Update files from the datasets library (from 1.2.0)

Release notes: https://github.com/huggingface/datasets/releases/tag/1.2.0

.gitattributes

@@ -0,0 +1,27 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bin.* filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,251 @@
+---
+annotations_creators:
+  convai2_inferred:
+  - machine-generated
+  funpedia:
+  - found
+  gendered_words:
+  - found
+  image_chat:
+  - found
+  light_inferred:
+  - machine-generated
+  name_genders:
+  - found
+  new_data:
+  - crowdsourced
+  - found
+  opensubtitles_inferred:
+  - machine-generated
+  wizard:
+  - found
+  yelp_inferred:
+  - machine-generated
+language_creators:
+  convai2_inferred:
+  - found
+  funpedia:
+  - found
+  gendered_words:
+  - found
+  image_chat:
+  - found
+  light_inferred:
+  - found
+  name_genders:
+  - found
+  new_data:
+  - crowdsourced
+  - found
+  opensubtitles_inferred:
+  - found
+  wizard:
+  - found
+  yelp_inferred:
+  - found
+languages:
+- en
+licenses:
+- mit
+multilinguality:
+- monolingual
+size_categories:
+  convai2_inferred:
+  - 100K<n<1M
+  funpedia:
+  - 10K<n<100K
+  gendered_words:
+  - n<1K
+  image_chat:
+  - 100K<n<1M
+  light_inferred:
+  - 100K<n<1M
+  name_genders:
+  - n>1M
+  new_data:
+  - 1K<n<10K
+  opensubtitles_inferred:
+  - 100K<n<1M
+  wizard:
+  - 10K<n<100K
+  yelp_inferred:
+  - n>1M
+source_datasets:
+  convai2_inferred:
+  - extended|other-convai2
+  - original
+  funpedia:
+  - original
+  gendered_words:
+  - original
+  image_chat:
+  - original
+  light_inferred:
+  - extended|other-light
+  - original
+  name_genders:
+  - original
+  new_data:
+  - original
+  opensubtitles_inferred:
+  - extended|other-opensubtitles
+  - original
+  wizard:
+  - original
+  yelp_inferred:
+  - extended|other-yelp
+  - original
+task_categories:
+- text-classification
+task_ids:
+- text-classification-other-gender-bias
+---
+
+# Dataset Card for Multi-Dimensional Gender Bias Classification
+
+## Table of Contents
+- [Dataset Description](#dataset-description)
+  - [Dataset Summary](#dataset-summary)
+  - [Supported Tasks](#supported-tasks-and-leaderboards)
+  - [Languages](#languages)
+- [Dataset Structure](#dataset-structure)
+  - [Data Instances](#data-instances)
+  - [Data Fields](#data-instances)
+  - [Data Splits](#data-instances)
+- [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)
+
+## Dataset Description
+
+- **Homepage:** https://parl.ai/projects/md_gender/
+- **Repository:** [Needs More Information]
+- **Paper:** https://arxiv.org/abs/2005.00614
+- **Leaderboard:** [Needs More Information]
+- **Point of Contact:** edinan@fb.com
+
+### Dataset Summary
+
+Machine learning models are trained to find patterns in data.
+NLP models can inadvertently learn socially undesirable patterns when training on gender biased text.
+In this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:
+bias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.
+Using this fine-grained framework, we automatically annotate eight large scale datasets with gender information.
+In addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.
+Distinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.
+We show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,
+detecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.
+
+### Supported Tasks and Leaderboards
+
+[Needs More Information]
+
+### Languages
+
+The data is in English (`en`)
+
+## Dataset Structure
+
+### Data Instances
+
+[Needs More Information]
+
+### Data Fields
+
+The data has the following features.
+
+For the `new_data` config:
+- `text`: the text to be classified
+- `original`: the text before reformulation
+- `labels`: a `list` of classification labels, with possible values including `ABOUT:female`, `ABOUT:male`, `PARTNER:female`, `PARTNER:male`, `SELF:female`.
+- `class_type`: a classification label, with possible values including `about`, `partner`, `self`.
+- `turker_gender`: a classification label, with possible values including `man`, `woman`, `nonbinary`, `prefer not to say`, `no answer`.
+
+For the other annotated datasets:
+- `text`: the text to be classified.
+- `gender`: a classification label, with possible values including `gender-neutral`, `female`, `male`.
+
+For the `_inferred` configurations:
+- `text`: the text to be classified.
+- `binary_label`: a classification label, with possible values including `ABOUT:female`, `ABOUT:male`.
+- `binary_score`: a score between 0 and 1.
+- `ternary_label`: a classification label, with possible values including `ABOUT:female`, `ABOUT:male`, `ABOUT:gender-neutral`.
+- `ternary_score`: a score between 0 and 1.
+
+### Data Splits
+
+The different parts of the data can be accessed through the different configurations:
+- `gendered_words`: A list of common nouns with a masculine and feminine variant.
+- `new_data`: Sentences reformulated and annotated along all three axes.
+- `funpedia`, `wizard`: Sentences from Funpedia and Wizards of Wikipedia annotated with ABOUT gender with entity gender information.
+- `image_chat`: sentences about images annotated  with ABOUT gender based on gender information from the entities in the image
+- `convai2_inferred`, `light_inferred`, `opensubtitles_inferred`, `yelp_inferred`:  Data from several source datasets with ABOUT annotations inferred by a trined classifier.
+
+
+## Dataset Creation
+
+### Curation Rationale
+
+[Needs More Information]
+
+### Source Data
+
+#### Initial Data Collection and Normalization
+
+[Needs More Information]
+
+#### Who are the source language producers?
+
+[Needs More Information]
+
+### Annotations
+
+#### Annotation process
+
+[Needs More Information]
+
+#### Who are the annotators?
+
+[Needs More Information]
+
+### Personal and Sensitive Information
+
+[Needs More Information]
+
+## Considerations for Using the Data
+
+### Social Impact of Dataset
+
+[Needs More Information]
+
+### Discussion of Biases
+
+[Needs More Information]
+
+### Other Known Limitations
+
+[Needs More Information]
+
+## Additional Information
+
+### Dataset Curators
+
+[Needs More Information]
+
+### Licensing Information
+
+[Needs More Information]
+
+### Citation Information
+
+[Needs More Information]

