Three types of files: analytical files, texts file, and attribution files

In TLS, information is organised in three basic levels (that is: there are three basic categories of files):

A) The Texts file is a category of its own. It contains original Chinese texts (and information relevant to these). This is the "raw" source material for the analysis.
B) 10 analytical files (Phonology, Characters, Synonym Groups, Syntactic Categories, Semantic Categories, Lexeme Relations, Rhetorical Devices, Lexeme Representations, Lexemes, Lexeme Entries) provide the analytical framework of TLS. They are interrelated with each other and form an attempt to describe which elements are necessary to understand a unit of the language.
C) The 3 Attribution files (Lexeme Entry Attributions, Lexeme Relation Attributions, Rhetorical Device Attributions) bring the analysis and the source material together: These files record the exact text passages where a given Lexeme Entry, Lexeme Relation, or Rhetorical Device is used. These files therefore provide both the illustrative material to make the analysis understandable, and a means to verify (or falsify) the analysis on the basis of the sources.

There are two irregularities in this scheme:

- The Sources file is not part of the scheme outlined above, as it is not part of the analysis (neither on the side of analytical framework, nor on the side of the sources to which this is applied). Its functions is to list the scholarly literature used for TLS, and the individual database records for which this literature has been used. While "Sources" are not an analytic category, organising this information in its own file allows convenient retrieval from one central access point.
- The Lexeme Relations file really should be split in two files: a basic analytic category of Lexeme Relations such as 'Antonym', 'Synonym', 'Contrast', 'Converse', or 'Opposition'; and a second file registering the application of these Lexeme Relation categories to specific pairs of Lexemes.

While the later of these irregularities constitute inconsistencies in the database structure, they are of little practical consequence, as the information concerned can always be retrieved by field-specific searches.