Implemented specifications parse and msd->properties conversion

.gitignore

@@ -1 +1,2 @@
 *.pyc
+venv

conversion_utils/jos_msds_and_properties.py

@@ -0,0 +1,192 @@
+import lxml.etree as lxml
+import re
+import sys
+
+from conversion_utils.utils import xpath_find, get_xml_id
+
+LANGUAGE_INDEX_MAP = {'en':0, 'sl':1}
+
+LEXEME_FEATURE_MAP = {'noun':{1,2},
+                      'verb':{1,2},
+                      'adjective':{1},
+                      'adverb':{1},
+                      'pronoun':{1,2,6,7,8},
+                      'numeral':{1,2},
+                      'preposition':{1},
+                      'conjunction':{1},
+                      'particle':set(),
+                      'interjection':set(),
+                      'abbreviation':set(),
+                      'residual':{1},
+                      'punctuation':set()}
+
+LEVEL_EXCEPTIONS = {('pronoun', 2, 'čezme'),
+                    ('pronoun', 2, 'medme'),
+                    ('pronoun', 2, 'nadme'),
+                    ('pronoun', 2, 'name'),
+                    ('pronoun', 2, 'obme'),
+                    ('pronoun', 2, 'podme'),
+                    ('pronoun', 2, 'pome'),
+                    ('pronoun', 2, 'predme'),
+                    ('pronoun', 2, 'skozme'),
+                    ('pronoun', 2, 'vame'),
+                    ('pronoun', 2, 'zame'),
+                    ('pronoun', 3, 'tadva'),
+                    ('pronoun', 4, 'tadva'),
+                    ('pronoun', 5, 'čezme'),
+                    ('pronoun', 5, 'medme'),
+                    ('pronoun', 5, 'nadme'),
+                    ('pronoun', 5, 'name'),
+                    ('pronoun', 5, 'obme'),
+                    ('pronoun', 5, 'podme'),
+                    ('pronoun', 5, 'pome'),
+                    ('pronoun', 5, 'predme'),
+                    ('pronoun', 5, 'skozme'),
+                    ('pronoun', 5, 'vame'),
+                    ('pronoun', 5, 'zame'),
+                    ('pronoun', 7, 'njegov'),
+                    ('pronoun', 8, 'jaz'),
+                    ('pronoun', 8, 'on'),
+                    ('pronoun', 8, 'se'),
+                    ('pronoun', 8, 'ti')}
+
+
+class Specifications:
+
+    def __init__(self, *categories):
+        self.categories = list(categories)
+
+    def add_category(self, category):
+        self.categories.append(category)
+        
+    def find_category(self, char, language_index):
+        return next((category for category in self.categories if category.char_pair[language_index] == char), None)
+        
+    def __str__(self):
+        return 'categories:{categories}'.format(categories=self.categories)
+
+
+class Category:
+
+    def __init__(self, string_pair, char_pair, *features):
+        self.string_pair = string_pair
+        self.char_pair = char_pair
+        self.features = list(features)
+
+    def add_feature(self, feature):
+        self.features.append(feature)
+        
+    def find_feature(self, position):
+        return next((feature for feature in self.features if feature.position == position), None)
+
+    def __str__(self):
+        return 'strings:{strings}, chars:{chars}, features:{features}'.\
+            format(strings=self.string_pair, chars=self.char_pair, features=self.features)
+
+
+class Feature:
+
+    def __init__(self, string_pair, position, lexeme_level_flag, *values):
+        self.string_pair = string_pair
+        self.position = position
+        self.lexeme_level_flag = lexeme_level_flag
+        self.values = list(values)
+
+    def add_value(self, value):
+        self.values.append(value)
+
+    def find_value(self, char, language_index):
+        return next((value for value in self.values if value.char_pair[language_index] == char), None)
+        
+    def __str__(self):
+        return 'strings:{strings}, position:{position}, level:{level}, values:{values}'.\
+            format(strings=self.string_pair, position=self.position, level='level' if self.lexeme_level_flag else 'form', values=self.values)
+
+
+class Value:
+
+    def __init__(self, string_pair, char_pair):
+        self.char_pair = char_pair
+        self.string_pair = string_pair
+
+    def __str__(self):
+        return 'chars:{chars}, strings:{strings}'.\
+            format(chars=self.char_pair, strings=self.strings_pair)
+
+
+class SpecificationsBuilder:
+