dataset_infos.json

@@ -0,0 +1 @@
+{"gendered_words": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"word_masculine": {"dtype": "string", "id": null, "_type": "Value"}, "word_feminine": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "gendered_words", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 4988, "num_examples": 222, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 4988, "size_in_bytes": 232633998}, "name_genders": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"name": {"dtype": "string", "id": null, "_type": "Value"}, "assigned_gender": {"num_classes": 2, "names": ["M", "F"], "names_file": null, "id": null, "_type": "ClassLabel"}, "count": {"dtype": "int32", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "name_genders", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"yob1880": {"name": "yob1880", "num_bytes": 43404, "num_examples": 2000, "dataset_name": "md_gender_bias"}, "yob1881": {"name": "yob1881", "num_bytes": 41944, "num_examples": 1935, "dataset_name": "md_gender_bias"}, "yob1882": {"name": "yob1882", "num_bytes": 46211, "num_examples": 2127, "dataset_name": "md_gender_bias"}, "yob1883": {"name": "yob1883", "num_bytes": 45221, "num_examples": 2084, "dataset_name": "md_gender_bias"}, "yob1884": {"name": "yob1884", "num_bytes": 49886, "num_examples": 2297, "dataset_name": "md_gender_bias"}, "yob1885": {"name": "yob1885", "num_bytes": 49810, "num_examples": 2294, "dataset_name": "md_gender_bias"}, "yob1886": {"name": "yob1886", "num_bytes": 51935, "num_examples": 2392, "dataset_name": "md_gender_bias"}, "yob1887": {"name": "yob1887", "num_bytes": 51458, "num_examples": 2373, "dataset_name": "md_gender_bias"}, "yob1888": {"name": "yob1888", "num_bytes": 57531, "num_examples": 2651, "dataset_name": "md_gender_bias"}, "yob1889": {"name": "yob1889", "num_bytes": 56177, "num_examples": 2590, "dataset_name": "md_gender_bias"}, "yob1890": {"name": "yob1890", "num_bytes": 58509, "num_examples": 2695, "dataset_name": "md_gender_bias"}, "yob1891": {"name": "yob1891", "num_bytes": 57767, "num_examples": 2660, "dataset_name": "md_gender_bias"}, "yob1892": {"name": "yob1892", "num_bytes": 63493, "num_examples": 2921, "dataset_name": "md_gender_bias"}, "yob1893": {"name": "yob1893", "num_bytes": 61525, "num_examples": 2831, "dataset_name": "md_gender_bias"}, "yob1894": {"name": "yob1894", "num_bytes": 63927, "num_examples": 2941, "dataset_name": "md_gender_bias"}, "yob1895": {"name": "yob1895", "num_bytes": 66346, "num_examples": 3049, "dataset_name": "md_gender_bias"}, "yob1896": {"name": "yob1896", "num_bytes": 67224, "num_examples": 3091, "dataset_name": "md_gender_bias"}, "yob1897": {"name": "yob1897", "num_bytes": 65886, "num_examples": 3028, "dataset_name": "md_gender_bias"}, "yob1898": {"name": "yob1898", "num_bytes": 71088, "num_examples": 3264, "dataset_name": "md_gender_bias"}, "yob1899": {"name": "yob1899", "num_bytes": 66225, "num_examples": 3042, "dataset_name": "md_gender_bias"}, "yob1900": {"name": "yob1900", "num_bytes": 81305, "num_examples": 3730, "dataset_name": "md_gender_bias"}, "yob1901": {"name": "yob1901", "num_bytes": 68723, "num_examples": 3153, "dataset_name": "md_gender_bias"}, "yob1902": {"name": "yob1902", "num_bytes": 73321, "num_examples": 3362, "dataset_name": "md_gender_bias"}, "yob1903": {"name": "yob1903", "num_bytes": 74019, "num_examples": 3389, "dataset_name": "md_gender_bias"}, "yob1904": {"name": "yob1904", "num_bytes": 77751, "num_examples": 3560, "dataset_name": "md_gender_bias"}, "yob1905": {"name": "yob1905", "num_bytes": 79802, "num_examples": 3655, "dataset_name": "md_gender_bias"}, "yob1906": {"name": "yob1906", "num_bytes": 79392, "num_examples": 3633, "dataset_name": "md_gender_bias"}, "yob1907": {"name": "yob1907", "num_bytes": 86342, "num_examples": 3948, "dataset_name": "md_gender_bias"}, "yob1908": {"name": "yob1908", "num_bytes": 87965, "num_examples": 4018, "dataset_name": "md_gender_bias"}, "yob1909": {"name": "yob1909", "num_bytes": 92591, "num_examples": 4227, "dataset_name": "md_gender_bias"}, "yob1910": {"name": "yob1910", "num_bytes": 101491, "num_examples": 4629, "dataset_name": "md_gender_bias"}, "yob1911": {"name": "yob1911", "num_bytes": 106787, "num_examples": 4867, "dataset_name": "md_gender_bias"}, "yob1912": {"name": "yob1912", "num_bytes": 139448, "num_examples": 6351, "dataset_name": "md_gender_bias"}, "yob1913": {"name": "yob1913", "num_bytes": 153110, "num_examples": 6968, "dataset_name": "md_gender_bias"}, "yob1914": {"name": "yob1914", "num_bytes": 175167, "num_examples": 7965, "dataset_name": "md_gender_bias"}, "yob1915": {"name": "yob1915", "num_bytes": 205921, "num_examples": 9357, "dataset_name": "md_gender_bias"}, "yob1916": {"name": "yob1916", "num_bytes": 213468, "num_examples": 9696, "dataset_name": "md_gender_bias"}, "yob1917": {"name": "yob1917", "num_bytes": 218446, "num_examples": 9913, "dataset_name": "md_gender_bias"}, "yob1918": {"name": "yob1918", "num_bytes": 229209, "num_examples": 10398, "dataset_name": "md_gender_bias"}, "yob1919": {"name": "yob1919", "num_bytes": 228656, "num_examples": 10369, "dataset_name": "md_gender_bias"}, "yob1920": {"name": "yob1920", "num_bytes": 237286, "num_examples": 10756, "dataset_name": "md_gender_bias"}, "yob1921": {"name": "yob1921", "num_bytes": 239616, "num_examples": 10857, "dataset_name": "md_gender_bias"}, "yob1922": {"name": "yob1922", "num_bytes": 237569, "num_examples": 10756, "dataset_name": "md_gender_bias"}, "yob1923": {"name": "yob1923", "num_bytes": 235046, "num_examples": 10643, "dataset_name": "md_gender_bias"}, "yob1924": {"name": "yob1924", "num_bytes": 240113, "num_examples": 10869, "dataset_name": "md_gender_bias"}, "yob1925": {"name": "yob1925", "num_bytes": 235098, "num_examples": 10638, "dataset_name": "md_gender_bias"}, "yob1926": {"name": "yob1926", "num_bytes": 230970, "num_examples": 10458, "dataset_name": "md_gender_bias"}, "yob1927": {"name": "yob1927", "num_bytes": 230004, "num_examples": 10406, "dataset_name": "md_gender_bias"}, "yob1928": {"name": "yob1928", "num_bytes": 224583, "num_examples": 10159, "dataset_name": "md_gender_bias"}, "yob1929": {"name": "yob1929", "num_bytes": 217057, "num_examples": 