Init LaCourFinished

.gitattributes

@@ -53,3 +53,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.jpg filter=lfs diff=lfs merge=lfs -text
 *.jpeg filter=lfs diff=lfs merge=lfs -text
 *.webp filter=lfs diff=lfs merge=lfs -text
+lacourxml.tar.gz filter=lfs diff=lfs merge=lfs -text

LaCourfinished.py

@@ -0,0 +1,223 @@
+import xml.etree.ElementTree as ET
+import datasets
+import pandas as pd
+from huggingface_hub import hf_hub_url
+
+logger = datasets.logging.get_logger(__name__)
+
+class LaCourConfig(datasets.BuilderConfig):
+    def __init__(self, **kwargs):
+        super(LaCourConfig, self).__init__(**kwargs)
+
+
+class LaCourDataset(datasets.GeneratorBasedBuilder):
+    """
+    A class used to represent a Dataset.
+
+    ...
+
+    Attributes
+    ----------
+    VERSION : datasets.Version
+        a version number for the dataset
+
+    BUILDER_CONFIGS : list
+        a list of BuilderConfig instances
+
+    Methods
+    -------
+    _info():
+        Returns the dataset information.
+    _split_generators(download_manager: datasets.DownloadManager):
+        Returns SplitGenerators.
+    _generate_examples():
+        Yields examples.
+    """
+    # Version history
+    # 0.1.0 initial release
+    VERSION = datasets.Version("0.1.0")
+
+    BUILDER_CONFIGS = [
+        datasets.BuilderConfig(
+            name="transcripts",
+            version=VERSION,
+            description="transcript dataset based on xml files",
+        ),
+        datasets.BuilderConfig(
+            name="documents",
+            version=VERSION,
+            description="linked documents associated with the webcast"
+        )
+
+    ]
+
+    DEFAULT_CONFIG_NAME = "transcripts"
+
+    def _info(self):
+        """
+        Returns the dataset information.
+
+        ...
+
+        Returns
+        -------
+        datasets.DatasetInfo
+            a DatasetInfo instance containing information about the dataset
+        """
+        if self.config.name == "transcripts":
+            return datasets.DatasetInfo(
+                features=datasets.Features(
+                    {
+                        "id": datasets.Value("int32"),
+                        "webcast_id": datasets.Value("string"),
+                        "segment_id": datasets.features.Value("int32"),
+                        "speaker_name": datasets.features.Value("string"),
+                        "speaker_role": datasets.features.Value("string"),
+                        "data": datasets.features.Sequence({
+                            "begin": datasets.features.Value("float32"),
+                            "end": datasets.features.Value("float32"),
+                            "language": datasets.features.Value("string"),
+                            "text": datasets.features.Value("string"),
+                        })
+                    }
+                ),
+                supervised_keys=None,
+            )
+        else:
+            return datasets.DatasetInfo(
+                features=datasets.Features(
+                    {
+                        "id": datasets.Value("int32"),
+                        "webcast_id": datasets.Value("string"),
+                        "hearing_title": datasets.Value("string"),
+                        "hearing_date": datasets.Value("string"),
+                        "hearing_type": datasets.Value("string"),
+                        "application_number": datasets.features.Sequence(datasets.Value("string")),
+                        "case_id": datasets.Value("string"),
+                        "case_name": datasets.Value("string"),
+                        "case_url": datasets.Value("string"),
+                        "ecli": datasets.Value("string"),
+                        "type": datasets.Value("string"),
+                        "document_date": datasets.Value("string"),
+                        "importance": datasets.Value("int32"),
+                        "articles": datasets.features.Sequence(datasets.Value("string")),
+                        "respondent_government": datasets.features.Sequence(datasets.Value("string")),
+                        "issue": datasets.Value("string"),
+                        "strasbourg_caselaw": datasets.Value("string"),
+                        "external_sources": datasets.Value("string"),
+                        "conclusion": datasets.Value("string"),
+                        "separate_opinion": datasets.Value("bool")
+                    }
+                ),
+                supervised_keys=None,
+            )
+
+
+    def _split_generators(self, dl_manager):
+        """
+        Returns SplitGenerators.
+
+        Parameters
+        ----------
+        download_manager : datasets.DownloadManager
+            a DownloadManager instance
