feat: first commit

README.md

@@ -0,0 +1,144 @@
+---
+annotations_creators:
+- expert-generated
+language_creators:
+- found
+languages:
+- en-US
+licenses:
+- bsd-3-clause
+multilinguality:
+- monolingual
+size_categories:
+- 10K<n<100K
+source_datasets:
+- 
+task_categories:
+- question-answering
+task_ids:
+- open-domain-qa
+- extractive-qa
+paperswithcode_id: quasar
+---
+
+# Dataset Card Creation Guide
+
+## Table of Contents
+- [Dataset Description](#dataset-description)
+  - [Dataset Summary](#dataset-summary)
+  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
+  - [Languages](#languages)
+- [Dataset Structure](#dataset-structure)
+  - [Data Instances](#data-instances)
+  - [Data Fields](#data-fields)
+  - [Data Splits](#data-splits)
+- [Dataset Creation](#dataset-creation)
+  - [Curation Rationale](#curation-rationale)
+  - [Source Data](#source-data)
+  - [Annotations](#annotations)
+  - [Personal and Sensitive Information](#personal-and-sensitive-information)
+- [Considerations for Using the Data](#considerations-for-using-the-data)
+  - [Social Impact of Dataset](#social-impact-of-dataset)
+  - [Discussion of Biases](#discussion-of-biases)
+  - [Other Known Limitations](#other-known-limitations)
+- [Additional Information](#additional-information)
+  - [Dataset Curators](#dataset-curators)
+  - [Licensing Information](#licensing-information)
+  - [Citation Information](#citation-information)
+  - [Contributions](#contributions)
+
+## Dataset Description
+
+- **Homepage:** N/A
+- **Repository:** [GitHub](https://github.com/mcobzarenco/mctest/)
+- **Paper:** [MCTest: A Challenge Dataset for the Open-Domain Machine Comprehension of Text](https://www.aclweb.org/anthology/D13-1020.pdf)
+- **Leaderboard:** N/A
+- **Point of Contact:** -
+
+### Dataset Summary
+
+[More Information Needed]
+
+### Supported Tasks and Leaderboards
+
+[More Information Needed]
+
+### Languages
+
+[More Information Needed]
+
+## Dataset Structure
+
+### Data Instances
+
+[More Information Needed]
+
+### Data Fields
+
+[More Information Needed]
+
+### Data Splits
+
+[More Information Needed]
+## Dataset Creation
+
+### Curation Rationale
+
+[More Information Needed]
+
+### Source Data
+
+[More Information Needed]
+
+#### Initial Data Collection and Normalization
+
+[More Information Needed]
+
+#### Who are the source language producers?
+
+[More Information Needed]
+
+### Annotations
+
+[More Information Needed]
+
+#### Annotation process
+
+[More Information Needed]
+
+#### Who are the annotators?
+
+[More Information Needed]
+
+### Personal and Sensitive Information
+
+[More Information Needed]
+
+## Considerations for Using the Data
+
+### Social Impact of Dataset
+
+[More Information Needed]
+
+### Discussion of Biases
+
+[More Information Needed]
+
+### Other Known Limitations
+
+[More Information Needed]
+
+## Additional Information
+
+### Dataset Curators
+
+[More Information Needed]
+
+### Licensing Information
+
+[More Information Needed]
+
+### Citation Information
+
+[More Information Needed]
+### Contributions

dataset_infos.json

@@ -0,0 +1 @@
+{"quasar-s": {"description": "We present two new large-scale datasets aimed at evaluating systems designed to comprehend a natural language query and extract its answer from a large corpus of text. The Quasar-S dataset consists of 37000 cloze-style (fill-in-the-gap) queries constructed from definitions of software entity tags on the popular website Stack Overflow. The posts and comments on the website serve as the background corpus for answering the cloze questions. The Quasar-T dataset consists of 43000 open-domain trivia questions and their answers obtained from various internet sources. ClueWeb09 serves as the background corpus for extracting these answers. We pose these datasets as a challenge for two related subtasks of factoid Question Answering: (1) searching for relevant pieces of text that include the correct answer to a query, and (2) reading the retrieved text to answer the query. \n", "citation": "@article{dhingra2017quasar,\n  