|
--- |
|
license: cc-by-nc-4.0 |
|
language: |
|
- en |
|
- fr |
|
size_categories: |
|
- 100K<n<1M |
|
--- |
|
# Refugee Law Lab: Canadian Legal Data |
|
|
|
## Dataset Summary |
|
|
|
The [Refugee Law Lab](https://refugeelab.ca) supports bulk open-access to Canadian legal data to facilitate research and advocacy. |
|
Bulk open-access helps avoid asymmetrical access-to-justice and amplification of marginalization that |
|
results when commercial actors leverage proprietary |
|
legal datasets for profit -- a particular concern in the border control setting. |
|
|
|
The Canadian Legal Data dataset includes the unofficial full text of thousands of court and tribunal |
|
decisions at the federal level. It can be used for legal analytics (i.e. identifying patterns in legal |
|
decision-making), to test ML and NLP tools on a bilingual dataset of Canadian legal materials, and to |
|
pretrain language models for various tasks. |
|
|
|
|
|
## Dataset Structure |
|
|
|
### Data Instances |
|
|
|
#### Court Decisions |
|
|
|
- SCC: Full text of Supreme Court of Canada decisions, based on the Refugee Law Lab's |
|
[Supreme Court of Canada Bulk Decisions Dataset](https://refugeelab.ca/bulk-data/scc/) (1877 – 2023) |
|
- FCA: Full text of Federal Court of Appeal (Canada) decisions that have been given a neutral citation, based on |
|
the Refugee Law Lab's [Federal Court of Appeal Bulk Decisions Dataset](https://refugeelab.ca/bulk-data/fca/) (2001-2023) |
|
- FC: Full text of Federal Court (Canada) decisions that have been given a neutral citation, based on |
|
the Refugee Law Lab's [Federal Court Bulk Decisions Dataset](https://refugeelab.ca/bulk-data/fc/) (2001-2023) |
|
- TCC: Full text of Tax Court of Canada decisions that have been given a neutral citation, based on |
|
the Refugee Law Lab's [Tax Court of Canada Bulk Decisions Dataset](https://refugeelab.ca/bulk-data/tcc/) (2003-2023) |
|
|
|
#### Tribunal Decisions |
|
|
|
- RLLR: Full text of Immigration and Refugee Board, Refugee Protection Division Decisions, as reported in the |
|
[Refugee Law Lab Reporter](https://refugeelab.ca/rllr), based on the Refugee Law Lab's [RLLR Bulk Decisions Dataset](https://refugeelab.ca/bulk-data/rllr/) (2019 – 2022) |
|
|
|
|
|
### Data Fields |
|
|
|
- citation1 (string): Legal citation for the document (neutral citation where available) |
|
|
|
- citation2 (string): For some documents multiple citations are available (e.g. for some periods |
|
the Supreme Court of Canada provided both official reported citation and neutral citation) |
|
|
|
- dataset (string): Name of the data instance (e.g. "SCC", "FCA", "FC", "TCC", etc) |
|
|
|
- year (int32): Year of the document date, which can be useful for filtering |
|
|
|
- name (string): Name of the document, typically the style of cause of a case |
|
|
|
- language (string): Language of the document, "en" for English, "fr" for French, "" for no language specified |
|
|
|
- document_date (string): Date of the document, typically the date of a decision (yyyy-mm-dd) |
|
|
|
- source_url (string): URL where the document was scraped and where the official version can be found |
|
|
|
- scraped_timestamp (string): Date the document was scraped (yyyy-mm-dd) |
|
|
|
- unofficial_text (string): Full text of the document (unofficial version, for official version see source_url) |
|
|
|
- other (string): Field for additional metadata in JSON format, currently a blank string for most datasets |
|
|
|
### Data Languages |
|
|
|
Many documents are available in both English and French. Some are only available in one of the two languages. |
|
|
|
### Data Splits |
|
|
|
The data has not been split, so all files are in the train split. If splitting for training/validation, |
|
some thought should be given to whether it is necessary to limit to one language or to ensure that both |
|
English and French versions of the same documents (where available) are put into the same split. |
|
|
|
### Data Loading |
|
|
|
To load all data instances: |
|
|
|
```python |
|
from datasets import load_dataset |
|
dataset = load_dataset("refugee-law-lab/canadian-legal-data", split="train") |
|
``` |
|
|
|
To load only a specific data instance, for example only the SCC data instance: |
|
|
|
```python |
|
from datasets import load_dataset |
|
dataset = load_dataset("refugee-law-lab/canadian-legal-data", split="train", data_dir="SCC") |
|
``` |
|
|
|
## Dataset Creation |
|
|
|
### Curation Rationale |
|
|
|
The dataset includes all the [Bulk Legal Data](https://refugeelab.ca/bulk-data) made publicly available by |
|
the Refugee Law Lab. The Lab has focused on federal courts (e.g. Supreme Court of Canada, Federal Court of |
|
Appeal, Federal Court) as well as federal administrative tribunals (e.g. Immigration and Refugee Board) because |
|
immigration and refugee law, which is the main area of interest of the Lab, operates mostly at the federal level. |
|
|
|
### Source Data |
|
|
|
#### Initial Data Collection and Normalization |
|
|
|
Details (including links to github repos with code) are available via links on the Refugee Law Lab's |
|
[Bulk Legal Data](https://refugeelab.ca/bulk-data/) page. |
|
|
|
### Personal and Sensitive Information |
|
|
|
Documents may include personal and sensitive information. All documents have been published online or |
|
otherwise released publicly by the relevant court or tribunal. While the open court principle mandates |
|
that court (and some tribunal) materials be made available to the public, there are privacy risks when these |
|
materials become easily and widely available. These privacy risks are particularly acute for marginalized groups, |
|
including refugees and other non-citizens whose personal and sensitive information is included in some of the |
|
documents in this dataset. For example, imagine a repressive government working with private data aggregators to |
|