+
+        Returns
+        -------
+        list
+            a list of SplitGenerator instances
+        """
+        base_url_xml = hf_hub_url("TrustHLT-ECALP/LaCourfinished]", filename="lacourxml.tar.gz", repo_type="dataset")
+        base_url_json = hf_hub_url("TrustHLT-ECALP/LaCourfinished]", filename="lacour_linked_documents.json", repo_type="dataset")
+
+        if self.config.name == "transcripts":
+            path = dl_manager.download(base_url_xml)
+            xmlpath = dl_manager.iter_archive(path)
+            return [
+                datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepaths": xmlpath}),
+                ]
+        else:
+            jsonpath = dl_manager.download(base_url_json)
+            return [
+                datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepaths": jsonpath}),
+                ]
+
+    def _generate_examples(self, filepaths):
+        """
+        This method reads the files in the provided transcripts, parses the data, and yields it in a structured format.
+
+        For the configuration "xml", it reads XML files and extracts speaker segments and associated metadata.
+
+        Parameters
+        ----------
+        filepaths : list
+            A list of filepaths to the data files.
+
+        Yields
+        ------
+        tuple
+            A tuple containing an ID and a dictionary with the data. The dictionary keys include 'id' and 'data'.
+            'data' is a list of lists, where each inner list contains a key and a value extracted from the data file.
+        """
+        if self.config.name == "transcripts":
+            id_ = 0
+            for fpath, file in filepaths:
+                logger.info("generating examples from = %s", fpath)
+                tree = ET.parse(file)
+                root = tree.getroot()
+                segment_id = 0
+                for speakerSegment in root.findall('SpeakerSegment'):
+                    text_segments = []
+                    for segment in speakerSegment.findall('Segment'):
+                        meta_data = segment.find('meta_data')
+                        text_segments.append({                                
+                            "begin": meta_data.findtext('TimestampBegin', ''),
+                            "end": meta_data.findtext('TimestampEnd', ''),
+                            "language": meta_data.findtext('Language', ''),
+                            "text": segment.findtext('text', '').strip(),
+                        })
+                    feature = id_, {
+                        "id": id_,
+                        "webcast_id": fpath.split('_')[1] + "_" + fpath.split('_')[2].split('.')[0],
+                        "segment_id": segment_id,
+                        "speaker_role": meta_data.findtext('Role', ''),
+                        "speaker_name": meta_data.findtext('Name', ''),
+                        "data": text_segments
+                    }
+                    yield feature
+                    id_ += 1
+                    segment_id += 1
+                    
+        elif self.config.name == "documents":
+            id_ = 0
+            df = pd.read_json(filepaths, orient="index", dtype={"webcast_id": str})
+            logger.info("generating examples from = %s", filepaths)
+            cols = df.columns.tolist()
+            cols.remove('appno')
+            # group appnos to avoid duplicates
+            df = df.groupby(cols)['appno'].apply(';'.join).reset_index()
+            for _, row in df.iterrows():
+                feature = id_,{
+                    "id": id_,
+                    "webcast_id": row["webcast_id"],
+                    "hearing_title": row["hearing_title"], 
+                    "hearing_date": row["hearing_date"],
+                    "hearing_type": row["hearing_type"],
+                    "application_number": row["appno"].split(';'),
+                    "case_id": row["case_id"],
+                    "case_name": row["case_name"],
+                    "case_url": row["case_url"],
+                    "ecli": row["ecli"],
+                    "type": row["type"],
+                    "document_date": row["document_date"],
+                    "importance": row["importance"],
+                    "articles": row["articles"].split(';'),
+                    "respondent_government": row["respondent"].split(';'),
+                    "issue": row["issue"],
+                    "strasbourg_caselaw": row["strasbourg_caselaw"],
+                    "external_sources": row["external_sources"],
+                    "conclusion": row["conclusion"],
+                    "separate_opinion": row["separate_opinion"]
+                }
+                yield feature
+                id_ += 1

