Skip to content

adding_pos.md

Latest commit

 

History

History
32 lines (32 loc) · 2.72 KB

adding_pos.md

File metadata and controls

32 lines (32 loc) · 2.72 KB

Example Use: Adding a New Part-of-Speech (POS)

The following outlines how to add support for a new POS to the FST, using the example of proper nouns being added to the Ojibwe FST.

  1. Create a config file in the morphological source.
    • For example: create proper_nouns.json in OjibweMorph/config/.
    • You also need to make sure this config file is being given as a command-line argument to csv2lexc.py, but this was taken care of automatically by the Makefile in OjibweMorph which generates the list of config files based on the contents of config/.
    • Check out the docs for more info on what the config values mean. Not all are necessary for basic functioning.
      • The following must be set and their values must be correct
        • morphology_source_path -regular_csv_files
        • lexical_database
        • morph_features (assuming you want the analysis to contain any info!)
      • The following must be set, but their value can really be anything -- just choose something specific to this POS:
        • pos
        • root_lexicon
        • regular_lexc_file
      • The following must be set, but you can just use the default/null value (described in the config docs):
        • irregular_csv_files
        • irregular_lexc_file
        • missing_tag_marker
        • missing_form_marker
        • multichar_symbols
        • template_path
  2. Add the root name to the template for building root.lexc.
    • For example: add ProperNounRoot to the end of OjibweMorph/templates/root.lexc.j2.
  3. Create a paradigm spreadsheet in the morphological source.
    • For example: create PROPER_NOUN.csv in one of the OtherSpreadsheets/ folders in OjibweMorph.
    • There's more info in the docs, but basically there are six mandatory columns for this spreadsheet: Paradigm, Class, Lemma, Stem, Form1Surface, and Form1Split.
    • There has to be at least one row for every possible paradigm and class value.
      • For example, there are two possibilities for proper nouns: paradigm and class being PersonName, or paradigm and class being PlaceName. Therefore, there are two rows in PROPER_NOUN.csv.
  4. Not required, but if you want the FST to handle more than a few cases, you'll need a spreadsheet in the lexical source.
    • For example: add PROPER_NOUNS.csv to OjibweLexicon/OPD/.
    • As outlined in the docs, the six columns are mandatory (Lemma, Stem, Paradigm, Class, Translation, and Source), but only four require real values; Translation and Source can be NONE.