|
Download docs/tables.md from mediabiasgroup/MBDef: direct link, hf CLI and curl.
- Browser
- Download file 2.84 kB
-
https://huggingface.co/datasets/mediabiasgroup/MBDef/resolve/main/docs/tables.md
- Command line
-
hf download hf://datasets/mediabiasgroup/MBDef/docs/tables.md
-
curl -L -o tables.md https://huggingface.co/datasets/mediabiasgroup/MBDef/resolve/main/docs/tables.md
2.84 kB
| # What one row is | |
| Part of [MBDef](https://huggingface.co/datasets/mediabiasgroup/MBDef). The dataset has five tables; **start with `terms`**. The other four hold the rows it is built from and share ids. What every column means: [Columns](https://huggingface.co/datasets/mediabiasgroup/MBDef/blob/main/docs/columns.md). | |
| | Table | One row is | Rows | | |
| |---|---|---:| | |
| | `terms` | a concept with all its definitions, description, synonyms, related terms, subtypes, supertypes and examples, each item merged across papers, with how many papers state it | 34,916 | | |
| | `concepts` | a concept: every term-in-paper whose label is the same up to spelling | 34,916 | | |
| | `instances` | a term as one paper defines it ("term-in-paper"), with the relevance scores and everything else extracted for it (description, synonyms, related terms, subtypes, supertypes, examples) | 62,066 | | |
| | `definitions` | one definition text of a term-in-paper, quoted from the paper | 103,476 | | |
| | `papers` | a source paper: DOI, title, year, open-access status, licence | 6,115 | | |
| ## Worked example | |
| The concept “fake news” (`concept_id` gf8d45d620d38) has 6 spellings, among them “fake news”, “Fake news”, “Fake News”, “Fake news (FN)”, “fake-news”, “fake news (FN)”. It appears as a defined term in 659 papers (659 term-in-paper rows) with 1,891 definitions. | |
| One of its rows in `instances` is `m08713153bd204656`: the term “Fake news” in *Fog of War: the role of fake news in initiating and prolonging conflict* (2026, DOI 10.3389/fcomm.2026.1790164, licence cc-by). Its relevance score is 0.967, combined from three scores: embedding classifier 0.988, LLM decision model 0.997 and retrieval judge 1.000 ([how they are combined](https://huggingface.co/datasets/mediabiasgroup/MBDef/blob/main/docs/relevance.md)). | |
| One of its rows in `definitions` (`o04be1d9f462da565`) reads: | |
| > The term itself describes "a news story that has been released through media and has information that is either greatly exaggerated or completely untrue" (Mara, 2019, p. 4). | |
| For it the paper cites “Mara, 2019, p. 4” (`citations` column). | |
| Besides definitions, a row lists what else the paper says about its term. The row `mc22171ef9e31acfe` (“fake news” in *An Overview of the Fake News Phenomenon: From Untruth-Driven to Post-Truth-Driven Approaches*) lists: `description` “Fake news connotes intent to deceive and often to do harm.”, “fake news does not exist independently of real news; instead, it exists as a critique o…” (+9 more); `synonyms` “parody”, “satire” (+6 more); `related_terms` “post-truth”, “information disorder” (+17 more); `subtypes` “untruth-driven definition”, “post-truth-driven definition” (+8 more); `supertypes` “disinformation”; `examples` “Everything against my opinion is fake news.”. | |