README.md

@@ -0,0 +1,208 @@
+---
+language:
+- en
+- fr
+- ru
+- es
+- hr
+- it
+- pt
+- tr
+- pl
+- lt
+- de
+- uk
+- hu
+- nl
+- sq
+- ro
+- sr
+pretty_name: LaCour!
+language_creators:
+- found
+- machine-generated
+license:
+- cc-by-sa-4.0
+multilinguality:
+- multilingual
+size_categories:
+- 1K<n<10K
+- n<1K
+tags:
+- legal
+- hearing
+- oral argument
+- transcript
+- echr
+- dialog
+---
+
+## Dataset Description
+
+- **Homepage: https://trusthlt.org/lacour**
+- **Repository: https://github.com/trusthlt/lacour-corpus**
+- **Paper: tbd**
+
+### Dataset Summary
+
+This dataset contains transcribed court hearings sourced from official hearings of the __European Court of Human Rights__ ([https://www.echr.coe.int/webcasts-of-hearings](https://www.echr.coe.int/webcasts-of-hearings)). The hearings are 154 selected webcasts (videos) from 2012-2022 in their original language (no interpretation). With manual annotation for language labels and automatic processing of the extracted audio with [pyannote](https://huggingface.co/pyannote/speaker-diarization) and [whisper-large-v2](https://huggingface.co/openai/whisper-large-v2), the resulting dataset contains 4000 speaker turns and 88920 individual lines. The dataset contains two subsets, the transcripts and the metadata with linked documents. The transcripts are additionally available as .txt or .xml.
+
+### Languages
+
+The largest amounts in the transcripts are:
+
+English, French
+
+A smaller portion also contains the following languages:
+
+Russian, Spanish, Croatian, Italian, Portuguese, Turkish, Polish, Lithuanian, German, Ukrainian, Hungarian, Dutch, Albanian, Romanian, Serbian
+
+The collected metadata is:
+
+English
+
+## Dataset Structure
+
+### Data Instances
+
+Each instance in transcripts represents an entire segment of a transcript, similar to a conversation turn in a dialog.  
+
+```
+{
+ 'id': 0,
+ 'webcast_id': '1021112_29112017',
+ 'segment_id': 0,
+ 'speaker_name': 'UNK',
+ 'speaker_role': 'Announcer',
+ 'data': {
+   'begin': [12.479999542236328],
+   'end': [13.359999656677246],
+   'language': ['fr'],
+   'text': ['La Cour!']
+  }
+}
+```
+
+Each instance in documents represents a information on a document in hudoc associated with a hearing and the metadata associated with a hearing. The actual document is linked and can also be found in [hudoc](https://hudoc.echr.coe.int) with the case_id. Note: `hearing_type` states the type of the hearing, `type` states the type of the document. If the hearing is a "Grand Chamber hearing", the "CHAMBER" document refers to a different hearing.
+
+```
+{
+  'id': 16,
+  'webcast_id': '1232311_02102012',
+  'hearing_title': 'Michaud v. France (nos. 12323/11)',
+  'hearing_date': '2012-10-02 00:00:00',
+  'hearing_type': 'Chamber hearing',
+  'application_number': ['12323/11'],
+  'case_id': '001-115377',
+  'case_name': 'CASE OF MICHAUD v. FRANCE',
+  'case_url': 'https://hudoc.echr.coe.int/eng?i=001-115377',
+  'ecli': 'ECLI:CE:ECHR:2012:1206JUD001232311',
+  'type': 'CHAMBER',
+  'document_date': '2012-12-06 00:00:00',
+  'importance': 1,
+  'articles': ['8', '8-1', '8-2', '34', '35'],
+  'respondent_government': ['FRA'],
+  'issue': 'Decision of the National Bar Council of 12 July 2007 “adopting regulations on internal procedures for implementing the obligation to combat money laundering and terrorist financing, and an internal supervisory mechanism to guarantee compliance with those procedures” ; Article 21-1 of the Law of 31 December 1971 ; Law no. 2004-130 of 11 February 2004 ; Monetary and Financial Code',
+  'strasbourg_caselaw': 'André and Other v. France, no 18603/03, 24 July 2008;Bosphorus Hava Yollari Turizm ve Ticaret Anonim Sirketi v. Ireland [GC], no 45036/98, ECHR 2005-VI;[...]',
+  'external_sources': 'Directive 91/308/EEC, 10 June 1991;Article 6 of the Treaty on European Union;Charter of Fundamental Rights of the European Union;Articles 169, 170, 173, 175, 177, 184 and 189 of the Treaty establishing the European Community;Recommendations 12 and 16 of the financial action task  force (“FATF”) on money laundering;Council of Europe Convention on Laundering, Search, Seizure and Confiscation of the Proceeds from Crime and on the Financing of Terrorism  (16 May 2005)',
+  'conclusion': 'Remainder inadmissible;No violation of Article 8 - Right to respect for private and family life (Article 8-1 - Respect for correspondence;Respect for private life)',
+  'separate_opinion': True
+  }
+
+
+```
+