title={Quasar: Datasets for Question Answering by Search and Reading},\n  author={Dhingra, Bhuwan and Mazaitis, Kathryn and Cohen, William W},\n  journal={arXiv preprint arXiv:1707.03904},\n  year={2017}\n}\n", "homepage": "https://github.com/bdhingra/quasar", "license": "", "features": {"uid": {"dtype": "string", "id": null, "_type": "Value"}, "question": {"dtype": "string", "id": null, "_type": "Value"}, "context_short": {"feature": {"confidence": {"dtype": "float32", "id": null, "_type": "Value"}, "content": {"dtype": "string", "id": null, "_type": "Value"}}, "length": -1, "id": null, "_type": "Sequence"}, "context_long": {"feature": {"confidence": {"dtype": "float32", "id": null, "_type": "Value"}, "content": {"dtype": "string", "id": null, "_type": "Value"}}, "length": -1, "id": null, "_type": "Sequence"}, "tags": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}, "relation": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "quasar", "config_name": "quasar-s", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 1187409339, "num_examples": 31049, "dataset_name": "quasar"}, "validation": {"name": "validation", "num_bytes": 120067134, "num_examples": 3139, "dataset_name": "quasar"}, "test": {"name": "test", "num_bytes": 120290406, "num_examples": 3174, "dataset_name": "quasar"}}, "download_checksums": {"http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/questions/train_questions.json.gz": {"num_bytes": 1957049, "checksum": "bc1540ca81df8bceb89c78c39e4f734bf19d08fc0e6f0893ae8b69fd7816a202"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/contexts/long/train_contexts.json.gz": {"num_bytes": 244642137, "checksum": "3a0bb6294ab54bc96ee3097cb98fbbd1b3e0f990c9a7812bd6d48e27416677b7"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/contexts/short/train_contexts.json.gz": {"num_bytes": 122615621, "checksum": "ebb1df435d899d560866daac6f9c91715414d2e6db1ea41f780d6b95780b23a9"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/questions/dev_questions.json.gz": {"num_bytes": 195290, "checksum": "ecb287ac7d862af7ced0a9e27320ff12b3a731a2642c18d7d209f0c5cb2d7958"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/contexts/long/dev_contexts.json.gz": {"num_bytes": 24782055, "checksum": "6ee2185911add3dbb6b2a3c81bb6d1ddec39d0d5b84607c599ccaacdda427eba"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/contexts/short/dev_contexts.json.gz": {"num_bytes": 12425372, "checksum": "17e726d5ec62847d42eb2bf9b28d370f39ba2212d1b40305f549ffa4308052de"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/questions/test_questions.json.gz": {"num_bytes": 192764, "checksum": "eca89e7f63e728d6bdd861f5dc68b5066a648f5a2d84ece730150cd8ceb8a0ca"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/contexts/long/test_contexts.json.gz": {"num_bytes": 24707451, "checksum": "2a68d90e137f46d67454fbdc9730b4da70e07c36531a385c37fafbc3753176f2"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/contexts/short/test_contexts.json.gz": {"num_bytes": 12382791, "checksum": "a62e667efebb6e168277e93c26668c75d8b39c787b21f6d1fe1768a40ee274c5"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-s/relation_annotations.json": {"num_bytes": 2833, "checksum": "ee8e90131357e0137425fc10c764476d0c12a4245ef9a8c59f4a9836a7be02aa"}}, "download_size": 443903363, "post_processing_size": null, "dataset_size": 1427766879, "size_in_bytes": 1871670242}, "quasar-t": {"description": "We present two new large-scale datasets aimed at evaluating systems designed to comprehend a natural language query and extract its answer from a large corpus of text. The Quasar-S dataset consists of 37000 cloze-style (fill-in-the-gap) queries constructed from definitions of software entity tags on the popular website Stack Overflow. The posts and comments on the website serve as the background corpus for answering the cloze questions. The Quasar-T dataset consists of 43000 open-domain trivia questions and their answers obtained from various internet sources. ClueWeb09 serves as the background corpus for extracting these