def setUp(self): super().setUp() question_tokens = [ Token(x) for x in ['what', 'was', 'the', 'last', 'year', '2013', '?'] ] self.table_file = self.FIXTURES_ROOT / 'data' / 'wikitables' / 'sample_table.tsv' self.table_kg = TableQuestionKnowledgeGraph.read_from_file( self.table_file, question_tokens) self.world = WikiTablesVariableFreeWorld(self.table_kg) table_file_with_date = self.FIXTURES_ROOT / 'data' / 'wikitables' / 'sample_table_with_date.tsv' table_kg_with_date = TableQuestionKnowledgeGraph.read_from_file( table_file_with_date, question_tokens) self.world_with_date = WikiTablesVariableFreeWorld(table_kg_with_date)
def test_world_parses_logical_forms_with_decimals(self): question_tokens = [Token(x) for x in ['0.2']] table_kg = TableQuestionKnowledgeGraph.read_from_file( "tests/fixtures/data/wikitables/sample_table.tsv", question_tokens) world = WikiTablesWorld(table_kg) sempre_form = "(fb:cell.cell.number (number 0.200))" expression = world.parse_logical_form(sempre_form) assert str(expression) == "I1(I(num:0_200))"
def test_world_parses_logical_forms_with_decimals(self): question_tokens = [Token(x) for x in ['0.2']] table_kg = TableQuestionKnowledgeGraph.read_from_file( self.FIXTURES_ROOT / "data" / "wikitables" / "sample_table.tsv", question_tokens) world = WikiTablesWorld(table_kg) sempre_form = "(fb:cell.cell.number (number 0.200))" expression = world.parse_logical_form(sempre_form) f = types.name_mapper.get_alias assert str(expression) == f"{f('fb:cell.cell.number')}({f('number')}(num:0_200))"
def setUp(self): super().setUp() self.world_without_recursion = FakeWorldWithoutRecursion() self.world_with_recursion = FakeWorldWithRecursion() question_tokens = [Token(x) for x in ['what', 'was', 'the', 'last', 'year', '2004', '?']] table_file = self.FIXTURES_ROOT / 'data' / 'wikitables' / 'sample_table.tsv' table_kg = TableQuestionKnowledgeGraph.read_from_file(table_file, question_tokens) self.wikitables_world = WikiTablesWorld(table_kg)
def test_world_parses_logical_forms_with_decimals(self): question_tokens = [Token(x) for x in [u'0.2']] table_kg = TableQuestionKnowledgeGraph.read_from_file( self.FIXTURES_ROOT / u"data" / u"wikitables" / u"sample_table.tsv", question_tokens) world = WikiTablesWorld(table_kg) sempre_form = u"(fb:cell.cell.number (number 0.200))" expression = world.parse_logical_form(sempre_form) assert unicode(expression) == u"I1(I(num:0_200))"
def setUp(self): super(TestWikiTablesWorld, self).setUp() question_tokens = [ Token(x) for x in [u'what', u'was', u'the', u'last', u'year', u'2000', u'?'] ] self.table_file = self.FIXTURES_ROOT / u'data' / u'wikitables' / u'sample_table.tsv' self.table_kg = TableQuestionKnowledgeGraph.read_from_file( self.table_file, question_tokens) self.world = WikiTablesWorld(self.table_kg)
def setUp(self): super().setUp() question_tokens = [ Token(x) for x in ['what', 'was', 'the', 'last', 'year', '2000', '?'] ] self.table_file = 'tests/fixtures/data/wikitables/sample_table.tsv' self.table_kg = TableQuestionKnowledgeGraph.read_from_file( self.table_file, question_tokens) self.world = WikiTablesWorld(self.table_kg)
def test_world_parses_logical_forms_with_decimals(self): question_tokens = [Token(x) for x in ['0.2']] table_kg = TableQuestionKnowledgeGraph.read_from_file( self.FIXTURES_ROOT / "data" / "wikitables" / "sample_table.tsv", question_tokens) world = WikiTablesWorld(table_kg) sempre_form = "(fb:cell.cell.number (number 0.200))" expression = world.parse_logical_form(sempre_form) f = types.name_mapper.get_alias assert str(expression ) == f"{f('fb:cell.cell.number')}({f('number')}(num:0_200))"
def setUp(self): super(TestWorld, self).setUp() self.world_without_recursion = FakeWorldWithoutRecursion() self.world_with_recursion = FakeWorldWithRecursion() test_filename = self.FIXTURES_ROOT / u"data" / u"nlvr" / u"sample_ungrouped_data.jsonl" data = [json.loads(line)[u"structured_rep"] for line in open(test_filename).readlines()] self.nlvr_world = NlvrWorld(data[0]) question_tokens = [Token(x) for x in [u'what', u'was', u'the', u'last', u'year', u'2004', u'?']] table_file = self.FIXTURES_ROOT / u'data' / u'wikitables' / u'sample_table.tsv' table_kg = TableQuestionKnowledgeGraph.read_from_file(table_file, question_tokens) self.wikitables_world = WikiTablesWorld(table_kg)
