may have several words linked to it (or none, if the concept is missing: most likely, ancient Greek did not have a word for car or spacecraft). The connection is “many to many”: more than one lexeme can be linked to one family, and a lexeme can be linked to multiple families.

For example, a family describing trucks may have words truck and motortruck linked to it in English; they both mean the same. In Tisane, families are cross-lingual structures. That is, the same family ID (e.g. 35101) is linked to English lexemes truck and motortruck, as well as its Vietnamese equivalent xe tải or camión in Spanish.

Multiple families make up a knowledge graph; they can be linked to each other as hypernyms (“super-types”) and hyponyms (“sub-types”).

Families are used as a way to reference meaning from other parts of Tisane.

While families are a construct unique to Tisane, there are similar structures in other frameworks. For example, synsets in WordNet.

In LaMP, families can be viewed from the Knowledge Graph page.

Knowledge Graph in LaMP

Families are preserved as a separate collection when compiling. However, for the sake of portability, related tables like hypernyms, hyponyms, and domains, are embedded in the family Tisane Runtime Datastore.