answers. We pose these datasets as a challenge for two related subtasks of factoid Question Answering: (1) searching for relevant pieces of text that include the correct answer to a query, and (2) reading the retrieved text to answer the query. \n", "citation": "@article{dhingra2017quasar,\n  title={Quasar: Datasets for Question Answering by Search and Reading},\n  author={Dhingra, Bhuwan and Mazaitis, Kathryn and Cohen, William W},\n  journal={arXiv preprint arXiv:1707.03904},\n  year={2017}\n}\n", "homepage": "https://github.com/bdhingra/quasar", "license": "", "features": {"uid": {"dtype": "string", "id": null, "_type": "Value"}, "question": {"dtype": "string", "id": null, "_type": "Value"}, "context_short": {"feature": {"confidence": {"dtype": "float32", "id": null, "_type": "Value"}, "content": {"dtype": "string", "id": null, "_type": "Value"}}, "length": -1, "id": null, "_type": "Sequence"}, "context_long": {"feature": {"confidence": {"dtype": "float32", "id": null, "_type": "Value"}, "content": {"dtype": "string", "id": null, "_type": "Value"}}, "length": -1, "id": null, "_type": "Sequence"}, "tags": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}, "answer_type": {"dtype": "string", "id": null, "_type": "Value"}, "genre": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "quasar", "config_name": "quasar-t", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 1973207987, "num_examples": 37012, "dataset_name": "quasar"}, "validation": {"name": "validation", "num_bytes": 159766129, "num_examples": 3000, "dataset_name": "quasar"}, "test": {"name": "test", "num_bytes": 160121123, "num_examples": 3000, "dataset_name": "quasar"}}, "download_checksums": {"http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/questions/train_questions.json.gz": {"num_bytes": 1466304, "checksum": "ab3b68e842793dc3ed31839438986d63b7ef20a94bb347b9cd8644a6527d7840"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/long/train_contexts.json.gz": {"num_bytes": 445078274, "checksum": "aa2b5722f1003736919dca4de64b33f9066154fb357a7010d3a0fb4e11a4d2f2"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/short/train_contexts.json.gz": {"num_bytes": 172850853, "checksum": "76e9bd6c806c36136c55603e2439834863f0ffcb748ae6fe5aaac75d63c47f1b"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/questions/dev_questions.json.gz": {"num_bytes": 121433, "checksum": "d917cdcfef65b700225c41b863cd96b76f6c569f7a356aed60a54dfb7f515bc0"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/long/dev_contexts.json.gz": {"num_bytes": 36267682, "checksum": "56923cf7738e5b12e859eefbf07a29a1bb4cd6ddd0361c34b635f6f3550a825c"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/short/dev_contexts.json.gz": {"num_bytes": 13976824, "checksum": "22bf4221715c9aea49506cb578121ac26eff5c896d6199dfa2173be615241486"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/questions/test_questions.json.gz": {"num_bytes": 121488, "checksum": "0bb161f31aaac93058f0079f87d3be5ec928a74c741bd9d81e1005a3cdf9bd5a"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/long/test_contexts.json.gz": {"num_bytes": 35996879, "checksum": "039d8f57867820b4659b75d03cfd182f2d8179acee5a0b273d8a8e03dcaeadd3"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/short/test_contexts.json.gz": {"num_bytes": 14023655, "checksum": "cb967682e2ab06cad41f7a66c0e737a2e3f825b3684e30137848d7020737635a"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/answer_annotations.json": {"num_bytes": 1502, "checksum": "d410026ffe62557d289d7bd2f230c3af7695d6081cee9329db3de9143ad4ac26"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/genre_annotations.json": {"num_bytes": 2635, "checksum": "cc02cb14c8c9ccf8d07c94e2e845612a700277f4b0da578fee8fbc50642ff4f2"}}, "download_size": 719907529, "post_processing_size": null, "dataset_size": 2293095239, "size_in_bytes": 3013002768}, "quasar-t-nps": {"description": "We present two new large-scale datasets aimed at evaluating systems designed to comprehend a natural language query and extract its answer from a large corpus of