{"about":{"non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","site":"https://codewithpapers.app","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page","syntology":{"site":"https://syntology.ai","developers":"https://syntology.ai/developers","mcp":{"server":"https://syntology.ai/mcp","transport":"streamable-http","server_card":"https://syntology.ai/.well-known/mcp/server-card.json","auth":{"type":"trial token, no account","trial_token":"https://syntology.ai/api/oauth/trial/token","method":"POST","docs":"https://syntology.ai/developers"}},"have":"https://syntology.ai/api/graph/have?x=<method, arXiv id or title> (free, answers coverage only)","paper_base":"https://syntology.ai/paper/","atlas_base":"https://app.syntology.ai/?focus="},"machine_readable":[{"url":"https://codewithpapers.app/llms.txt","what":"the machine catalog: every machine-readable file, counted"},{"url":"https://codewithpapers.app/index/manifest.json","what":"paper-to-code index by arXiv id, with Syntology's counts"},{"url":"https://codewithpapers.app/search/manifest.json","what":"site search index (titles, authors) and its files"},{"url":"https://codewithpapers.app/download","what":"bulk files: Syntology's layer, described there"},{"url":"https://codewithpapers.app/build_manifest.json","what":"the build record: inputs, counts, exclusions, probes"}]},"url":"/dataset/gerdalir","name":"GerDaLIR","full_name":"A German Dataset for Legal Information Retrieval","description_markdown":"# GerDaLIR\r\n\r\nThe **Ger**man **Da**taset for **L**egal **I**nformation **R**etrieval (GerDaLIR) is a legal information retrieval dataset comprising a large collection of documents, passages and relevance labels. The large amount of training data we provide enables GerDaLIR to be used as a downstream task for German or multilingual language models. The task provided is a precedent retrieval task based on case documents from the open legal information platform [Open Legal Data](http://openlegaldata.io/). Relevance labels are derived from references: If a passage contains a reference to one or more available documents, the passage is used as a query while the referenced cases are labelled as relevant.\r\n \r\nMore information about the dataset and its generation can be taken from the official [GerDaLIR paper](https://aclanthology.org/2021.nllp-1.13/) \r\n\r\n\r\n## Download\r\n\r\nAll files are formatted with tab-separators (*.tsv) and packed with gzip. Some files are packed to tarballs and compressed (\\*.tar.gz). Sizes specified refer to decompressed size. Details about the individual Files can be found in the [dataset details section](#dataset-details)\r\n\r\n\r\n### Essential Files\r\n\r\nThe collection, the queries and the relevance labels can be downloaded with the links in the table below.\r\n\r\n| Filename               | Description                          | Num Records | Size    | Format                  |\r\n|------------------------|-------------------------------------:|------------:|--------:|------------------------:|\r\n| [`collection.tsv.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/collection.tsv.gz)    | Collection - one passage per line    | 3.095.383   | 2.0 GB  | `d_id, passage`         |    \r\n| [`queries.tar.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/queries.tar.gz)       | Queries - train, dev and test        |   122.975   | 127 MB  | `q_id, query`           |   \r\n| [`qrels.tar.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/qrels.tar.gz)         | Labels - train, dev and test         |   144.324   | 1.7 MB  | `q_id, d_id`            | \r\n\r\n\r\n### Optional Files\r\n\r\nBeneath the essential files, also a bunch of optional files can be downloaded. Among them is a file that maps document-ids back to the original file numbers (see [details](#reference-number-map)) . For convenience, we also provide passage-wise and document-wise BM25 rankings (candidates) that can be used for training or re-ranking. Passage-wise candidate files contain the whole passages as text, while document-wise candidate files only specify document-ids.\r\n  \r\n\r\n| Filename               | Description                          | Num Records | Size    | Format                  |\r\n|------------------------|-------------------------------------:|------------:|--------:|------------------------:|\r\n| [`refmap.tsv.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/filemap.tsv.gz)        | Doc-Ids to reference number          |   131.446   | 6.2 MB  | `d_id, slug, file`      | \r\n| [`pass-candidates.train.tsv.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/pass-candidates.train.tsv.gz)        | Top-1000 _passage_ candidates with text (train)   |  98.380.000    | 85 GB  | `q_id, d_id, rank, text`      | \r\n| [`pass-candidates.dev.tsv.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/pass-candidates.dev.tsv.gz)        | Top-1000 _passage_ candidates with text (dev)  |   12.297.000   | 11 GB  | `q_id, d_id, rank, text`      | \r\n| [`pass-candidates.test.tsv.