def setUp(self): super().setUp() self.world_without_recursion = FakeWorldWithoutRecursion() self.world_with_recursion = FakeWorldWithRecursion() test_filename = self.FIXTURES_ROOT / "data" / "nlvr" / "sample_ungrouped_data.jsonl" data = [json.loads(line)["structured_rep"] for line in open(test_filename).readlines()] self.nlvr_world = NlvrWorld(data[0]) question_tokens = [Token(x) for x in ['what', 'was', 'the', 'last', 'year', '2004', '?']] table_file = self.FIXTURES_ROOT / 'data' / 'wikitables' / 'sample_table.tsv' table_kg = TableQuestionKnowledgeGraph.read_from_file(table_file, question_tokens) self.wikitables_world = WikiTablesWorld(table_kg)
def test_with_deeply_nested_logical_form(self): question_tokens = [Token(x) for x in ['what', 'was', 'the', 'district', '?']] table_filename = 'tests/fixtures/data/wikitables/table/109.tsv' table_kg = TableQuestionKnowledgeGraph.read_from_file(table_filename, question_tokens) world = WikiTablesWorld(table_kg) logical_form = ("(count ((reverse fb:cell.cell.number) (or (or (or (or (or (or (or (or " "(or (or (or (or (or (or (or (or (or (or (or (or (or fb:cell.virginia_1 " "fb:cell.virginia_10) fb:cell.virginia_11) fb:cell.virginia_12) " "fb:cell.virginia_13) fb:cell.virginia_14) fb:cell.virginia_15) " "fb:cell.virginia_16) fb:cell.virginia_17) fb:cell.virginia_18) " "fb:cell.virginia_19) fb:cell.virginia_2) fb:cell.virginia_20) " "fb:cell.virginia_21) fb:cell.virginia_22) fb:cell.virginia_3) " "fb:cell.virginia_4) fb:cell.virginia_5) fb:cell.virginia_6) " "fb:cell.virginia_7) fb:cell.virginia_8) fb:cell.virginia_9)))") print("Parsing...") world.parse_logical_form(logical_form)
def test_with_deeply_nested_logical_form(self): question_tokens = [Token(x) for x in ['what', 'was', 'the', 'district', '?']] table_filename = self.FIXTURES_ROOT / 'data' / 'wikitables' / 'table' / '109.tsv' table_kg = TableQuestionKnowledgeGraph.read_from_file(table_filename, question_tokens) world = WikiTablesWorld(table_kg) logical_form = ("(count ((reverse fb:cell.cell.number) (or (or (or (or (or (or (or (or " "(or (or (or (or (or (or (or (or (or (or (or (or (or fb:cell.virginia_1 " "fb:cell.virginia_10) fb:cell.virginia_11) fb:cell.virginia_12) " "fb:cell.virginia_13) fb:cell.virginia_14) fb:cell.virginia_15) " "fb:cell.virginia_16) fb:cell.virginia_17) fb:cell.virginia_18) " "fb:cell.virginia_19) fb:cell.virginia_2) fb:cell.virginia_20) " "fb:cell.virginia_21) fb:cell.virginia_22) fb:cell.virginia_3) " "fb:cell.virginia_4) fb:cell.virginia_5) fb:cell.virginia_6) " "fb:cell.virginia_7) fb:cell.virginia_8) fb:cell.virginia_9)))") print("Parsing...") world.parse_logical_form(logical_form)
def test_world_adds_numbers_from_question(self): question_tokens = [Token(x) for x in ['what', '2007', '2,107', '0.2', '1800s', '1950s', '?']] table_kg = TableQuestionKnowledgeGraph.read_from_file("tests/fixtures/data/wikitables/sample_table.tsv", question_tokens) world = WikiTablesWorld(table_kg) valid_actions = world.get_valid_actions() assert 'n -> 2007' in valid_actions['n'] assert 'n -> 2107' in valid_actions['n'] # It appears that sempre normalizes floating point numbers. assert 'n -> 0.200' in valid_actions['n'] # We want to add the end-points to things like "1800s": 1800 and 1900. assert 'n -> 1800' in valid_actions['n'] assert 'n -> 1900' in valid_actions['n'] assert 'n -> 1950' in valid_actions['n'] assert 'n -> 1960' in valid_actions['n']
def test_world_adds_numbers_from_question(self): question_tokens = [Token(x) for x in ['what', '2007', '2,107', '0.2', '1800s', '1950s', '?']] table_kg = TableQuestionKnowledgeGraph.read_from_file( self.FIXTURES_ROOT / "data" / "wikitables" / "sample_table.tsv", question_tokens) world = WikiTablesWorld(table_kg) valid_actions = world.get_valid_actions() assert 'n -> 2007' in valid_actions['n'] assert 'n -> 2107' in valid_actions['n'] # It appears that sempre normalizes floating point numbers. assert 'n -> 0.200' in valid_actions['n'] # We want to add the end-points to things like "1800s": 1800 and 1900. assert 'n -> 1800' in valid_actions['n'] assert 'n -> 1900' in valid_actions['n'] assert 'n -> 1950' in valid_actions['n'] assert 'n -> 1960' in valid_actions['n']