text. The Quasar-S dataset consists of 37000 cloze-style (fill-in-the-gap) queries constructed from definitions of software entity tags on the popular website Stack Overflow. The posts and comments on the website serve as the background corpus for answering the cloze questions. The Quasar-T dataset consists of 43000 open-domain trivia questions and their answers obtained from various internet sources. ClueWeb09 serves as the background corpus for extracting these answers. We pose these datasets as a challenge for two related subtasks of factoid Question Answering: (1) searching for relevant pieces of text that include the correct answer to a query, and (2) reading the retrieved text to answer the query. \n", "citation": "@article{dhingra2017quasar,\n  title={Quasar: Datasets for Question Answering by Search and Reading},\n  author={Dhingra, Bhuwan and Mazaitis, Kathryn and Cohen, William W},\n  journal={arXiv preprint arXiv:1707.03904},\n  year={2017}\n}\n", "homepage": "https://github.com/bdhingra/quasar", "license": "", "features": {"uid": {"dtype": "string", "id": null, "_type": "Value"}, "question": {"dtype": "string", "id": null, "_type": "Value"}, "context_short": {"feature": {"confidence": {"dtype": "float32", "id": null, "_type": "Value"}, "content": {"dtype": "string", "id": null, "_type": "Value"}, "content_tokens": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "nps": {"feature": {"content": {"dtype": "string", "id": null, "_type": "Value"}, "start_token_id": {"dtype": "int32", "id": null, "_type": "Value"}}, "length": -1, "id": null, "_type": "Sequence"}}, "length": -1, "id": null, "_type": "Sequence"}, "context_long": {"feature": {"confidence": {"dtype": "float32", "id": null, "_type": "Value"}, "content": {"dtype": "string", "id": null, "_type": "Value"}, "content_tokens": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "nps": {"feature": {"content": {"dtype": "string", "id": null, "_type": "Value"}, "start_token_id": {"dtype": "int32", "id": null, "_type": "Value"}}, "length": -1, "id": null, "_type": "Sequence"}}, "length": -1, "id": null, "_type": "Sequence"}, "tags": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}, "answer_type": {"dtype": "string", "id": null, "_type": "Value"}, "genre": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "quasar", "config_name": "quasar-t-nps", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 6428881377, "num_examples": 37012, "dataset_name": "quasar"}, "validation": {"name": "validation", "num_bytes": 520694542, "num_examples": 3000, "dataset_name": "quasar"}, "test": {"name": "test", "num_bytes": 521524682, "num_examples": 3000, "dataset_name": "quasar"}}, "download_checksums": {"http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/questions/train_questions.json.gz": {"num_bytes": 1466304, "checksum": "ab3b68e842793dc3ed31839438986d63b7ef20a94bb347b9cd8644a6527d7840"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/long/train_contexts.json.gz": {"num_bytes": 445078274, "checksum": "aa2b5722f1003736919dca4de64b33f9066154fb357a7010d3a0fb4e11a4d2f2"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/short/train_contexts.json.gz": {"num_bytes": 172850853, "checksum": "76e9bd6c806c36136c55603e2439834863f0ffcb748ae6fe5aaac75d63c47f1b"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/questions/dev_questions.json.gz": {"num_bytes": 121433, "checksum": "d917cdcfef65b700225c41b863cd96b76f6c569f7a356aed60a54dfb7f515bc0"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/long/dev_contexts.json.gz": {"num_bytes": 36267682, "checksum": "56923cf7738e5b12e859eefbf07a29a1bb4cd6ddd0361c34b635f6f3550a825c"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/short/dev_contexts.json.gz": {"num_bytes": 13976824, "checksum": "22bf4221715c9aea49506cb578121ac26eff5c896d6199dfa2173be615241486"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/questions/test_questions.json.gz": {"num_bytes": 121488, "checksum": "0bb161f31aaac93058f0079f87d3be5ec928a74c741bd9d81e1005a3cdf9bd5a"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/long/test_contexts.json.gz": {"num_bytes": 