+    def build(self, file_name):
+        root = lxml.parse(file_name).getroot()
+        specifications = Specifications()
+        div_elements = xpath_find(root, 'tei:div')
+        for div_element in div_elements:
+            if (re.match(r'^msd\..-sl', get_xml_id(div_element))):
+                category_element = xpath_find(div_element, 'tei:table/tei:row[@role="type"]')[0]
+                category_string_pair = self.get_cell_pair(category_element, 'value')
+                category_char_pair = self.get_cell_pair(category_element, 'code')
+                category = Category(category_string_pair, category_char_pair)
+                specifications.add_category(category)
+                feature_elements = xpath_find(div_element, 'tei:table/tei:row[@role="attribute"]')
+                for feature_element in feature_elements:
+                    feature_string_pair = self.get_cell_pair(feature_element, 'name')
+                    feature_position = int(self.get_cell(feature_element, 'position'))
+                    lexeme_level_flag = feature_position in LEXEME_FEATURE_MAP[category_string_pair[0]]
+                    feature = Feature(feature_string_pair, feature_position, lexeme_level_flag)
+                    category.add_feature(feature)
+                    value_elements = xpath_find(feature_element, 'tei:cell[@role="values"]/tei:table/tei:row[@role="value"]')
+                    for value_element in value_elements:
+                        value_char_pair = self.get_cell_pair(value_element, 'name')
+                        value_string_pair = self.get_cell_pair(value_element, 'code')
+                        value = Value(value_char_pair, value_string_pair)
+                        feature.add_value(value)
+        return specifications
+        
+    def get_cell(self, row, role, language=None):
+        language_condition = ' and @xml:lang="' + language + '"' if language is not None else ''
+        expression = 'tei:cell[@role="' + role + '"' + language_condition + ']'
+        text = xpath_find(row, expression)[0].text.lower()
+        if (text == 'adposition'): text = 'preposition'
+        return text
+
+    def get_cell_pair(self, row, role):
+        return (self.get_cell(row, role, 'en'), self.get_cell(row, role, 'sl'))
+
+
+class Properties:
+
+    def __init__(self, lemma, category, feature_value_list, language_index):
+        self.language = next(l for (l, i) in LANGUAGE_INDEX_MAP.items() if i == language_index)
+        self.category = category.string_pair[language_index]
+        self.lexeme_feature_map = {}
+        self.form_feature_map = {}
+        for (feature, value) in feature_value_list:
+            feature_name = feature.string_pair[language_index]
+            feature_value = value.string_pair[language_index]
+            level_exception_flag = (category.string_pair[0], feature.position, lemma) in LEVEL_EXCEPTIONS
+            lexeme_level_flag = feature.lexeme_level_flag if not level_exception_flag else not feature.lexeme_level_flag
+            if (lexeme_level_flag):
+                self.lexeme_feature_map[feature_name] = feature_value
+            else:
+               self.form_feature_map[feature_name] = feature_value
+
+    def __str__(self):
+        return 'language={language}, category={category}, lexeme features={lexeme_features}, form_features={form_features}'.\
+            format(language=self.language, category=self.category, lexeme_features=str(self.lexeme_feature_map), form_features=str(self.form_feature_map))
+
+
+class MsdToProperties:
+
+    def convert(self, specifications, lemma, msd, msd_language, properties_language):
+        msd_language_index = LANGUAGE_INDEX_MAP[msd_language]
+        properties_language_index = LANGUAGE_INDEX_MAP[properties_language]
+        category_char = msd[0].lower()
+        value_chars = msd[1:]
+        category = specifications.find_category(category_char, msd_language_index)
+        feature_value_list = []
+        for (index, value_char) in enumerate(value_chars, start=1):
+            if (value_char != '-'):
+                feature = category.find_feature(index)
+                value = feature.find_value(value_char, msd_language_index)
+                feature_value_list.append((feature, value))
+        return Properties(lemma, category, feature_value_list, properties_language_index)

