{ "overview": { "what": { "dataset": "Existing data-to-text generation datasets are mostly limited to English. Table-to-Text in African languages (TaTA) addresses this lack of data as the first large multilingual table-to-text dataset with a focus on African languages. TaTA was created by transcribing figures and accompanying text in bilingual reports by the Demographic and Health Surveys Program, followed by professional translation to make the dataset fully parallel. TaTA includes 8,700 examples in nine languages including four African languages (Hausa, Igbo, Swahili, and Yor\u00f9b\u00e1) and a zero-shot test language (Russian). " }, "where": { "has-leaderboard": "yes", "leaderboard-url": "[Github](https://github.com/google-research/url-nlp)", "leaderboard-description": "The paper introduces a metric StATA which is trained on human ratings and which is used to rank approaches submitted to the leaderboard.", "website": "[Github](https://github.com/google-research/url-nlp)", "data-url": "[Github](https://github.com/google-research/url-nlp)", "paper-url": "[ArXiv](https://arxiv.org/abs/2211.00142)", "paper-bibtext": "```\n@misc{gehrmann2022TaTA,\n Author = {Sebastian Gehrmann and Sebastian Ruder and Vitaly Nikolaev and Jan A. Botha and Michael Chavinda and Ankur Parikh and Clara Rivera},\n Title = {TaTa: A Multilingual Table-to-Text Dataset for African Languages},\n Year = {2022},\n Eprint = {arXiv:2211.00142},\n}\n```", "contact-name": "Sebastian Ruder", "contact-email": "ruder@google.com" }, "languages": { "is-multilingual": "yes", "license": "cc-by-sa-4.0: Creative Commons Attribution Share Alike 4.0 International", "task-other": "N/A", "language-names": [ "English", "Portuguese", "Arabic", "French", "Hausa", "Swahili (macrolanguage)", "Igbo", "Yoruba", "Russian" ], "language-speakers": "The language is taken from reports by the demographic and health surveys program.", "intended-use": "The dataset poses significant reasoning challenges and is thus meant as a way to asses the verbalization and reasoning capabilities of structure-to-text models.", "license-other": "N/A", "task": "Data-to-Text", "communicative": "Summarize key information from a table in a single sentence.\n" }, "credit": { "organization-type": [ "industry" ], "organization-names": "Google Research", "creators": "Sebastian Gehrmann, Sebastian Ruder , Vitaly Nikolaev, Jan A. Botha, Michael Chavinda, Ankur Parikh, Clara Rivera", "funding": "Google Research", "gem-added-by": "Sebastian Gehrmann (Google Research)" }, "structure": { "data-fields": "- `example_id`: The ID of the example. Each ID (e.g., `AB20-ar-1`) consists of three parts: the document ID, the language ISO 639-1 code, and the index of the table within the document.\n- `title`: The title of the table.\n- `unit_of_measure`: A description of the numerical value of the data. E.g., percentage of households with clean water.\n- `chart_type`: The kind of chart associated with the data. We consider the following (normalized) types: horizontal bar chart, map chart, pie graph, tables, line chart, pie chart, vertical chart type, line graph, vertical bar chart, and other.\n- `was_translated`: Whether the table was transcribed in the original language of the report or translated.\n- `table_data`: The table content is a JSON-encoded string of a two-dimensional list, organized by row, from left to right, starting from the top of the table. Number of items varies per table. Empty cells are given as empty string values in the corresponding table cell.\n- `table_text`: The sentences forming the description of each table are encoded as a JSON object. In the case of more than one sentence, these are separated by commas. Number of items varies per table.\n- `linearized_input`: A single string that contains the table content separated by vertical bars, i.e., |. Including title, unit of measurement, and the content of each cell including row and column headers in between brackets, i.e., (Medium Empowerment, Mali, 17.9).", "structure-description": "The structure includes all available information for the infographics on which the dataset is based.", "structure-labels": "Annotators looked through English text to identify sentences that describe an infographic. They then identified the corresponding location of the parallel non-English document. All sentences were extracted.", "structure-example": "```\n{\n \"example_id\": \"FR346-en-39\",\n \"title\": \"Trends in early childhood mortality rates\",\n \"unit_of_measure\": \"Deaths per 1,000 live births for the 5-year period before the survey\",\n \"chart_type\": \"Line chart\",\n \"was_translated\": \"False\",\n \"table_data\": \"[[\\\"\\\", \\\"Child mortality\\\", \\\"Neonatal mortality\\\", \\\"Infant mortality\\\", \\\"Under-5 mortality\\\"], [\\\"1990 JPFHS\\\", 5, 21, 34, 39], [\\\"1997 JPFHS\\\", 6, 19, 29, 34], [\\\"2002 JPFHS\\\", 5, 16, 22, 27], [\\\"2007 JPFHS\\\", 2, 14, 19, 21], [\\\"2009 JPFHS\\\", 5, 15, 23, 28], [\\\"2012 JPFHS\\\", 4, 14, 17, 21], [\\\"2017-18 JPFHS\\\", 3, 11, 17, 19]]\",\n \"table_text\": [\n \"neonatal, infant, child, and under-5 mortality rates for the 5 years preceding each of seven JPFHS surveys (1990 to 2017-18).\",\n \"Under-5 mortality declined by half over the period, from 39 to 19 deaths per 1,000 live births.\",\n \"The decline in mortality was much greater between the 1990 and 2007 surveys than in the most recent period.\",\n \"Between 2012 and 2017-18, under-5 mortality decreased only modestly, from 21 to 19 deaths per 1,000 live births, and infant mortality remained stable at 17 deaths per 1,000 births.