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/pass-candidates.test.tsv.gz)        | Top-1000 _passage_ candidates with text (test)   |  12.298.000 | 11 GB  | `q_id, d_id, rank, text`      | \r\n| [`doc-candidates.train.tsv.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/doc-candidates.train.tsv.gz)        | Top-1000 _document_ candidates (train)   |  98.380.000    | 1.5 GB  | `q_id, d_id, rank`      |\r\n| [`doc-candidates.dev.tsv.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/doc-candidates.dev.tsv.gz)        | Top-1000 _document_ candidates (dev)   |  12.297.000     | 189 MB  | `q_id, d_id, rank`      |\r\n| [`doc-candidates.test.tsv.gz`](https://lavis.cs.hs-rm.de/storage/gerdalir/data/doc-candidates.test.tsv.gz)        | Top-1000 _document_ candidates (test)   |  12.298.000   | 189 MB  | `q_id, d_id, rank`      |\r\n\r\n\r\n\r\n## Dataset details\r\n\r\n### Pre-processing\r\n\r\nAll documents have been pre-processed with the goal to remove all text that is not natural language and remove hints that neural models might exploit. This includes html markup, margin numbers, references, dates and numbers in general. We parse dates as well as references to statutes and other cases using regular expressions, and replace the occurrences with a *[DATE]* or *[REF]* token respectively. Braced contents are removed (including the braces), which in the most cases are comprehensive reference descriptions. All remaining numbers are replaced by zeros.\r\n\r\n### Collection   \r\n\r\nThe collection file contains potentially relevant documents split into passages. Note that passages have no own passage-ids assigned to them. Each line represents one passage and begins with the corresponding document-id (d_id). For document-level retrieval, all passages with the same document-id must be concatenated. \r\n```\r\n1       Das Zulassungsvorbringen der Klägerin begründet keine ernstlichen Zweifel an der Richtigkeit des angefochtenen Urteils . Zweifel in diesem Sinn si..\r\n1       Daran fehlt es hier. Die Antragsbegründung, wonach die Klägerin entgegen den Ausführungen im angefochtenen Urteil zuverlässig sei und die Urteil...\r\n1       In der Antragsbegründung fehlt es an jeglichen Ausführungen zu der ausführlichen Würdigung des Verwaltungsgerichts, die sich im Einzelnen mit de...\r\n2       Tenor Auf die Beschwerden der Antragsteller wird der Beschluss des Verwaltungsgerichts Göttingen 0. Kammer vom [DATE] geändert. Die Antragsgegneri..\r\n2       Durch Beschluss vom [DATE] , auf den wegen der Einzelheiten des Sachverhalts und der Begründung Bezug genommen wird, hat das Verwaltungsgericht den...\r\n```\r\n\r\n### Queries\r\n\r\nQueries are passages that referenced one or more collection documents. There is one query file each for training, development and testing. Query ids (q_id) are assigned globally so that the query files could be concatenated without any problem.\r\n\r\n```\r\n2       Nach [REF] ist eine Erlaubnis zu widerrufen, wenn nachträglich bekannt wird, dass die Voraussetzung nach § 0 Nummer 0 nicht erfüllt ist. Gem...\r\n3       Erforderlich ist mithin eine Prognoseentscheidung unter Berücksichtigung aller Umstände des Einzelfalls dahingehend, ob der Betreffende wille...\r\n4       [REF] ist in Reaktion auf das Urteil des Schleswig-Holsteinischen Landesverfassungsgerichts neu gefasst worden, vor dem Hintergrund, dass sich...\r\n5       Die streitgegenständliche Satzung gibt nicht die Rechtsvorschriften an, welche zum Erlass der Satzung berechtigen, [REF] . Dies ist aber insbe...\r\n6       Insofern gehört zur zutreffenden Angabe der zum Erlass der Satzung berechtigenden Rechtsvorschriften im Sinne des [REF] nicht nur die genaue A...\r\n```\r\n\r\n### Qrels\r\n\r\nThe relevance labels to the queries are also split into sets for training, development and testing. To each query at least one relevance label exist. Multiple relevance labels for a single query result in multiple lines with one target document-id each.\r\n\r\n```\r\n2       118149\r\n3       72511\r\n4       74503\r\n5       4240\r\n5       72939\r\n```\r\n\r\n## How do I cite this work?\r\n\r\nIf you use this dataset for your research, please consider citing our Paper: \r\n\r\n```\r\n@inproceedings{wrzalik-krechel-2021-gerdalir,\r\n    title = \"{G}er{D}a{LIR}: A {G}erman Dataset for Legal Information Retrieval\",\r\n    author = \"Wrzalik, Marco  and\r\n      Krechel, Dirk\",\r\n    booktitle = \"Proceedings of the Natural Legal Language Processing Workshop 2021\",\r\n    month = nov,\r\n    year = \"2021\",\r\n    address = \"Punta Cana, Dominican Republic\",\r\n    publisher = \"Association for Computational Linguistics\",\r\n    url = \"https://aclanthology.org/2021.nllp-1.13\",\r\n    pages = \"123--128\",\r\n    abstract = \"We present GerDaLIR, a German Dataset for Legal Information Retrieval based on case documents from the open legal information platform Open Legal Data. The dataset consists of 123K queries, each labelled with at least one relevant document in a collection of 131K case documents. We conduct several baseline experiments including BM25 and a state-of-the-art neural re-ranker. With our dataset, we aim to provide a standardized benchmark for German LIR and promote open research in this area. Beyond that, our dataset comprises sufficient training data to be used as a downstream task for German or multilingual language models.