9820, "dataset_name": "md_gender_bias"}, "yob1930": {"name": "yob1930", "num_bytes": 216352, "num_examples": 9791, "dataset_name": "md_gender_bias"}, "yob1931": {"name": "yob1931", "num_bytes": 205361, "num_examples": 9298, "dataset_name": "md_gender_bias"}, "yob1932": {"name": "yob1932", "num_bytes": 207268, "num_examples": 9381, "dataset_name": "md_gender_bias"}, "yob1933": {"name": "yob1933", "num_bytes": 199031, "num_examples": 9013, "dataset_name": "md_gender_bias"}, "yob1934": {"name": "yob1934", "num_bytes": 202758, "num_examples": 9180, "dataset_name": "md_gender_bias"}, "yob1935": {"name": "yob1935", "num_bytes": 199614, "num_examples": 9037, "dataset_name": "md_gender_bias"}, "yob1936": {"name": "yob1936", "num_bytes": 196379, "num_examples": 8894, "dataset_name": "md_gender_bias"}, "yob1937": {"name": "yob1937", "num_bytes": 197757, "num_examples": 8946, "dataset_name": "md_gender_bias"}, "yob1938": {"name": "yob1938", "num_bytes": 199603, "num_examples": 9032, "dataset_name": "md_gender_bias"}, "yob1939": {"name": "yob1939", "num_bytes": 196979, "num_examples": 8918, "dataset_name": "md_gender_bias"}, "yob1940": {"name": "yob1940", "num_bytes": 198141, "num_examples": 8961, "dataset_name": "md_gender_bias"}, "yob1941": {"name": "yob1941", "num_bytes": 200858, "num_examples": 9085, "dataset_name": "md_gender_bias"}, "yob1942": {"name": "yob1942", "num_bytes": 208363, "num_examples": 9425, "dataset_name": "md_gender_bias"}, "yob1943": {"name": "yob1943", "num_bytes": 207940, "num_examples": 9408, "dataset_name": "md_gender_bias"}, "yob1944": {"name": "yob1944", "num_bytes": 202227, "num_examples": 9152, "dataset_name": "md_gender_bias"}, "yob1945": {"name": "yob1945", "num_bytes": 199478, "num_examples": 9025, "dataset_name": "md_gender_bias"}, "yob1946": {"name": "yob1946", "num_bytes": 214614, "num_examples": 9705, "dataset_name": "md_gender_bias"}, "yob1947": {"name": "yob1947", "num_bytes": 229327, "num_examples": 10371, "dataset_name": "md_gender_bias"}, "yob1948": {"name": "yob1948", "num_bytes": 226615, "num_examples": 10241, "dataset_name": "md_gender_bias"}, "yob1949": {"name": "yob1949", "num_bytes": 227278, "num_examples": 10269, "dataset_name": "md_gender_bias"}, "yob1950": {"name": "yob1950", "num_bytes": 227946, "num_examples": 10303, "dataset_name": "md_gender_bias"}, "yob1951": {"name": "yob1951", "num_bytes": 231613, "num_examples": 10462, "dataset_name": "md_gender_bias"}, "yob1952": {"name": "yob1952", "num_bytes": 235483, "num_examples": 10646, "dataset_name": "md_gender_bias"}, "yob1953": {"name": "yob1953", "num_bytes": 239654, "num_examples": 10837, "dataset_name": "md_gender_bias"}, "yob1954": {"name": "yob1954", "num_bytes": 242389, "num_examples": 10968, "dataset_name": "md_gender_bias"}, "yob1955": {"name": "yob1955", "num_bytes": 245652, "num_examples": 11115, "dataset_name": "md_gender_bias"}, "yob1956": {"name": "yob1956", "num_bytes": 250674, "num_examples": 11340, "dataset_name": "md_gender_bias"}, "yob1957": {"name": "yob1957", "num_bytes": 255370, "num_examples": 11564, "dataset_name": "md_gender_bias"}, "yob1958": {"name": "yob1958", "num_bytes": 254520, "num_examples": 11522, "dataset_name": "md_gender_bias"}, "yob1959": {"name": "yob1959", "num_bytes": 260051, "num_examples": 11767, "dataset_name": "md_gender_bias"}, "yob1960": {"name": "yob1960", "num_bytes": 263474, "num_examples": 11921, "dataset_name": "md_gender_bias"}, "yob1961": {"name": "yob1961", "num_bytes": 269493, "num_examples": 12182, "dataset_name": "md_gender_bias"}, "yob1962": {"name": "yob1962", "num_bytes": 270244, "num_examples": 12209, "dataset_name": "md_gender_bias"}, "yob1963": {"name": "yob1963", "num_bytes": 271872, "num_examples": 12282, "dataset_name": "md_gender_bias"}, "yob1964": {"name": "yob1964", "num_bytes": 274590, "num_examples": 12397, "dataset_name": "md_gender_bias"}, "yob1965": {"name": "yob1965", "num_bytes": 264889, "num_examples": 11952, "dataset_name": "md_gender_bias"}, "yob1966": {"name": "yob1966", "num_bytes": 269321, "num_examples": 12151, "dataset_name": "md_gender_bias"}, "yob1967": {"name": "yob1967", "num_bytes": 274867, "num_examples": 12397, "dataset_name": "md_gender_bias"}, "yob1968": {"name": "yob1968", "num_bytes": 286774, "num_examples": 12936, "dataset_name": "md_gender_bias"}, "yob1969": {"name": "yob1969", "num_bytes": 304909, "num_examples": 13749, "dataset_name": "md_gender_bias"}, "yob1970": {"name": "yob1970", "num_bytes": 328047, "num_examples": 14779, "dataset_name": "md_gender_bias"}, "yob1971": {"name": "yob1971", "num_bytes": 339657, "num_examples": 15295, "dataset_name": "md_gender_bias"}, "yob1972": {"name": "yob1972", "num_bytes": 342321, "num_examples": 15412, "dataset_name": "md_gender_bias"}, "yob1973": {"name": "yob1973", "num_bytes": 348414, "num_examples": 15682, "dataset_name": "md_gender_bias"}, "yob1974": {"name": "yob1974", "num_bytes": 361188, "num_examples": 16249, "dataset_name": "md_gender_bias"}, "yob1975": {"name": "yob1975", "num_bytes": 376491, "num_examples": 16944, "dataset_name": "md_gender_bias"}, "yob1976": {"name": "yob1976", "num_bytes": 386565, "num_examples": 17391, "dataset_name": "md_gender_bias"}, "yob1977": {"name": "yob1977", "num_bytes": 403994, "num_examples": 18175, "dataset_name": "md_gender_bias"}, "yob1978": {"name": "yob1978", "num_bytes": 405430, "num_examples": 18231, "dataset_name": "md_gender_bias"}, "yob1979": {"name": "yob1979", "num_bytes": 423423, "num_examples": 19039, "dataset_name": "md_gender_bias"}, "yob1980": {"name": "yob1980", "num_bytes": 432317, "num_examples": 19452, "dataset_name": "md_gender_bias"}, "yob1981": {"name": "yob1981", "num_bytes": 432980, "num_examples": 19475, "dataset_name": "md_gender_bias"}, "yob1982": {"name": "yob1982", "num_bytes": 437986, "num_examples": 19694, "dataset_name": "md_gender_bias"}, "yob1983": {"name": "yob1983", "num_bytes": 431531, "num_examples": 19407, "dataset_name": "md_gender_bias"}, "yob1984": {"name": "yob1984", "num_bytes": 434085, "num_examples": 19506, "dataset_name": "md_gender_bias"}, "yob1985": {"name": "yob1985", "num_bytes": 447113, "num_examples": 20085, "dataset_name": "md_gender_bias"}, "yob1986": {"name": "yob1986", "num_bytes": 