def setUp(self): super().setUp() self.world_without_recursion = FakeWorldWithoutRecursion() self.world_with_recursion = FakeWorldWithRecursion() test_filename = "tests/fixtures/data/nlvr/sample_ungrouped_data.jsonl" data = [ json.loads(line)["structured_rep"] for line in open(test_filename).readlines() ] self.nlvr_world = NlvrWorld(data[0]) question_tokens = [ Token(x) for x in ['what', 'was', 'the', 'last', 'year', '2004', '?'] ] table_file = 'tests/fixtures/data/wikitables/sample_table.tsv' table_kg = TableQuestionKnowledgeGraph.read_from_file( table_file, question_tokens) self.wikitables_world = WikiTablesWorld(table_kg)
def test_with_deeply_nested_logical_form(self): question_tokens = [ Token(x) for x in [u'what', u'was', u'the', u'district', u'?'] ] table_filename = self.FIXTURES_ROOT / u'data' / u'wikitables' / u'table' / u'109.tsv' table_kg = TableQuestionKnowledgeGraph.read_from_file( table_filename, question_tokens) world = WikiTablesWorld(table_kg) logical_form = ( u"(count ((reverse fb:cell.cell.number) (or (or (or (or (or (or (or (or " u"(or (or (or (or (or (or (or (or (or (or (or (or (or fb:cell.virginia_1 " u"fb:cell.virginia_10) fb:cell.virginia_11) fb:cell.virginia_12) " u"fb:cell.virginia_13) fb:cell.virginia_14) fb:cell.virginia_15) " u"fb:cell.virginia_16) fb:cell.virginia_17) fb:cell.virginia_18) " u"fb:cell.virginia_19) fb:cell.virginia_2) fb:cell.virginia_20) " u"fb:cell.virginia_21) fb:cell.virginia_22) fb:cell.virginia_3) " u"fb:cell.virginia_4) fb:cell.virginia_5) fb:cell.virginia_6) " u"fb:cell.virginia_7) fb:cell.virginia_8) fb:cell.virginia_9)))") print(u"Parsing...") world.parse_logical_form(logical_form)
def text_to_instance( self, # type: ignore question: str, table_info: Union[str, JsonDict], example_lisp_string: str = None, dpd_output: List[str] = None, tokenized_question: List[Token] = None) -> Instance: """ Reads text inputs and makes an instance. WikitableQuestions dataset provides tables as TSV files, which we use for training. For running a demo, we may want to provide tables in a JSON format. To make this method compatible with both, we take ``table_info``, which can either be a filename, or a dict. We check the argument's type and call the appropriate method in ``TableQuestionKnowledgeGraph``. Parameters ---------- question : ``str`` Input question table_info : ``str`` or ``JsonDict`` Table filename or the table content itself, as a dict. See ``TableQuestionKnowledgeGraph.read_from_json`` for the expected format. example_lisp_string : ``str``, optional The original (lisp-formatted) example string in the WikiTableQuestions dataset. This comes directly from the ``.examples`` file provided with the dataset. We pass this to SEMPRE for evaluating logical forms during training. It isn't otherwise used for anything. dpd_output : List[str], optional List of logical forms, produced by dynamic programming on denotations. Not required during test. tokenized_question : ``List[Token]``, optional If you have already tokenized the question, you can pass that in here, so we don't duplicate that work. You might, for example, do batch processing on the questions in the whole dataset, then pass the result in here. """ # pylint: disable=arguments-differ tokenized_question = tokenized_question or self._tokenizer.tokenize( question.lower()) question_field = TextField(tokenized_question, self._question_token_indexers) if isinstance(table_info, str): table_knowledge_graph = TableQuestionKnowledgeGraph.read_from_file( table_info, tokenized_question) table_metadata = MetadataField(open(table_info).readlines()) else: table_knowledge_graph = TableQuestionKnowledgeGraph.read_from_json( table_info) table_metadata = MetadataField(table_info) table_field = KnowledgeGraphField( table_knowledge_graph, tokenized_question, self._table_token_indexers, tokenizer=self._tokenizer, feature_extractors=self._linking_feature_extractors, include_in_vocab=self._use_table_for_vocab, max_table_tokens=self._max_table_tokens) world = WikiTablesWorld(table_knowledge_graph) world_field = MetadataField(world) production_rule_fields: List[Field] = [] for production_rule in world.all_possible_actions(): _, rule_right_side = production_rule.split(' -> ') is_global_rule = not world.is_table_entity(rule_right_side) field = ProductionRuleField(production_rule, is_global_rule) production_rule_fields.append(field) action_field = ListField(production_rule_fields) fields = { 'question': question_field, 'table': table_field, 'world': world_field, 'actions': action_field } if self._include_table_metadata: fields['table_metadata'] = table_metadata if example_lisp_string: fields['example_lisp_string'] = MetadataField(example_lisp_string) # We'll make each target action sequence a List[IndexField], where the index is into # the action list we made above. We need to ignore the type here because mypy doesn't # like `action.rule` - it's hard to tell mypy that the ListField is made up of # ProductionRuleFields. action_map = { action.rule: i for i, action in enumerate(action_field.field_list) } # type: ignore if dpd_output: action_sequence_fields: List[Field] = [] for logical_form in dpd_output: if not self._should_keep_logical_form(logical_form): logger.debug(f'Question was: {question}') logger.debug(f'Table info was: {table_info}') continue try: expression = world.parse_logical_form(logical_form) except ParsingError as error: logger.debug( f'Parsing error: {error.message}, skipping logical form' ) logger.debug(f'Question was: {question}') logger.debug(f'Logical form was: {logical_form}') logger.debug(f'Table info was: {table_info}') continue except: logger.error(logical_form) raise action_sequence = world.get_action_sequence(expression) try: index_fields: List[Field] = [] for production_rule in action_sequence: index_fields.append( IndexField(action_map[production_rule], action_field)) action_sequence_fields.append(ListField(index_fields)) except KeyError as error: logger.debug( f'Missing production rule: {error.args}, skipping logical form' ) logger.debug(f'Question was: {question}') logger.debug(f'Table info was: {table_info}') logger.debug(f'Logical form was: {logical_form}') continue if len(action_sequence_fields) >= self._max_dpd_logical_forms: break if not action_sequence_fields: # This is not great, but we're only doing it when we're passed logical form # supervision, so we're expecting labeled logical forms, but we can't actually # produce the logical forms. We should skip this instance. Note that this affects # _dev_ and _test_ instances, too, so your metrics could be over-estimates on the # full test data. return None fields['target_action_sequences'] = ListField( action_sequence_fields) if self._output_agendas: agenda_index_fields: List[Field] = [] for agenda_string in world.get_agenda(): agenda_index_fields.append( IndexField(action_map[agenda_string], action_field)) if not agenda_index_fields: agenda_index_fields = [IndexField(-1, action_field)] fields['agenda'] = ListField(agenda_index_fields) return Instance(fields)
def _get_world_with_question_tokens(self, tokens): table_kg = TableQuestionKnowledgeGraph.read_from_file( self.table_file, tokens) world = WikiTablesWorld(table_kg) return world
def setUp(self): super().setUp() question_tokens = [Token(x) for x in ['what', 'was', 'the', 'last', 'year', '2000', '?']] self.table_file = self.FIXTURES_ROOT / 'data' / 'wikitables' / 'sample_table.tsv' self.table_kg = TableQuestionKnowledgeGraph.read_from_file(self.table_file, question_tokens) self.world = WikiTablesWorld(self.table_kg)
def _get_world_with_question_tokens(self, tokens: List[Token]) -> WikiTablesWorld: table_kg = TableQuestionKnowledgeGraph.read_from_file(self.table_file, tokens) world = WikiTablesWorld(table_kg) return world
def _get_world_with_question_tokens( self, tokens: List[Token]) -> WikiTablesVariableFreeWorld: table_kg = TableQuestionKnowledgeGraph.read_from_file( self.table_file, tokens) world = WikiTablesVariableFreeWorld(table_kg) return world