cjvt-srl-tagging/tools/parser/msd/msdmap.py

import pkg_resources
import pandas
import sys

# msd mappings from slo to ang
class Msdmap:
    def __init__(self):
        # http://nl.ijs.si/ME/V4/msd/html/msd.categories-sl.html
        self.pos_slo_ang = [
            ("samostalnik", "S", "Noun", "N"),
            ("glagol", "G", "Verb", "V"),
            ("pridevnik", "P", "Adjective", "A"),
            ("prislov", "R", "Adverb", "R"),
            ("zaimek", "Z", "Pronoun", "P"),
            ("števnik", "K", "Numeral", "M"),
            ("predlog", "D", "Adposition", "S"),
            ("veznik", "V", "Conjunction", "C"),
            ("členek", "L", "Particle", "Q"),
            ("medmet", "M", "Interjection", "I"),
            ("okrajšava", "O", "Abbreviation", "Y"),
            ("neuvrščeno", "N", "Residual", "X"),
        ]

        table_name = "msd-human-sl.tbl"
        table_path = pkg_resources.resource_filename(__name__, table_name)
        with open(table_path, "rb") as fp:
            # TODO: pandas is awfully slow --- might need optimization
            self.msd_table = pandas.read_csv(
                fp,
                "\t",
                names=["id", "eng_msd", "eng_long", "slo_msd", "slo_long", "slo_very_long"]
            )
            self.msd_table.set_index('slo_msd')

    def slo_msd_to_eng_long(self, slo_msd):
        # old, slow
        # return (self.msd_table[self.msd_table["slo_msd"] == slo_msd]["eng_long"]).values[0]
        # return self.msd_table.query("slo_msd == '{}'".format(slo_msd))["eng_long"].values[0]
        query = self.msd_table.query("slo_msd == '{}'".format(slo_msd))
        if query.empty:
            return "No-matching-msd-found"
        return query["eng_long"].values[0]

    def slo_msd_to_eng_pos(self, slo_msd):
        # first letter in slo_msd == slo_pos 
        return self.pos_slo_ang_map(1, slo_msd[0])[3]

    def pos_slo_ang_map(self, col, query):
        for pos in self.pos_slo_ang:
            if pos[col] == query:
                return pos
        raise ValueError("Wrong part of speech.")


if __name__ == "__main__":
    msdmap = Msdmap()
    test_msds = [
        "Soser",
        "Ppnzmm",
        "Gp-d-mz"
    ]

    for msd in test_msds:
        print(msd)
        print(msdmap.slo_msd_to_eng_long(msd))
        print(msdmap.slo_msd_to_eng_pos(msd))
        print()