\",\r\n}\r\n```","description_withheld":null,"homepage":"","introduced_date":null,"introduced_date_note":null,"introduced_by":null,"license":null,"modalities":[],"tasks":[],"languages":[],"variants":["GerDaLIR"],"data_loaders":[],"num_papers_in_archive":1,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[],"papers":{"order":"date (newest first), then slug","population":"every paper with a leaderboard row on this dataset's benchmarks (the benchmark-backed subset): the archive's own papers-using-this-dataset list was never published, so this is not that list; num_papers_in_archive is the archive's own count","page":1,"pages_in_order":1,"rows_per_page":100,"shown":0,"of":0,"counts":{"papers_with_a_benchmark_row":0,"with_a_code_link":0,"where_syntology_ran_a_sample":0,"not_listed_spam_title":0,"listed":0,"listed_where_code_ran":0,"where_syntology_ran_a_sample_split":{"with_a_run_with_no_instrument_failure":0,"every_run_a_failure_of_syntologys_instrument":0,"listed_with_a_run_with_no_instrument_failure":0,"listed_every_run_a_failure_of_syntologys_instrument":0,"filter":{"states":["a run with no instrument failure","any run, instrument failures included"],"default":"a run with no instrument failure","note":"on the 'only where code ran' pages the default hides, in the browser, the rows where every run was a failure of Syntology's instrument; the second state shows them again. Rows are hidden, never re-ordered; these twins list every row"}},"definition":"distinct papers with at least one leaderboard row on this dataset's benchmarks; 'where Syntology ran a sample' counts papers with at least one harvested sample that ran, which is not a correctness claim"},"items":[],"full_list":"papers_with_a_benchmark_row: every listed paper, in this order"},"papers_with_a_benchmark_row":[],"papers_where_code_ran":0,"papers_where_code_ran_first_page":null,"papers_where_code_ran_caption":"We ran code from the paper's repository; we did not run it on this dataset or check it against this dataset's benchmarks.","pages":[],"row_route":"each paper's syntology object carries sample_list (this site's paper page, #ran), syntology_url (syntology.ai's page for the paper, only when that route probed live) and mcp (the agent call that lists the paper's harvested samples; not a verdict on the paper)","developers":"https://syntology.ai/developers","syntology_totals":{"read_at":"2026-10-05T00:36:11+00:00","papers_with_samples":0,"samples_harvested":0,"samples_ran":0,"samples_constructed":0,"samples_ran_checked":0,"samples_ran_instrument_failed":0,"samples_unverified":0,"pointer_only_for_licence":0,"papers_with_no_sample_that_ran":0,"papers_with_a_sample_that_ran":0,"papers_with_a_run_with_no_instrument_failure":0,"papers_where_every_run_was_an_instrument_failure":0,"note":"the per-paper counts of every listed paper (all pages), summed; not a rate"},"key_notes":{"samples_ran_checked":"legacy name, kept unchanged so existing readers do not break: it counts the samples that ran with no instrument failure (honoured, violated, and ran with no contract checked); it does not mean a contract was checked, and the pages print it as 'K with no instrument failure', not 'K checked'","samples_constructed":"a sub-count of the samples that ran, never subtracted from them and never a failure: an executed sample whose run returned an instance of its own class (fixture_out_type equals the entry name): the run built an object and did not compute a result (Syntology's RAN record, counts.constructed)"},"since_archive_judged":{"papers":0,"papers_placed":0,"judged_by":{"model":"a model","models_as_recorded":[],"from":"lb_rows_run_verified.jsonl: each record's `model` field (the extractor's call that wrote decisions[])"},"note":"a model's judgement from the paper's own tables that they report on this leaderboard, not a result: no value from a paper that was not placed is printed or carried here; papers counts each paper once over these leaderboards","backfill":"8,886 of 18,637 newer papers checked so far (733 more had nothing to place; 41 could not be read)","backfill_complete":false,"leaderboards":[]},"papers_note":"The archive never published its papers-using-dataset list; these are papers with a leaderboard row on this dataset's benchmarks.","record_sha256":"e0a1d14c58a138a4c60f57f99175322391ccb14ecdcbb920d279498588225ccc","record_changed_at":"2026-10-05","record_changed_at_basis":"changed"}