35996879, "checksum": "039d8f57867820b4659b75d03cfd182f2d8179acee5a0b273d8a8e03dcaeadd3"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/short/test_contexts.json.gz": {"num_bytes": 14023655, "checksum": "cb967682e2ab06cad41f7a66c0e737a2e3f825b3684e30137848d7020737635a"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/answer_annotations.json": {"num_bytes": 1502, "checksum": "d410026ffe62557d289d7bd2f230c3af7695d6081cee9329db3de9143ad4ac26"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/genre_annotations.json": {"num_bytes": 2635, "checksum": "cc02cb14c8c9ccf8d07c94e2e845612a700277f4b0da578fee8fbc50642ff4f2"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/long/train_nps.json.gz": {"num_bytes": 377526504, "checksum": "99464c6edab03208dbb3b482ac499e9a49df9a69421eb9411cd73dbbae98ec53"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/short/train_nps.json.gz": {"num_bytes": 110182511, "checksum": "a69dbc82d1da05e7a9c0991623b3d3cc211421b8b85260ef9055d4b52adf1feb"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/long/dev_nps.json.gz": {"num_bytes": 30690232, "checksum": "50896eea80865d554f15c9b6b44d5a93903e06b060510dbedb9229567bae5477"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/short/dev_nps.json.gz": {"num_bytes": 8951977, "checksum": "67b99fd89cb6fe7aed9c9666a08b4e84538fd4a5a563c99e1db6d980fdcd1424"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/long/test_nps.json.gz": {"num_bytes": 30594385, "checksum": "0dc158ec67ff5e4ee32961eee6235b6bd90e399e28c130cd96bb17ef17f4f90b"}, "http://curtis.ml.cmu.edu/datasets/quasar/quasar-t/contexts/short/test_nps.json.gz": {"num_bytes": 8920932, "checksum": "3ce4a4fbb9de4f67c8728d5c96c7c077bec6811d508ea6a57862867ed71c7116"}}, "download_size": 1286774070, "post_processing_size": null, "dataset_size": 7471100601, "size_in_bytes": 8757874671}}

quasar.py

@@ -0,0 +1,356 @@
+# coding=utf-8
+# Copyright 2020 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.
+"""Quasar: Datasets for Question Answering by Search and Reading"""
+
+
+import gzip
+import datasets
+import json
+from collections import defaultdict
+from tqdm import tqdm
+
+_CITATION = """\
+@article{dhingra2017quasar,
+  title={Quasar: Datasets for Question Answering by Search and Reading},
+  author={Dhingra, Bhuwan and Mazaitis, Kathryn and Cohen, William W},
+  journal={arXiv preprint arXiv:1707.03904},
+  year={2017}
+}
+"""
+_UNKNOWN_RELATION = "UNK_RELATION"
+_UNKNOWN_ANS_TYPE = "UNK_ANS_TYPE"
+_UNKNOWN_GENRE = "UNK_GENRE"
+_QUASAR_S = "quasar-s"
+_QUASAR_T = "quasar-t"
+_QUASAR_T_NPS = "quasar-t-nps"
+_WHITE_SPACE = " "
+_DESCRIPTION = """\
+We present two new large-scale datasets aimed at evaluating systems designed to comprehend a natural language query and extract its answer from a large corpus of text. The Quasar-S dataset consists of 37000 cloze-style (fill-in-the-gap) queries constructed from definitions of software entity tags on the popular website Stack Overflow. The posts and comments on the website serve as the background corpus for answering the cloze questions. The Quasar-T dataset consists of 43000 open-domain trivia questions and their answers obtained from various internet sources. ClueWeb09 serves as the background corpus for extracting these answers. We pose these datasets as a challenge for two related subtasks of factoid Question Answering: (1) searching for relevant pieces of text that include the correct answer to a query, and (2) reading the retrieved text to answer the query. 
+"""
+
+_HOMEPAGE = "https://github.com/bdhingra/quasar"
+
+_DATA_URL = "http://curtis.ml.cmu.edu/datasets/quasar"
+
+QUASAR_S_DESC = """\
+Quasar-S consists of cloze style questions over software entities. The following information is provided.
+uid: Unique id
+question: Text of the question
+answer: Text of the answer
+context_short: List[{confidence: float, content: str}]
+context_long: The same as context_short, but from a different data source. see the paper for more info.