collect information that is used to target families of political opponents who have sought asylum abroad. |
|
One mechanism used to try to achieve a balance between the open court principle |
|
and privacy is that in publishing the documents in this dataset, the relevant courts and tribunals prohibit |
|
search engines from indexing the documents. Users of this data are required to do the same. |
|
|
|
### Non-Official Versions |
|
|
|
Documents included in this dataset are unofficial copies. For official versions published by |
|
the Government of Canada, please see the source URLs. |
|
|
|
### Non-Affiliation / Endorsement |
|
|
|
The reproduction of documents in this dataset was not done in affiliation with, or with the endorsement of |
|
the Government of Canada. |
|
|
|
## Considerations for Using the Data |
|
|
|
### Social Impact of Dataset |
|
|
|
The Refugee Law Lab recognizes that this dataset -- and further research using the dataset -- raises challenging |
|
questions about how to balance protecting privacy, enhancing government transparency, addressing information |
|
asymmetries, and building technologies that leverage data to advance the rights and interests of |
|
refugees and other displaced people, as well as assisting those working with them (rather than technologies that |
|
[enhance the power of states](https://citizenlab.ca/2018/09/bots-at-the-gate-human-rights-analysis-automated-decision-making-in-canadas-immigration-refugee-system/) |
|
to control the movement of people across borders). |
|
|
|
More broadly, the Refugee Law Lab also recognizes that considerations around privacy and data protection are complex |
|
and evolving. When working on migration, refugee law, data, technology and surveillance, we strive to foreground |
|
intersectional understandings of the systemic harms perpetuated against groups historically made marginalized. We |
|
encourage other users to do the same. |
|
|
|
We also encourage users to try to avoid participating in building technologies that harm refugees and other |
|
marginalized groups, as well as to connect with [community organizations](https://www.migrationtechmonitor.com/ways-to-help) |
|
working in this space, and to [listen directly](https://www.migrationtechmonitor.com/about-us) and learn from people who are affected by new technologies. |
|
|
|
We will review the use these datasets periodically to examine whether continuing to publicly release these datasets achieves |
|
the Refugee Law Lab's goals of advancing the rights and interests of refugees and other marginalized groups without creating |
|
disproportionate risks and harms, including risks related to privacy and human rights. |
|
|
|
|
|
### Discussion of Biases |
|
|
|
The dataset reflects many biases present in legal decision-making, including biases based on race, immigration status, gender, sexual orientation, religion, disability, socio-economic class, and other intersecting categories of discrimination. |
|
|
|
### Other Known Limitations |
|
|
|
Publicly available court and tribunal decisions are not a representative sample of legal decision-making -- and in some cases may reflect |
|
significantly skewed samples. To give one example, the vast majority of Federal Court judicial reviews of refugee determinations involve negative |
|
first instance decisions even thought most first instance decisions are positive (this occurs because the government seldom applies for judicial |
|
reviews of positive first instance decisions whereas claimants frequently apply for judicial review of negative decisions). As such, generative models |
|
built partly on this dataset risk amplifying negative refugee decision-making (rather than more common positive refugee decision-making). Due to the ways that |
|
legal datasets may be skewed, users of this dataset are encouraged to collaborate with or consult domain experts. |
|
|
|
## Additional Information |
|
|
|
|
|
### Licensing Information |
|
|
|
Attribution-NonCommercial 4.0 International ([CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/)) |
|
|
|
NOTE: Users must also comply with upstream licensing for the [SCC](https://www.scc-csc.ca/terms-avis/notice-enonce-eng.aspx), |
|
[FCA](https://www.fca-caf.gc.ca/en/pages/important-notices) & [FC](https://www.fct-cf.gc.ca/en/pages/important-notices) data instances, as |
|
well as requests on source urls not to allow indexing of the documents by search engines to protect privacy. As a result, users must |
|
not make the data available in formats or locations that can be indexed by search engines. |
|
|
|
### Warranties / Representations |
|
|
|
We make no warranties or representations that the data included in this dataset is complete or accurate. Data |
|
were obtained through academic research projects, including projects that use automated processes. |
|
While we try to make the data as accurate as possible, our methodologies may result in |
|
inaccurate or outdated data. As such, data should be viewed as preliminary information aimed to prompt |
|
further research and discussion, rather than as definitive information. |
|
|
|
### Dataset Curators |
|
|
|
[Sean Rehaag](https://www.osgoode.yorku.ca/faculty-and-staff/rehaag-sean), Osgoode Hall Law School Professor & Director of the Refugee Law Lab |
|
|
|
### Citation Information |
|
|
|
Sean Rehaag, "Refugee Law Lab: Canadian Legal Data" (2023) online: Hugging Face: <https://huggingface.co/datasets/refugee-law-lab/canadian-legal-data>. |
|
|
|
### Acknowledgements |
|
|
|
This project draws on research supported by the Social Sciences and Humanities Research Council and the Law Foundation of Ontario. |
|
|
|
The project was inspired in part by the excellent prior work by [pile-of-law](https://huggingface.co/datasets/pile-of-law/pile-of-law) (Peter Henderson et al, "Pile of Law: Learning |
|
Responsible Data Filtering from the Law and a 256GB Open-Source Legal Dataset" (2022), online: arXiv: https://arxiv.org/abs/2207.00220). |