460315, "num_examples": 20657, "dataset_name": "md_gender_bias"}, "yob1987": {"name": "yob1987", "num_bytes": 477677, "num_examples": 21406, "dataset_name": "md_gender_bias"}, "yob1988": {"name": "yob1988", "num_bytes": 499347, "num_examples": 22367, "dataset_name": "md_gender_bias"}, "yob1989": {"name": "yob1989", "num_bytes": 531020, "num_examples": 23775, "dataset_name": "md_gender_bias"}, "yob1990": {"name": "yob1990", "num_bytes": 552114, "num_examples": 24716, "dataset_name": "md_gender_bias"}, "yob1991": {"name": "yob1991", "num_bytes": 560932, "num_examples": 25109, "dataset_name": "md_gender_bias"}, "yob1992": {"name": "yob1992", "num_bytes": 568151, "num_examples": 25427, "dataset_name": "md_gender_bias"}, "yob1993": {"name": "yob1993", "num_bytes": 579778, "num_examples": 25966, "dataset_name": "md_gender_bias"}, "yob1994": {"name": "yob1994", "num_bytes": 580223, "num_examples": 25997, "dataset_name": "md_gender_bias"}, "yob1995": {"name": "yob1995", "num_bytes": 581949, "num_examples": 26080, "dataset_name": "md_gender_bias"}, "yob1996": {"name": "yob1996", "num_bytes": 589131, "num_examples": 26423, "dataset_name": "md_gender_bias"}, "yob1997": {"name": "yob1997", "num_bytes": 601284, "num_examples": 26970, "dataset_name": "md_gender_bias"}, "yob1998": {"name": "yob1998", "num_bytes": 621587, "num_examples": 27902, "dataset_name": "md_gender_bias"}, "yob1999": {"name": "yob1999", "num_bytes": 635355, "num_examples": 28552, "dataset_name": "md_gender_bias"}, "yob2000": {"name": "yob2000", "num_bytes": 662398, "num_examples": 29772, "dataset_name": "md_gender_bias"}, "yob2001": {"name": "yob2001", "num_bytes": 673111, "num_examples": 30274, "dataset_name": "md_gender_bias"}, "yob2002": {"name": "yob2002", "num_bytes": 679392, "num_examples": 30564, "dataset_name": "md_gender_bias"}, "yob2003": {"name": "yob2003", "num_bytes": 692931, "num_examples": 31185, "dataset_name": "md_gender_bias"}, "yob2004": {"name": "yob2004", "num_bytes": 711776, "num_examples": 32048, "dataset_name": "md_gender_bias"}, "yob2005": {"name": "yob2005", "num_bytes": 723065, "num_examples": 32549, "dataset_name": "md_gender_bias"}, "yob2006": {"name": "yob2006", "num_bytes": 757620, "num_examples": 34088, "dataset_name": "md_gender_bias"}, "yob2007": {"name": "yob2007", "num_bytes": 776893, "num_examples": 34961, "dataset_name": "md_gender_bias"}, "yob2008": {"name": "yob2008", "num_bytes": 779403, "num_examples": 35079, "dataset_name": "md_gender_bias"}, "yob2009": {"name": "yob2009", "num_bytes": 771032, "num_examples": 34709, "dataset_name": "md_gender_bias"}, "yob2010": {"name": "yob2010", "num_bytes": 756717, "num_examples": 34073, "dataset_name": "md_gender_bias"}, "yob2011": {"name": "yob2011", "num_bytes": 752804, "num_examples": 33908, "dataset_name": "md_gender_bias"}, "yob2012": {"name": "yob2012", "num_bytes": 748915, "num_examples": 33747, "dataset_name": "md_gender_bias"}, "yob2013": {"name": "yob2013", "num_bytes": 738288, "num_examples": 33282, "dataset_name": "md_gender_bias"}, "yob2014": {"name": "yob2014", "num_bytes": 737219, "num_examples": 33243, "dataset_name": "md_gender_bias"}, "yob2015": {"name": "yob2015", "num_bytes": 734183, "num_examples": 33121, "dataset_name": "md_gender_bias"}, "yob2016": {"name": "yob2016", "num_bytes": 731291, "num_examples": 33010, "dataset_name": "md_gender_bias"}, "yob2017": {"name": "yob2017", "num_bytes": 721444, "num_examples": 32590, "dataset_name": "md_gender_bias"}, "yob2018": {"name": "yob2018", "num_bytes": 708657, "num_examples": 32033, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 43393095, "size_in_bytes": 276022105}, "new_data": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"text": {"dtype": "string", "id": null, "_type": "Value"}, "original": {"dtype": "string", "id": null, "_type": "Value"}, "labels": [{"num_classes": 6, "names": ["ABOUT:female", "ABOUT:male", "PARTNER:female", "PARTNER:male", "SELF:female", "SELF:male"], "names_file": null, "id": null, "_type": "ClassLabel"}], "class_type": {"num_classes": 3, "names": ["about", "partner", "self"], "names_file": null, "id": null, "_type": "ClassLabel"}, "turker_gender": {"num_classes": 5, "names": ["man", "woman", "nonbinary", "prefer not to say", "no answer"], "names_file": null, "id": null, "_type": "ClassLabel"}, "episode_done": {"dtype": "bool_", "id": null, "_type": "Value"}, "confidence": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "new_data", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 369753, "num_examples": 2345, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 369753, "size_in_bytes": 232998763}, "funpedia": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"text": {"dtype": "string", "id": null, "_type": "Value"}, "title": {"dtype": "string", "id": null, "_type": "Value"}, "persona": {"dtype": "string", "id": null, "_type": "Value"}, "gender": {"num_classes": 3, "names": ["gender-neutral", "female", "male"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "funpedia", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 3225542, "num_examples": 23897, "dataset_name": "md_gender_bias"}, "validation": {"name": "validation", "num_bytes": 402205, "num_examples": 2984, "dataset_name": "md_gender_bias"}, "test": {"name": "test", "num_bytes": 396417, "num_examples": 2938, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 4024164, "size_in_bytes": 236653174}, "image_chat": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"caption": {"dtype": "string", "id": null, "_type": "Value"}, "id": {"dtype": "string", "id": null, "_type": "Value"}, "male": {"dtype": "bool_", "id": null, "_type": "Value"}, "female": {"dtype": "bool_", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "image_chat", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 1061285, "num_examples": 9997, "dataset_name": "md_gender_bias"}, "validation": {"name": "validation", "num_bytes": 35868670, "num_examples": 338180, "dataset_name": "md_gender_bias"}, "test": {"name": "test", "num_bytes": 530126, "num_examples": 5000, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 37460081, "size_in_bytes": 