+relation: For some questions in Quasar-S, the relation type between head entity of the cloze question and the answer 
+entity is provided. For the other questions, this field takes the value "UNK_RELATION". For example,
+[question]: jarjar -- jar jar links http : code.google.com p @placeholder is a utility that 
+makes it easy to repackage java libraries and embed them into your own distribution ., 
+[answer]: jarjar, 
+[relationship]: synonym 
+"""
+
+QUASAR_T_DESC = """\
+The following information is provided.
+uid: unique id
+question: text of the question
+answer: text of the answer
+context_short: List[{confidence: float, content: str}]
+context_long: The same as context_short, but from a different data source. see the paper for more info.
+answer_type: Whether the answer is a date/time or number. This is known for some answers, for the others, this field 
+takes the value "UNK_ANS_TYPE"
+genre: Whether the question is from the genre of arts or math/science. This is known for some questions, for the others, 
+this field takes the value "UNK_GENRE"  
+"""
+
+QUASAR_T_NPS_DESC = """\
+Quasar-T consists of consists of trivia questions. The following information is provided.
+uid: unique id
+question: text of the question
+answer: text of the answer
+context_short: 
+List[
+  {
+    confidence: float, 
+    content: str, 
+    content_tokens: List[str],
+    nps: List[{'content': str, 'start_token_id': int}]
+  }
+]
+Here, context_tokens is a whitespace tokenization of content. `nps` are contiguous chunks of NN* tagged tokens from the 
+context as candidate answers.
+context_long: The same as context_short, but from a different data source. see the paper for more info.
+answer_type: Whether the answer is a date/time or number. This is known for some answers, for the others, this field 
+takes the value "UNK_ANS_TYPE"
+genre: Whether the question is from the genre of arts or math/science. This is known for some questions, for the others, 
+this field takes the value "UNK_GENRE"  
+"""
+
+
+class Quasar(datasets.GeneratorBasedBuilder):
+    """MCTest: Machine comprehension test: http://research.microsoft.com/mct"""
+
+    VERSION = datasets.Version("1.0.0")
+
+    BUILDER_CONFIGS = [
+        datasets.BuilderConfig(
+            name=_QUASAR_S,
+            version=VERSION,
+            description=QUASAR_S_DESC,
+        ),
+        datasets.BuilderConfig(
+            name=_QUASAR_T,
+            version=VERSION,
+            description=QUASAR_T_DESC,
+        ),
+        datasets.BuilderConfig(
+            name=_QUASAR_T_NPS,
+            version=VERSION,
+            description=QUASAR_T_NPS_DESC,
+        )
+    ]
+
+    DEFAULT_CONFIG_NAME = _QUASAR_S
+
+    def _info(self):
+        features = datasets.Features(
+            {
+                "uid": datasets.Value("string"),
+                "question": datasets.Value("string"),
+                "context_short": datasets.Sequence(
+                    dict(
+                        {
+                            "confidence": datasets.Value("float"),
+                            "content": datasets.Value("string")
+                        }
+                    )),
+                "context_long": datasets.Sequence(
+                    dict(
+                        {
+                            "confidence": datasets.Value("float"),
+                            "content": datasets.Value("string")
+                        }
+                    )),
+                "tags": datasets.Sequence(datasets.Value("string")),
+                "answer": datasets.Value("string"),
+            }
+        )
+        # for some questions in Quasar-S, relation type between head entity of the cloze question and the answer entity
+        # is provided. For the other questions, we provide an UNK
+
+        # [relationship]: component-of, [question]: putchar -- anything related to c or @placeholder functions putchar
+        # c or std : : putchar c++ ., [answer]: c++-standard-library
+
+        # [relationship]: synonym, [question]: jarjar -- jar jar links http : code.google.com p @placeholder is a
+        # utility that makes it easy to repackage java libraries and embed them into your own distribution .,
+        # [answer]: jarjar
+
+        # [relationship]: runs-on, [question]: web-audio -- web-audio is a javascript api providing low-level
+        # low-latency audio playback and manipulation functions in html5 capable @placeholder browsers ., [answer]: web
+
+        # [relationship]: used-with, [question]: audio-video-sync -- questions related to synchronization between audio
+        # and @placeholder during creation transmission reception and playback of content with both audio and video .,
+        # [answer]: video
+
+        if self.config.name == _QUASAR_S:
+            features.update({
+                "relation": datasets.Value("string")
+            })
+        elif self.config.name.startswith(_QUASAR_T):
+            features.update({
+                "answer_type": datasets.Value("string"),
+                "genre": datasets.Value("string")
+            })
+            # (only for quasar-T): We also provide contiguous chunks of
+            # NN* tagged tokens from the context as candidate answers (only for quasar-T).
