# coding=utf-8 # Copyright 2020 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # Lint as: python3 """SQUALL: Lexical-level Supervised Table Question Answering Dataset.""" import json import re import datasets from datasets.tasks import QuestionAnsweringExtractive logger = datasets.logging.get_logger(__name__) _CITATION = """\ @inproceedings{Shi:Zhao:Boyd-Graber:Daume-III:Lee-2020, Title = {On the Potential of Lexico-logical Alignments for Semantic Parsing to {SQL} Queries}, Author = {Tianze Shi and Chen Zhao and Jordan Boyd-Graber and Hal {Daum\'{e} III} and Lillian Lee}, Booktitle = {Findings of EMNLP}, Year = {2020}, } """ _DESCRIPTION = """\ To explore the utility of fine-grained, lexical-level supervision, authors \ introduce SQUALL, a dataset that enriches 11,276 WikiTableQuestions \ English-language questions with manually created SQL equivalents plus \ alignments between SQL and question fragments. """ _URL = "https://raw.githubusercontent.com/tzshi/squall/main/data/" _URLS = { "squall": _URL + "squall.json", "wtq-test": _URL + "wtq-test.json", "dev-0": _URL + "dev-0.ids", "dev-1": _URL + "dev-1.ids", "dev-2": _URL + "dev-2.ids", "dev-3": _URL + "dev-3.ids", "dev-4": _URL + "dev-4.ids", } class SquallConfig(datasets.BuilderConfig): """BuilderConfig for Squall.""" def __init__(self, **kwargs): """BuilderConfig for Squall. Args: **kwargs: keyword arguments forwarded to super. """ super(SquallConfig, self).__init__(**kwargs) class Squall(datasets.GeneratorBasedBuilder): """SQUALL: Lexical-level Supervised Table Question Answering Dataset.""" BUILDER_CONFIGS = [ SquallConfig(name = '0'), SquallConfig(name = '1'), SquallConfig(name = '2'), SquallConfig(name = '3'), SquallConfig(name = '4') ] def _info(self): return datasets.DatasetInfo( description=_DESCRIPTION, features=datasets.Features( { "nt": datasets.Value("string"), "tbl": datasets.Value("string"), "columns": { "raw_header": datasets.features.Sequence(datasets.Value("string")), "tokenized_header": datasets.features.Sequence(datasets.features.Sequence(datasets.Value("string"))), "column_suffixes": datasets.features.Sequence(datasets.features.Sequence(datasets.Value("string"))), "column_dtype": datasets.features.Sequence(datasets.Value("string")), "example": datasets.features.Sequence(datasets.Value("string")) }, "nl": datasets.features.Sequence(datasets.Value("string")), "nl_pos": datasets.features.Sequence(datasets.Value("string")), "nl_ner": datasets.features.Sequence(datasets.Value("string")), "nl_incolumns": datasets.features.Sequence(datasets.Value("bool_")), "nl_incells": datasets.features.Sequence(datasets.Value("bool_")), "columns_innl": datasets.features.Sequence(datasets.Value("bool_")), "tgt": datasets.Value("string"), "sql": { "sql_type": datasets.features.Sequence(datasets.Value("string")), "value": datasets.features.Sequence(datasets.Value("string")), "span_indices": datasets.features.Sequence(datasets.features.Sequence(datasets.Value("int32"))) }, "nl_ralign": { "aligned_sql_token_type":datasets.features.Sequence(datasets.Value("string")), "aligned_sql_token_info":datasets.features.Sequence(datasets.Value("string")), }, "align":{ "nl_indices": datasets.features.Sequence(datasets.features.Sequence(datasets.Value("int32"))), "sql_indices": datasets.features.Sequence(datasets.features.Sequence(datasets.Value("int32"))) } } ), # No default supervised_keys (as we have to pass both question # and context as input). supervised_keys=None, homepage="https://github.com/tzshi/squall/", citation=_CITATION, task_templates=[ QuestionAnsweringExtractive( question_column="nl", context_column="columns", answers_column="tgt" ) ], ) def _split_generators(self, dl_manager): urls_to_download = { "squall": _URLS["squall"], "wtq-test": _URLS["wtq-test"], "dev-0": _URLS["dev-0"], "dev-1": _URLS["dev-1"], "dev-2": _URLS["dev-2"], "dev-3": _URLS["dev-3"], "dev-4": _URLS["dev-4"], } downloaded_files = dl_manager.download_and_extract(urls_to_download) return [ datasets.SplitGenerator( name=datasets.Split.TRAIN, gen_kwargs={"split_key": "train", "filepath": downloaded_files}), datasets.SplitGenerator( name=datasets.Split.VALIDATION, gen_kwargs={"split_key": "dev", "filepath": downloaded_files}), datasets.SplitGenerator( name=datasets.Split.TEST, gen_kwargs={"split_key": "test", "filepath": downloaded_files}), ] def _generate_examples(self, split_key, filepath): """This function returns the examples in the raw (text) form.""" logger.info("generating examples from = %s", filepath) squall_full = filepath["squall"] dev_ids = filepath["dev-" + self.config.name] test = filepath["wtq-test"] # transform the original squall data structure (list of things) to dict def transform(sample, sample_key, keys): cols = {} n_col = len(sample[sample_key]) for k in range(len(keys)): tmp = [] for j in range(n_col): tmp.append(sample[sample_key][j][k]) cols[keys[k]] = tmp return cols columns_keys = ["raw_header", "tokenized_header", "column_suffixes", "column_dtype", "example"] sql_keys = ["sql_type", "value", "span_indices"] align_keys = ["nl_indices", "sql_indices"] ralign_keys = ["aligned_sql_token_type", "aligned_sql_token_info"] with open(squall_full, encoding="utf-8") as f: squall_full_data = json.load(f) NUM_MAPPING = { 'half': 0.5, 'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6, 'seven': 7, 'eight': 8, 'nine': 9, 'ten': 10, 'eleven': 11, 'twelve': 12, 'twenty': 20, 'thirty': 30, 'once': 1, 'twice': 2, 'first': 1, 'second': 2, 'third': 3, 'fourth': 4, 'fifth': 5, 'sixth': 6, 'seventh': 7, 'eighth': 8, 'ninth': 9, 'tenth': 10, 'hundred': 100, 'thousand': 1000, 'million': 1000000, 'jan': 1, 'feb': 2, 'mar': 3, 'apr': 4, 'may': 5, 'jun': 6, 'jul': 7, 'aug': 8, 'sep': 9, 'oct': 10, 'nov': 11, 'dec': 12, 'january': 1, 'february': 2, 'march': 3, 'april': 4, 'june': 6, 'july': 7, 'august': 8, 'september': 9, 'october': 10, 'november': 11, 'december': 12, } def parse_number(s): if s in NUM_MAPPING: return NUM_MAPPING[s] s = s.replace(',', '') # https://stackoverflow.com/questions/4289331/python-extract-numbers-from-a-string ret = re.findall(r"[-+]?[.]?[\d]+(?:,\d\d\d)*[\.]?\d*(?:[eE][-+]?\d+)?", s) if len(ret) > 0: return ret[0] return None for instance in squall_full_data: has_number = False numbers = [] for x in instance["nl"]: numbers.append(parse_number(x)) if numbers[-1] is not None: has_number = True instance["numbers"] = numbers instance["has_number"] = has_number if split_key != 'test': with open(dev_ids) as f: dev_ids = json.load(f) if split_key == "train": set = [x for x in squall_full_data if x["tbl"] not in dev_ids] else: set = [x for x in squall_full_data if x["tbl"] in dev_ids] idx = 0 for sample in set: # transform columns cols = transform(sample, "columns", columns_keys) # transform sql sqls = transform(sample, "sql", sql_keys) # transform align aligns = transform(sample, "align", align_keys) # transform ralign raligns = transform(sample, "nl_ralign", ralign_keys) yield idx, { "nt": sample["nt"], "tbl": sample["tbl"], "columns": cols, "nl": sample["nl"], "nl_pos": sample["nl_pos"], "nl_ner": sample["nl_ner"], "nl_ralign": raligns, "nl_incolumns": sample["nl_incolumns"], "nl_incells": sample["nl_incells"], "columns_innl": sample["columns_innl"], "tgt": sample["tgt"], "sql": sqls, "align": aligns } idx += 1 else: with open(test, encoding="utf-8") as f: test_data = json.load(f) idx = 0 for sample in test_data: # transform columns keys = ["raw_header", "tokenized_header", "column_suffixes", "column_dtype", "example"] cols = transform(sample, "columns", keys) yield idx, { "nt": sample["nt"], "tbl": sample["tbl"], "columns": cols, "nl": sample["nl"], "nl_pos": sample["nl_pos"], "nl_ner": sample["nl_ner"], "nl_ralign": {k:[] for k in ralign_keys}, "nl_incolumns": sample["nl_incolumns"], "nl_incells": sample["nl_incells"], "columns_innl": sample["columns_innl"], "tgt": '', "sql": {k:[] for k in sql_keys}, "align": {k:[] for k in align_keys} } idx += 1