270089091}, "wizard": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"text": {"dtype": "string", "id": null, "_type": "Value"}, "chosen_topic": {"dtype": "string", "id": null, "_type": "Value"}, "gender": {"num_classes": 3, "names": ["gender-neutral", "female", "male"], "names_file": null, "id": null, "_type": "ClassLabel"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "wizard", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 1158785, "num_examples": 10449, "dataset_name": "md_gender_bias"}, "validation": {"name": "validation", "num_bytes": 57824, "num_examples": 537, "dataset_name": "md_gender_bias"}, "test": {"name": "test", "num_bytes": 53126, "num_examples": 470, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 1269735, "size_in_bytes": 233898745}, "convai2_inferred": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"text": {"dtype": "string", "id": null, "_type": "Value"}, "binary_label": {"num_classes": 2, "names": ["ABOUT:female", "ABOUT:male"], "names_file": null, "id": null, "_type": "ClassLabel"}, "binary_score": {"dtype": "float32", "id": null, "_type": "Value"}, "ternary_label": {"num_classes": 3, "names": ["ABOUT:female", "ABOUT:male", "ABOUT:gender-neutral"], "names_file": null, "id": null, "_type": "ClassLabel"}, "ternary_score": {"dtype": "float32", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "convai2_inferred", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 9853669, "num_examples": 131438, "dataset_name": "md_gender_bias"}, "validation": {"name": "validation", "num_bytes": 608046, "num_examples": 7801, "dataset_name": "md_gender_bias"}, "test": {"name": "test", "num_bytes": 608046, "num_examples": 7801, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 11069761, "size_in_bytes": 243698771}, "light_inferred": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"text": {"dtype": "string", "id": null, "_type": "Value"}, "binary_label": {"num_classes": 2, "names": ["ABOUT:female", "ABOUT:male"], "names_file": null, "id": null, "_type": "ClassLabel"}, "binary_score": {"dtype": "float32", "id": null, "_type": "Value"}, "ternary_label": {"num_classes": 3, "names": ["ABOUT:female", "ABOUT:male", "ABOUT:gender-neutral"], "names_file": null, "id": null, "_type": "ClassLabel"}, "ternary_score": {"dtype": "float32", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "light_inferred", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 10931355, "num_examples": 106122, "dataset_name": "md_gender_bias"}, "validation": {"name": "validation", "num_bytes": 679692, "num_examples": 6362, "dataset_name": "md_gender_bias"}, "test": {"name": "test", "num_bytes": 1375745, "num_examples": 12765, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 12986792, "size_in_bytes": 245615802}, "opensubtitles_inferred": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"text": {"dtype": "string", "id": null, "_type": "Value"}, "binary_label": {"num_classes": 2, "names": ["ABOUT:female", "ABOUT:male"], "names_file": null, "id": null, "_type": "ClassLabel"}, "binary_score": {"dtype": "float32", "id": null, "_type": "Value"}, "ternary_label": {"num_classes": 3, "names": ["ABOUT:female", "ABOUT:male", "ABOUT:gender-neutral"], "names_file": null, "id": null, "_type": "ClassLabel"}, "ternary_score": {"dtype": "float32", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "opensubtitles_inferred", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 27966476, "num_examples": 351036, "dataset_name": "md_gender_bias"}, "validation": {"name": "validation", "num_bytes": 3363802, "num_examples": 41957, "dataset_name": "md_gender_bias"}, "test": {"name": "test", "num_bytes": 3830528, "num_examples": 49108, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 35160806, "size_in_bytes": 267789816}, "yelp_inferred": {"description": "Machine learning models are trained to find patterns in data.\nNLP models can inadvertently learn socially undesirable patterns when training on gender biased text.\nIn this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:\nbias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.\nUsing this fine-grained framework, we automatically annotate eight large scale datasets with gender information.\nIn addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.\nDistinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.\nWe show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,\ndetecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.\n", "citation": "@inproceedings{md_gender_bias,\n  author    = {Emily Dinan and\n               Angela Fan and\n               Ledell Wu and\n               Jason Weston and\n               Douwe Kiela and\n               Adina Williams},\n  editor    = {Bonnie Webber and\n               Trevor Cohn and\n               Yulan He and\n               Yang Liu},\n  title     = {Multi-Dimensional Gender Bias Classification},\n  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural\n               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},\n  pages     = {314--331},\n  publisher = {Association for Computational Linguistics},\n  year      = {2020},\n  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}\n}\n", "homepage": "https://parl.ai/projects/md_gender/", "license": "MIT License", "features": {"text": {"dtype": "string", "id": null, "_type": "Value"}, "binary_label": {"num_classes": 2, "names": ["ABOUT:female", "ABOUT:male"], "names_file": null, "id": null, "_type": "ClassLabel"}, "binary_score": {"dtype": "float32", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "md_gender_bias", "config_name": "yelp_inferred", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 260582945, "num_examples": 2577862, "dataset_name": "md_gender_bias"}, "validation": {"name": "validation", "num_bytes": 324349, "num_examples": 4492, "dataset_name": "md_gender_bias"}, "test": {"name": "test", "num_bytes": 53887700, "num_examples": 534460, "dataset_name": "md_gender_bias"}}, "download_checksums": {"http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz": {"num_bytes": 232629010, "checksum": "c2c03257c53497b9e453600201fc7245b55dec1d98965093b4657fdb54822e9d"}}, "download_size": 232629010, "post_processing_size": null, "dataset_size": 314794994, "size_in_bytes": 547424004}}