+            # Again each line corresponds to the question in <split>_questions.json.gz,
+            # in the format:
+            # {
+            #     "nps": [
+            #         ...
+            #         [
+            #             "aerosol spray",
+            #             69,
+            #             29
+            #         ],
+            #     ],
+            #     "uid": "s3q41931"
+            # }
+            #
+            # Each element in "nps" is a list with three elements -
+            # [candidate, context_id, token_id]. The context_id is the index into the
+            # list of context documents, and token_id is the position of the start of
+            # the np in the context, when tokenized by white-space. Both are 0-based
+            # indices.
+            #
+            # If the correct answer is not detected as an NN* chunk we add it to the
+            # list of NPs above. The context_id and token_id are set to -1 in this
+            # case.
+
+            # since this will increase the size by quite a bit, we use a separate configuration for this, called
+            # quasar-t-nps
+            if self.config.name == _QUASAR_T_NPS:
+                for _type in ["short", "long"]:
+                    features[f"context_{_type}"] = datasets.Sequence(
+                        dict(
+                            {
+                                "confidence": datasets.Value("float"),
+                                "content": datasets.Value("string"),
+                                "content_tokens": datasets.Sequence(datasets.Value("string")),
+                                "nps": datasets.Sequence(dict(
+                                    {
+                                        "content": datasets.Value("string"),
+                                        "start_token_id": datasets.Value("int32")
+                                    }
+                                ))
+                            }
+                        )
+                    )
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=features,
+            homepage=_HOMEPAGE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        paths = {}
+        phases = ["train", "dev", "test"]
+        if self.config.name == _QUASAR_S:
+            data_path = f"{_DATA_URL}/{_QUASAR_S}"
+            for phase in phases:
+                paths[phase] = {
+                    "qa": dl_manager.download(f"{data_path}/questions/{phase}_questions.json.gz"),
+                    "contexts_long": dl_manager.download(f"{data_path}/contexts/long/{phase}_contexts.json.gz"),
+                    "contexts_short": dl_manager.download(f"{data_path}/contexts/short/{phase}_contexts.json.gz"),
+                }
+            paths["relations"] = dl_manager.download(f"{data_path}/relation_annotations.json")
+        elif self.config.name.startswith(_QUASAR_T):
+            data_path = f"{_DATA_URL}/{_QUASAR_T}"
+            for phase in phases:
+                paths[phase] = {
+                    "qa": dl_manager.download(f"{data_path}/questions/{phase}_questions.json.gz"),
+                    "contexts_long": dl_manager.download(f"{data_path}/contexts/long/{phase}_contexts.json.gz"),
+                    "contexts_short": dl_manager.download(f"{data_path}/contexts/short/{phase}_contexts.json.gz"),
+                }
+            paths["answer_types"] = dl_manager.download(f"{data_path}/answer_annotations.json")
+            paths["genres"] = dl_manager.download(f"{data_path}/genre_annotations.json")
+            if self.config.name == _QUASAR_T_NPS:
+                for phase in phases:
+                    paths[phase].update(
+                        {
+                            "nps_long": dl_manager.download(f"{data_path}/contexts/long/{phase}_nps.json.gz"),
+                            "nps_short": dl_manager.download(f"{data_path}/contexts/short/{phase}_nps.json.gz"),
+                        }
+                    )
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={"filepath": paths, "phase": "train"},
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.VALIDATION,
+                gen_kwargs={"filepath": paths, "phase": "dev"},
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={"filepath": paths, "phase": "test"},
+            ),
+        ]
+
+    @staticmethod
+    def _read_file(path):