conversion_utils/utils.py

@@ -0,0 +1,10 @@
+TEI_NAMESPACE = 'http://www.tei-c.org/ns/1.0'
+XML_ID_ATTRIBUTE_NAME = '{http://www.w3.org/XML/1998/namespace}id'
+
+def xpath_find(element,expression):
+    """Executes XPath expression, with TEI namespace."""
+    return element.xpath(expression, namespaces={'tei':TEI_NAMESPACE})
+
+def get_xml_id(element):
+    """Returns the element's @xml:id attribute."""
+    return element.get(XML_ID_ATTRIBUTE_NAME)

tests/test_jos_msd_to_properties.py

@@ -0,0 +1,62 @@
+import os.path
+import lxml.etree as lxml
+import unittest
+
+from conversion_utils.jos_msds_and_properties import SpecificationsBuilder, MsdToProperties
+
+class JosMsdToPropertiesTestCase(unittest.TestCase):
+
+    def setUp(self):
+        specifications_file_name = os.path.join(os.path.dirname(__file__), '../resources/msd-sl.spc.xml')
+        builder = SpecificationsBuilder()
+        self.specifications = builder.build(specifications_file_name)
+        self.converter = MsdToProperties()
+
+    def test_en_en(self):
+        properties = self.converter.convert(self.specifications, 'miza', 'Ncfpd', 'en', 'en')
+        self.assertEqual(properties.language, 'en')
+        self.assertEqual(properties.category, 'noun')
+        self.assertEqual(properties.lexeme_feature_map, {'type':'common', 'gender':'feminine'})
+        self.assertEqual(properties.form_feature_map, {'number':'plural', 'case':'dative'})
+
+    def test_en_sl(self):
+        properties = self.converter.convert(self.specifications, 'miza', 'Ncfpd', 'en', 'sl')
+        self.assertEqual(properties.language, 'sl')
+        self.assertEqual(properties.category, 'samostalnik')
+        self.assertEqual(properties.lexeme_feature_map, {'vrsta':'občno_ime', 'spol':'ženski'})
+        self.assertEqual(properties.form_feature_map, {'število':'množina', 'sklon':'dajalnik'})
+
+    def test_sl_en(self):
+        properties = self.converter.convert(self.specifications, 'miza', 'Sozmd', 'sl', 'en')
+        self.assertEqual(properties.language, 'en')
+        self.assertEqual(properties.category, 'noun')
+        self.assertEqual(properties.lexeme_feature_map, {'type':'common', 'gender':'feminine'})
+        self.assertEqual(properties.form_feature_map, {'number':'plural', 'case':'dative'})
+
+    def test_sl_sl(self):
+        properties = self.converter.convert(self.specifications, 'miza', 'Sozmd', 'sl', 'sl')
+        self.assertEqual(properties.language, 'sl')
+        self.assertEqual(properties.category, 'samostalnik')
+        self.assertEqual(properties.lexeme_feature_map, {'vrsta':'občno_ime', 'spol':'ženski'})
+        self.assertEqual(properties.form_feature_map, {'število':'množina', 'sklon':'dajalnik'})
+
+    def test_exception_feature_level(self): # testing level exception: ('pronoun', 8, 'ti')
+        properties = self.converter.convert(self.specifications, 'ti', 'Pp2-sd--y', 'en', 'sl')
+        self.assertEqual(properties.language, 'sl')
+        self.assertEqual(properties.category, 'zaimek')
+        self.assertEqual(properties.lexeme_feature_map, {'vrsta':'osebni', 'oseba':'druga'})
+        self.assertEqual(properties.form_feature_map, {'število':'ednina', 'sklon':'dajalnik', 'naslonskost':'klitična'})
+
+    def test_normal_feature_level(self): # invalid msd, but useful for testing contrast with previous test
+        properties = self.converter.convert(self.specifications, 'vi', 'Pp2-pd--y', 'en', 'sl')
+        self.assertEqual(properties.language, 'sl')
+        self.assertEqual(properties.category, 'zaimek')
+        self.assertEqual(properties.lexeme_feature_map, {'vrsta':'osebni', 'oseba':'druga', 'naslonskost':'klitična'})
+        self.assertEqual(properties.form_feature_map, {'število':'množina', 'sklon':'dajalnik'})
+
+    def test_featureless(self):
+        properties = self.converter.convert(self.specifications, 'in', 'V', 'sl', 'en')
+        self.assertEqual(properties.language, 'en')
+        self.assertEqual(properties.category, 'conjunction')
+        self.assertEqual(properties.lexeme_feature_map, {})
+        self.assertEqual(properties.form_feature_map, {})