dummy/convai2_inferred/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd5fe75c22d3ad2aa9be6f40c3048a67d950f7173d28484772b3bc4ffab379f1
+size 21054

dummy/funpedia/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:588d083b3055694557ded3c576b14a9a21948ad2b45f5520a02046363493cfbb
+size 21054

dummy/gendered_words/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c03c0e060bfacad4bc7ea29ca4d0087288042552f5345ece6415cfac819f836
+size 21054

dummy/image_chat/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c85a5e125c11927219c6078230747be0ab4ed8283a3cc71a97b49debe3d872dc
+size 21054

dummy/light_inferred/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e587a526d23eb565a4f93a76ad7a0a877efb9c7a337ad99e31d942810d64693
+size 21054

dummy/name_genders/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31924a7d5477faf2db6945c9ea41e4ed03f90e6f543a94a56e718c884a0ce5ed
+size 56490

dummy/new_data/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7d76ecbe7368e6e1cafb960bcff9688b34495c36f2ead68e41542f69ee5f6b7
+size 21054

dummy/opensubtitles_inferred/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4453ed97b5c1069a5d95f5089569c7ded6cd6222d1ae79826cc0d9958e58a54c
+size 21054

dummy/wizard/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bf7813fee98e70c908c3eca27b33081447ba09ae4b79e23949b118f16b2c4b9
+size 21054

dummy/yelp_inferred/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ffa7355ecd8215f9e3d50f5cc7b63cac07bef916f04829bca4e131d97a7abc1
+size 21054