\"\n ],\n \"linearized_input\": \"Trends in early childhood mortality rates | Deaths per 1,000 live births for the 5-year period before the survey | (Child mortality, 1990 JPFHS, 5) (Neonatal mortality, 1990 JPFHS, 21) (Infant mortality, 1990 JPFHS, 34) (Under-5 mortality, 1990 JPFHS, 39) (Child mortality, 1997 JPFHS, 6) (Neonatal mortality, 1997 JPFHS, 19) (Infant mortality, 1997 JPFHS, 29) (Under-5 mortality, 1997 JPFHS, 34) (Child mortality, 2002 JPFHS, 5) (Neonatal mortality, 2002 JPFHS, 16) (Infant mortality, 2002 JPFHS, 22) (Under-5 mortality, 2002 JPFHS, 27) (Child mortality, 2007 JPFHS, 2) (Neonatal mortality, 2007 JPFHS, 14) (Infant mortality, 2007 JPFHS, 19) (Under-5 mortality, 2007 JPFHS, 21) (Child mortality, 2009 JPFHS, 5) (Neonatal mortality, 2009 JPFHS, 15) (Infant mortality, 2009 JPFHS, 23) (Under-5 mortality, 2009 JPFHS, 28) (Child mortality, 2012 JPFHS, 4) (Neonatal mortality, 2012 JPFHS, 14) (Infant mortality, 2012 JPFHS, 17) (Under-5 mortality, 2012 JPFHS, 21) (Child mortality, 2017-18 JPFHS, 3) (Neonatal mortality, 2017-18 JPFHS, 11) (Infant mortality, 2017-18 JPFHS, 17) (Under-5 mortality, 2017-18 JPFHS, 19)\"\n }\n```", "structure-splits": "- `Train`: Training set, includes examples with 0 or more references.\n- `Validation`: Validation set, includes examples with 3 or more references.\n- `Test`: Test set, includes examples with 3 or more references.\n- `Ru`: Russian zero-shot set. Includes English and Russian examples (Russian is not includes in any of the other splits).", "structure-splits-criteria": "The same table across languages is always in the same split, i.e., if table X is in the test split in language A, it will also be in the test split in language B. In addition to filtering examples without transcribed table values, every example of the development and test splits has at least 3 references. \nFrom the examples that fulfilled these criteria, 100 tables were sampled for both development and test for a total of 800 examples each. A manual review process excluded a few tables in each set, resulting in a training set of 6,962 tables, a development set of 752 tables, and a test set of 763 tables.\n", "structure-outlier": "There are tables without references, without values, and others that are very large. The dataset is distributed as-is, but the paper describes multiple strategies to deal with data issues." } }, "considerations": { "pii": {}, "licenses": { "dataset-restrictions": [ "open license - commercial use allowed" ], "dataset-restrictions-other": "N/A", "data-copyright": [ "open license - commercial use allowed" ], "data-copyright-other": "N/A" }, "limitations": { "data-technical-limitations": "While tables were transcribed in the available languages, the majority of the tables were published in English as the first language. Professional translators were used to translate the data, which makes it plausible that some translationese exists in the data. Moreover, it was unavoidable to collect reference sentences that are only partially entailed by the source tables. ", "data-unsuited-applications": "The domain of health reports includes potentially sensitive topics relating to reproduction, violence, sickness, and death. Perceived negative values could be used to amplify stereotypes about people from the respective regions or countries. The intended academic use of this dataset is to develop and evaluate models that neutrally report the content of these tables but not use the outputs to make value judgments, and these applications are thus discouraged." } }, "gem": { "rationale": { "contribution": "There is no other multilingual data-to-text dataset that is parallel over languages. Moreover, over 70% of references in the dataset require reasoning and it is thus of very high quality and challenging for models.", "sole-task-dataset": "yes", "sole-language-task-dataset": "yes", "distinction-description": "More languages, parallel across languages, grounded in infographics, not centered on Western entities or source documents", "model-ability": "reasoning, verbalization, content planning" }, "curation": { "has-additional-curation": "no", "modification-types": [], "modification-description": "N/A", "has-additional-splits": "no", "additional-splits-description": "N/A", "additional-splits-capacicites": "N/A" }, "starting": { "research-pointers": "The background section of the [paper](https://arxiv.org/abs/2211.00142) provides a list of related datasets.", "technical-terms": "- `data-to-text`: Term that refers to NLP tasks in which the input is structured information and the output is natural language.\n" } }, "results": { "results": { "metrics": [ "Other: Other Metrics" ], "other-metrics-definitions": "`StATA`: A new metric associated with TaTA that is trained on human judgments and which has a much higher correlation with them.", "original-evaluation": "The creators used a human evaluation that measured [attribution](https://arxiv.org/abs/2112.12870) and reasoning capabilities of various models. Based on these ratings, they trained a new metric and showed that existing metrics fail to measure attribution.", "has-previous-results": "no", "current-evaluation": "N/A", "previous-results": "N/A" } }, "curation": { "original": { "rationale": "The curation rationale is to create a multilingual data-to-text dataset that is high-quality and challenging.", "communicative": "The communicative goal is to describe a table in a single sentence.", "is-aggregated": "no", "aggregated-sources": "N/A" }, "language": { "obtained": [ "Found" ], "found": [ "Single website" ], "crowdsourced": [], "created": "N/A", "machine-generated": "N/A", "producers-description": "The language was produced by USAID as part of the Demographic and Health Surveys program (https://dhsprogram.com/).", "topics": "The topics are related to fertility, family planning, maternal and child health, gender, and nutrition.", "validated": "validated by crowdworker", "pre-processed": "N/A", "is-filtered": "not filtered", "filtered-criteria": "N/A" }, "annotations": { "origin": "expert created", "rater-number": "11