Skip to content

Language model files

Naomi Leclercq edited this page Jul 14, 2021 · 6 revisions

A language model or knowledgebase (KB) consists of seven or eight csv files:

Contents Filename Description
abbreviations acro.csv a list of abbreviations that should not be treated as sentence endings, and if needed also words that in contrary mark a sentence ending
filter filter.csv transcription rules that are applied on the concept clusters at the end of the Smart Indexing process in order to optimize the clusters
grammatical labels labels.csv a list of all labels that are used in the lexrep file
lexical representations lexreps.csv a list of words and word groups with (grammatical) labels
metadata metadata.csv language-specific settings for the language model
pre-processor prepro.csv transcription rules that are applied on the input text before the actual indexing starts
rules rules.csv a series of rules to disambiguate elements that can be a Concept or a Relation depending on their context and to detect attributes and their scope
regular expressions (optional) regex.csv extra lexical representations with counterparts in the lexreps file

A full description of the contents for these files can be found in /docs/KB-file-formats.doc.

More on how these files get translated into runnable code in the corresponding section on the Build Process