md_gender_bias.py

@@ -0,0 +1,410 @@
+# coding=utf-8
+# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# 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.
+"""Multi-Dimensional Gender Bias classification"""
+
+from __future__ import absolute_import, division, print_function
+
+import json
+import os
+
+import datasets
+
+
+# TODO: Add BibTeX citation
+# Find for instance the citation on arxiv or on the dataset repo/website
+_CITATION = """\
+@inproceedings{md_gender_bias,
+  author    = {Emily Dinan and
+               Angela Fan and
+               Ledell Wu and
+               Jason Weston and
+               Douwe Kiela and
+               Adina Williams},
+  editor    = {Bonnie Webber and
+               Trevor Cohn and
+               Yulan He and
+               Yang Liu},
+  title     = {Multi-Dimensional Gender Bias Classification},
+  booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural
+               Language Processing, {EMNLP} 2020, Online, November 16-20, 2020},
+  pages     = {314--331},
+  publisher = {Association for Computational Linguistics},
+  year      = {2020},
+  url       = {https://www.aclweb.org/anthology/2020.emnlp-main.23/}
+}
+"""
+
+# TODO: Add description of the dataset here
+# You can copy an official description
+_DESCRIPTION = """\
+Machine learning models are trained to find patterns in data.
+NLP models can inadvertently learn socially undesirable patterns when training on gender biased text.
+In this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:
+bias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.
+Using this fine-grained framework, we automatically annotate eight large scale datasets with gender information.
+In addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.
+Distinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.
+We show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,
+detecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.
+"""
+
+_HOMEPAGE = "https://parl.ai/projects/md_gender/"
+
+_LICENSE = "MIT License"
+
+_URL = "http://parl.ai/downloads/md_gender/gend_multiclass_10072020.tgz"
+
+_CONF_FILES = {
+    "funpedia": {
+        "train": "funpedia/train.jsonl",
+        "validation": "funpedia/valid.jsonl",
+        "test": "funpedia/test.jsonl",
+    },
+    "image_chat": {
+        "train": "image_chat/engaging_imagechat_gender_captions_hashed.test.jsonl",
+        "validation": "image_chat/engaging_imagechat_gender_captions_hashed.train.jsonl",
+        "test": "image_chat/engaging_imagechat_gender_captions_hashed.valid.jsonl",
+    },
+    "wizard": {
+        "train": "wizard/train.jsonl",
+        "validation": "wizard/valid.jsonl",
+        "test": "wizard/test.jsonl",
+    },
+    "convai2_inferred": {
+        "train": (
+            "inferred_about/convai2_train_binary.txt",
+            "inferred_about/convai2_train.txt",
+        ),
+        "validation": (
+            "inferred_about/convai2_valid_binary.txt",
+            "inferred_about/convai2_valid.txt",
+        ),
+        "test": (
+            "inferred_about/convai2_test_binary.txt",
+            "inferred_about/convai2_test.txt",
+        ),
+    },
+    "light_inferred": {
+        "train": (
+            "inferred_about/light_train_binary.txt",
+            "inferred_about/light_train.txt",
+        ),
+        "validation": (
+            "inferred_about/light_valid_binary.txt",
+            "inferred_about/light_valid.txt",
+        ),
+        "test": (
+            "inferred_about/light_test_binary.txt",
+            "inferred_about/light_test.txt",
+        ),
+    },
+    "opensubtitles_inferred": {
+        "train": (
+            "inferred_about/opensubtitles_train_binary.txt",
+            "inferred_about/opensubtitles_train.txt",
+        ),
+        "validation": (
+            "inferred_about/opensubtitles_valid_binary.txt",
+            "inferred_about/opensubtitles_valid.txt",
+        ),
+        "test": (
+            "inferred_about/opensubtitles_test_binary.txt",
+            "inferred_about/opensubtitles_test.txt",
+        ),
+    },
+    "yelp_inferred": {
+        "train": (
+            "inferred_about/yelp_train_binary.txt",
+            "",
+        ),
+        "validation": (
+            "inferred_about/yelp_valid_binary.txt",
+            "",
+        ),
+        "test": (
+            "inferred_about/yelp_test_binary.txt",
+            "",
+        ),
+    },
+}
+
+
+class MdGenderBias(datasets.GeneratorBasedBuilder):
+    """Multi-Dimensional Gender Bias classification"""
+
+    VERSION = datasets.Version("1.0.0")
+
+    BUILDER_CONFIGS = [
+        datasets.BuilderConfig(
+            name="gendered_words",
+            version=VERSION,
+            description="A list of common nouns with a masculine and feminine variant.",
+        ),
+        datasets.BuilderConfig(
+            name="name_genders",
+            version=VERSION,
+            description="A list of first names with their gender attribution by year in the US.",
+        ),
+        datasets.BuilderConfig(
+            name="new_data", version=VERSION, description="Some data reformulated and annotated along all three axes."
+        ),
+        datasets.BuilderConfig(
+            name="funpedia",
+            version=VERSION,
+            description="Data from Funpedia with ABOUT annotations based on Funpedia information on an entity's gender.",
+        ),
+        datasets.BuilderConfig(
+            name="image_chat",
+            version=VERSION,
+            description="Data from ImageChat with ABOUT annotations based on image recognition.",
+        ),
+        datasets.BuilderConfig(
+            name="wizard",
+            version=VERSION,
+            description="Data from WizardsOfWikipedia with ABOUT annotations based on Wikipedia information on an entity's gender.",
+        ),
+        datasets.BuilderConfig(
+            name="convai2_inferred",
+            version=VERSION,
+            description="Data from the ConvAI2 challenge with ABOUT annotations inferred by a trined classifier.",
+        ),
+        datasets.BuilderConfig(
+            name="light_inferred",
+            version=VERSION,
+            description="Data from LIGHT with ABOUT annotations inferred by a trined classifier.",
+        ),
+        datasets.BuilderConfig(
+            name="opensubtitles_inferred",
+            version=VERSION,
+            description="Data from OpenSubtitles with ABOUT annotations inferred by a trined classifier.",
+        ),
+        datasets.BuilderConfig(
+            name="yelp_inferred",
+            version=VERSION,
+            description="Data from Yelp reviews with ABOUT annotations inferred by a trined classifier.",
+        ),
+    ]
+
+    DEFAULT_CONFIG_NAME = (
+        "new_data"  # It's not mandatory to have a default configuration. Just use one if it make sense.
+    )
+
+    def _info(self):
+        # TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
+        if (
+            self.config.name == "gendered_words"
+        ):  # This is the name of the configuration selected in BUILDER_CONFIGS above
+            features = datasets.Features(
+                {
+                    "word_masculine": datasets.Value("string"),
+                    "word_feminine": datasets.Value("string"),
+                }
+            )
+        elif self.config.name == "name_genders":
+            features = datasets.Features(
+                {
+                    "name": datasets.Value("string"),
+                    "assigned_gender": datasets.ClassLabel(names=["M", "F"]),
+                    "count": datasets.Value("int32"),
+                }
+            )
+        elif self.config.name == "new_data":
+            features = datasets.Features(
+                {
+                    "text": datasets.Value("string"),
+                    "original": datasets.Value("string"),
+                    "labels": [
+                        datasets.ClassLabel(
+                            names=[