+### Data Fields
+
+transcripts:
+
+* id: the identifier
+* webcast_id: the identifier for the hearing
+* segment_id: the identifier of the current speaker segment in the current hearing
+* speaker_name: the name of the speaker (not given for Applicant, Government or Third Party)
+* speaker_role: the role/party the speaker represents (`Announcer` for announcements, `Judge` for judges, `JudgeP` for judge president, `Applicant` for representatives of the applicant, `Government` for representatives of the respondent government, `ThirdParty` for representatives of third party interveners)
+* data: sequence of the following fields
+  * begin: the timestamp for begin of line (in seconds)
+  * end: the timestamp for end of line (in seconds)
+  * language: the language spoken (in ISO 639-1)
+  * text: the spoken line
+
+documents:
+
+* id: the identifier
+* webcast_id: the identifier for the hearing (allows linking to transcripts)
+* hearing_title: the title of the hearing
+* hearing_date: the date of the hearing
+* hearing_type: the type of hearing (Grand Chamber, Chamber or Grand Chamber Judgment Hearing)
+* application_number: the application numbers which are associated with the hearing and case
+* case_id: the id of the case
+* case_name: the name of the case 
+* case_url: the direct link to the document
+* ecli: the ECLI (European Case Law Identifier)
+* type: the type of the document
+* document_date: the date of the document
+* importance: the importance score of the case (1 is the highest importance, key case)
+* articles: the concerning articles of the Convention of Human Rights
+* respondent_government: the code of the respondent government(s) (in ISO-3166 Alpha-3)
+* issue: the references to the issue of the case
+* strasbourg_caselaw: the list of cases in the ECHR which are relevant to the current case
+* external_sources: the relevant references outside of the ECHR
+* conclusion: the short textual description of the conclusion
+* separate_opinion: the indicator if there is a separate opinion
+
+### Data Splits
+
+The dataset is only split into a train set.
+
+
+## Dataset Creation
+
+### Curation Rationale
+
+This datasets provides partly corrected transcribed webcasts to enable the processing of hearings in legal NLP. No specific task is given.
+
+### Source Data
+
+#### Data Collection
+
+The data was collected by transcribing the publicly available [webcasts of the ECHR](https://www.echr.coe.int/webcasts-of-hearings) with the help of [pyannote](https://huggingface.co/pyannote/speaker-diarization) and [whisper-large-v2](https://huggingface.co/openai/whisper-large-v2). The documents were sourced from the [ECHR hudoc database](https://hudoc.echr.coe.int).
+
+#### Who are the source producers?
+
+Participants in hearings before the ECHR for the audio and video material. Employees and judges of the ECHR for the documents.
+
+### Annotations
+
+#### Annotation process
+
+**language identification** Spoken languages were manually identified by research assistants. Disagreements were discussed to achieve the final language label.
+
+**transcript correction** All parts spoken by Judge or Judge President are corrected for the languages English and French by research assistants with a high proficiency in the respective language.
+
+#### Personal and Sensitive Information
+
+The dataset contains names of judges and other participants in the hearings. Due to those names being available in the public court material, we did not remove them. The machine-generated transcripts may also contain names, which were neither checked nor removed. In case of sensitive information, we rely on the provided material to provide protection (occasionally bleeping out names which should not have been mentioned in webcasts, appropriate anonymization in the documents).
+
+
+## Additional Information
+
+Download the transcripts and linked documents:
+
+```
+from datasets import load_dataset
+
+lacour_transcripts = load_dataset("TrustHLT/LaCour", "transcripts") # default config
+lacour_documents = load_dataset("TrustHLT/LaCour", "documents")
+
+```
+
+Formatted versions of the transcripts in .txt and .xml and more information on the collection and creation can be found on [github](https://github.com/trusthlt/lacour-corpus).
+
+
+### Citation Information
+
+Please cite this data using:
+
+```
+[tbd]
+```

lacourxml.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d4b37dadbb952c48ff4bcc023ec585d7fb7d057e3d63f688a9b596f347a5a64
+size 5475134