+        """
+        read a json.gz file
+        :param path:
+        :return:
+        """
+        with gzip.open(path) as rf:
+            for line in rf:
+                yield eval(line)
+
+    @staticmethod
+    def _invert_dict(_dict):
+        """
+        converts a dict of Dict[str, List[str]] to Dict[str, str], where each key in the new dict is one of the
+        values in the original dict
+        :param _dict:
+        :return:
+        """
+        _d = {}
+        for k, v in _dict.items():
+            for _v in v:
+                _d[_v] = k
+        return _d
+
+    @staticmethod
+    def _get_nps(nps, context_sentences):
+        np_sentence_dict = defaultdict(list)
+        for candidate, context_id, token_id in nps:
+            np_sentence_dict[context_id].append((candidate, token_id))
+        _context_sentences = [{
+            "confidence": context_sentence["confidence"],
+            "content": context_sentence["content"],
+            "content_tokens": context_sentence["content"].split(_WHITE_SPACE),
+            "nps": [{"content": np[0], "start_token_id": np[1]} for np in np_sentence_dict[index]]
+        } for index, context_sentence in enumerate(context_sentences)]
+        return _context_sentences
+
+    @staticmethod
+    def _get_base_datum(qa, context_long, context_short):
+        uid = qa["uid"]
+        assert context_long["uid"] == uid
+        assert context_short["uid"] == uid
+        context_long = [{"confidence": context[0], "content": context[1]} for context in context_long["contexts"]]
+        context_short = [{"confidence": context[0], "content": context[1]} for context in context_short["contexts"]]
+        return {
+            "uid": qa["uid"],
+            "question": qa["question"],
+            "context_short": context_short,
+            "context_long": context_long,
+            "tags": qa["tags"],
+            "answer": qa["answer"]
+        }
+
+    def _generate_examples(self, filepath, phase):
+        qas = self._read_file(filepath[phase]["qa"])
+        contexts_long = self._read_file(filepath[phase]["contexts_long"])
+        contexts_short = self._read_file(filepath[phase]["contexts_short"])
+        if self.config.name == _QUASAR_S:
+            relations = self._invert_dict(json.load(open(filepath["relations"])))
+            for qa, context_long, context_short in zip(qas, contexts_long, contexts_short):
+                datum = self._get_base_datum(qa, context_long, context_short)
+                datum.update({"relation": relations.get(qa["uid"], _UNKNOWN_RELATION)})
+                yield qa["uid"], datum
+        elif self.config.name == _QUASAR_T:
+            answer_types = self._invert_dict(json.load(open(filepath["answer_types"])))
+            genres = self._invert_dict(json.load(open(filepath["genres"])))
+            for qa, context_long, context_short in zip(qas, contexts_long, contexts_short):
+                datum = self._get_base_datum(qa, context_long, context_short)
+                datum.update({"answer_type": answer_types.get(qa["uid"], _UNKNOWN_ANS_TYPE)})
+                datum.update({"genre": genres.get(qa["uid"], _UNKNOWN_GENRE)})
+                yield qa["uid"], datum
+        elif self.config.name == _QUASAR_T_NPS:
+            answer_types = self._invert_dict(json.load(open(filepath["answer_types"])))
+            genres = self._invert_dict(json.load(open(filepath["genres"])))
+            nps_long = self._read_file(filepath[phase]["nps_long"])
+            nps_short = self._read_file(filepath[phase]["nps_short"])
+            for qa, context_long, context_short, np_long, np_short in zip(qas, contexts_long, contexts_short, nps_long,
+                                                                          nps_short):
+                datum = self._get_base_datum(qa, context_long, context_short)
+                assert np_long["uid"] == qa["uid"]
+                assert np_short["uid"] == qa["uid"]
+                datum.update({"answer_type": answer_types.get(qa["uid"], _UNKNOWN_ANS_TYPE)})
+                datum.update({"genre": genres.get(qa["uid"], _UNKNOWN_GENRE)})
+                datum["context_long"] = self._get_nps(np_long["nps"], datum["context_long"])
+                datum["context_short"] = self._get_nps(np_short["nps"], datum["context_short"])
+                yield qa["uid"], datum