+                                "ABOUT:female",
+                                "ABOUT:male",
+                                "PARTNER:female",
+                                "PARTNER:male",
+                                "SELF:female",
+                                "SELF:male",
+                            ]
+                        )
+                    ],
+                    "class_type": datasets.ClassLabel(names=["about", "partner", "self"]),
+                    "turker_gender": datasets.ClassLabel(
+                        names=["man", "woman", "nonbinary", "prefer not to say", "no answer"]
+                    ),
+                    "episode_done": datasets.Value("bool_"),
+                    "confidence": datasets.Value("string"),
+                }
+            )
+        elif self.config.name == "funpedia":
+            features = datasets.Features(
+                {
+                    "text": datasets.Value("string"),
+                    "title": datasets.Value("string"),
+                    "persona": datasets.Value("string"),
+                    "gender": datasets.ClassLabel(names=["gender-neutral", "female", "male"]),
+                }
+            )
+        elif self.config.name == "image_chat":
+            features = datasets.Features(
+                {
+                    "caption": datasets.Value("string"),
+                    "id": datasets.Value("string"),
+                    "male": datasets.Value("bool_"),
+                    "female": datasets.Value("bool_"),
+                }
+            )
+        elif self.config.name == "wizard":
+            features = datasets.Features(
+                {
+                    "text": datasets.Value("string"),
+                    "chosen_topic": datasets.Value("string"),
+                    "gender": datasets.ClassLabel(names=["gender-neutral", "female", "male"]),
+                }
+            )
+        elif self.config.name == "yelp_inferred":
+            features = datasets.Features(
+                {
+                    "text": datasets.Value("string"),
+                    "binary_label": datasets.ClassLabel(names=["ABOUT:female", "ABOUT:male"]),
+                    "binary_score": datasets.Value("float"),
+                }
+            )
+        else:  # data with inferred labels
+            features = datasets.Features(
+                {
+                    "text": datasets.Value("string"),
+                    "binary_label": datasets.ClassLabel(names=["ABOUT:female", "ABOUT:male"]),
+                    "binary_score": datasets.Value("float"),
+                    "ternary_label": datasets.ClassLabel(names=["ABOUT:female", "ABOUT:male", "ABOUT:gender-neutral"]),
+                    "ternary_score": datasets.Value("float"),
+                }
+            )
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=features,  # Here we define them above because they are different between the two configurations
+            supervised_keys=None,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        """Returns SplitGenerators."""
+        data_dir = os.path.join(dl_manager.download_and_extract(_URL), "data_to_release")
+        if self.config.name == "gendered_words":
+            return [
+                datasets.SplitGenerator(
+                    name=datasets.Split.TRAIN,
+                    gen_kwargs={
+                        "filepath": None,
+                        "filepath_pair": (
+                            os.path.join(data_dir, "word_list/male_word_file.txt"),
+                            os.path.join(data_dir, "word_list/female_word_file.txt"),
+                        ),
+                    },
+                )
+            ]
+        elif self.config.name == "name_genders":
+            return [
+                datasets.SplitGenerator(
+                    name=f"yob{yob}",
+                    gen_kwargs={
+                        "filepath": os.path.join(data_dir, f"names/yob{yob}.txt"),
+                        "filepath_pair": None,
+                    },
+                )
+                for yob in range(1880, 2019)
+            ]
+        elif self.config.name == "new_data":
+            return [
+                datasets.SplitGenerator(
+                    name=datasets.Split.TRAIN,
+                    gen_kwargs={
+                        "filepath": os.path.join(data_dir, "new_data/data.jsonl"),
+                        "filepath_pair": None,
+                    },
+                )
+            ]
+        elif self.config.name in ["funpedia", "image_chat", "wizard"]:
+            return [
+                datasets.SplitGenerator(
+                    name=spl,
+                    gen_kwargs={
+                        "filepath": os.path.join(data_dir, fname),
+                        "filepath_pair": None,
+                    },
+                )
+                for spl, fname in _CONF_FILES[self.config.name].items()
+            ]
+        else:
+            return [
+                datasets.SplitGenerator(
+                    name=spl,
+                    gen_kwargs={
+                        "filepath": None,
+                        "filepath_pair": (
+                            os.path.join(data_dir, fname_1),
+                            os.path.join(data_dir, fname_2),
+                        ),
+                    },
+                )
+                for spl, (fname_1, fname_2) in _CONF_FILES[self.config.name].items()
+            ]
+
+    def _generate_examples(self, filepath, filepath_pair):
+        if self.config.name == "gendered_words":
+            with open(filepath_pair[0], encoding="utf-8") as f_m:
+                with open(filepath_pair[1], encoding="utf-8") as f_f:
+                    for id_, (l_m, l_f) in enumerate(zip(f_m, f_f)):
+                        yield id_, {
+                            "word_masculine": l_m.strip(),
+                            "word_feminine": l_f.strip(),
+                        }
+        elif self.config.name == "name_genders":
+            with open(filepath, encoding="utf-8") as f:
+                for id_, line in enumerate(f):
+                    name, g, ct = line.strip().split(",")
+                    yield id_, {
+                        "name": name,
+                        "assigned_gender": g,
+                        "count": int(ct),
+                    }
+        elif "_inferred" in self.config.name:
+            with open(filepath_pair[0], encoding="utf-8") as f_b:
+                if "yelp" in self.config.name:
+                    for id_, line_b in enumerate(f_b):
+                        text_b, label_b, score_b = line_b.split("\t")
+                        yield id_, {
+                            "text": text_b,
+                            "binary_label": label_b,
+                            "binary_score": float(score_b.strip()),
+                        }
+                else:
+                    with open(filepath_pair[1], encoding="utf-8") as f_t:
+                        for id_, (line_b, line_t) in enumerate(zip(f_b, f_t)):
+                            text_b, label_b, score_b = line_b.split("\t")
+                            text_t, label_t, score_t = line_t.split("\t")
+                            yield id_, {
+                                "text": text_b,
+                                "binary_label": label_b,
+                                "binary_score": float(score_b.strip()),
+                                "ternary_label": label_t,
+                                "ternary_score": float(score_t.strip()),
+                            }
+        else:
+            with open(filepath, encoding="utf-8") as f:
+                for id_, line in enumerate(f):
+                    example = json.loads(line.strip())
+                    if "turker_gender" in example and example["turker_gender"] is None:
+                        example["turker_gender"] = "no answer"
+                    yield id_, example