Update files from the datasets library (from 1.2.0)

Release notes: https://github.com/huggingface/datasets/releases/tag/1.2.0

.gitattributes

@@ -0,0 +1,27 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bin.* filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,189 @@
+---
+annotations_creators:
+- expert-generated
+language_creators:
+- found
+languages:
+- yue
+licenses:
+- cc-by-4-0
+multilinguality:
+- monolingual
+size_categories:
+- 10K<n<100K
+source_datasets:
+- original
+task_categories:
+- conditional-text-generation
+- sequence-modeling
+task_ids:
+- dialogue-modeling
+- machine-translation
+---
+
+# Dataset Card for The Hong Kong Cantonese Corpus (HKCanCor)
+
+## Table of Contents
+- [Dataset Description](#dataset-description)
+  - [Dataset Summary](#dataset-summary)
+  - [Supported Tasks](#supported-tasks-and-leaderboards)
+  - [Languages](#languages)
+- [Dataset Structure](#dataset-structure)
+  - [Data Instances](#data-instances)
+  - [Data Fields](#data-instances)
+  - [Data Splits](#data-instances)
+- [Dataset Creation](#dataset-creation)
+  - [Curation Rationale](#curation-rationale)
+  - [Source Data](#source-data)
+  - [Annotations](#annotations)
+  - [Personal and Sensitive Information](#personal-and-sensitive-information)
+- [Considerations for Using the Data](#considerations-for-using-the-data)
+  - [Social Impact of Dataset](#social-impact-of-dataset)
+  - [Discussion of Biases](#discussion-of-biases)
+  - [Other Known Limitations](#other-known-limitations)
+- [Additional Information](#additional-information)
+  - [Dataset Curators](#dataset-curators)
+  - [Licensing Information](#licensing-information)
+  - [Citation Information](#citation-information)
+
+## Dataset Description
+
+- **Homepage:** http://compling.hss.ntu.edu.sg/hkcancor/
+- **Repository:** https://github.com/fcbond/hkcancor
+- **Paper:** [Luke and Wang, 2015](https://github.com/fcbond/hkcancor/blob/master/data/LukeWong_Hong-Kong-Cantonese-Corpus.pdf)
+- **Leaderboard:** N/A
+- **Point of Contact:** Luke Kang Kwong
+
+### Dataset Summary
+The Hong Kong Cantonese Corpus (HKCanCor) comprise transcribed conversations recorded 
+between March 1997 and August 1998. It contains recordings of spontaneous speech (51 texts)
+and radio programmes (42 texts), which involve 2 to 4 speakers, with 1 text of monologue.
+
+In total, the corpus contains around 230,000 Chinese words. The text is word-segmented (i.e., tokenization is at word-level, and each token can span multiple Chinese characters). Tokens are annotated with part-of-speech (POS) tags and romanised Cantonese pronunciation. 
+
+* Romanisation
+  * Follows conventions set by the Linguistic Society of Hong Kong (LSHK).
+* POS
+  * The tagset used by this corpus extends the one in the Peita-Fujitsu-Renmin Ribao (PRF) corpus (Duan et al., 2000). Extensions were made to further capture Cantonese-specific phenomena. 
+  * To facilitate everyday usage and for better comparability across languages and/or corpora, this dataset also includes the tags mapped to the [Universal Dependencies 2.0](https://universaldependencies.org/u/pos/index.html) format. This mapping references the [PyCantonese](https://github.com/jacksonllee/pycantonese) library.
+
+
+### Supported Tasks and Leaderboards
+[More Information Needed]
+
+### Languages
+Yue Chinese / Cantonese (Hong Kong).
+
+## Dataset Structure
+This corpus has 10801 utterances and approximately 230000 Chinese words. 
+There is no predefined split. 
+
+### Data Instances
+Each instance contains a conversation id, speaker id within that conversation,
+turn number, part-of-speech tag for each Chinese word in the PRF format and UD2.0 format, 
+and the utterance written in Chinese characters as well as its LSHK format romanisation.
+
+
+For example: 
+```python
+{
+    'conversation_id': 'TNR016-DR070398-HAI6V'
+    'pos_tags_prf': ['v', 'w'], 
+    'pos_tags_ud': ['VERB', 'PUNCT'],
+    'speaker': 'B', 
+    'transcriptions': ['hai6', 'VQ1'], 
+    'turn_number': 112, 
+    'tokens': ['係', '。']
+}
+ ```
+
+### Data Fields
+- conversation_id: unique dialogue-level id
+- pos_tags_prf: POS tag using the PRF format at token-level
+- pos_tag_ud: POS tag using the UD2.0 format at token-level
+- speaker: unique speaker id within dialogue
+- transcriptions: token-level romanisation in the LSHK format
+- turn_number: turn number in dialogue
+- tokens: Chinese word or punctuation at token-level
+
+### Data Splits
+There are no specified splits in this dataset.
+
+## Dataset Creation
+
+### Curation Rationale
+
+[More Information Needed]
+
+### Source Data
+
+#### Initial Data Collection and Normalization
+
+[More Information Needed]
+
+#### Who are the source language producers?
+
+[More Information Needed]
+
+### Annotations
+
+#### Annotation process
+
+[More Information Needed]
+
+#### Who are the annotators?
+
+[More Information Needed]
+
+### Personal and Sensitive Information
+
+[More Information Needed]
+
+## Considerations for Using the Data
+
+### Social Impact of Dataset
+
+[More Information Needed]
+
+### Discussion of Biases
+
+[More Information Needed]
+
+### Other Known Limitations
+
+[More Information Needed]
+
+## Additional Information
+
+### Dataset Curators
+
+[More Information Needed]
+
+### Licensing Information
+This work is licensed under a [Creative Commons Attribution 4.0 International License](https://creativecommons.org/licenses/by/4.0/deed.ast).
+
+
+### Citation Information
+This corpus was developed by [Luke and Wong, 2015](http://compling.hss.ntu.edu.sg/hkcancor/data/LukeWong_Hong-Kong-Cantonese-Corpus.pdf).
+```
+@article{luke2015hong,
+  author={Luke, Kang-Kwong and Wong, May LY},
+  title={The Hong Kong Cantonese corpus: design and uses},
+  journal={Journal of Chinese Linguistics},
+  year={2015},
+  pages={309-330},
+  month={12}
+}
+```
+The POS tagset to Universal Dependency tagset mapping is provided by Jackson Lee, as a part of the [PyCantonese](https://github.com/jacksonllee/pycantonese) library. 
+```
+@misc{lee2020,
+  author = {Lee, Jackson},
+  title = {PyCantonese: Cantonese Linguistics and NLP in Python},
+  year = {2020},
+  publisher = {GitHub},
+  journal = {GitHub repository},
+  howpublished = {\url{https://github.com/jacksonllee/pycantonese}},
+  commit = {1d58f44e1cb097faa69de6b617e1d28903b84b98}
+}
+```

dataset_infos.json

@@ -0,0 +1 @@
+{"default": {"description": "The Hong Kong Cantonese Corpus (HKCanCor) comprise transcribed conversations\nrecorded between March 1997 and August 1998. It contains recordings of\nspontaneous speech (51 texts) and radio programmes (42 texts),\nwhich involve 2 to 4 speakers, with 1 text of monologue.\n\nIn total, the corpus contains around 230,000 Chinese words.\nThe text is word-segmented, annotated with part-of-speech (POS) tags and\nromanised Cantonese pronunciation.\n\nRomanisation scheme - Linguistic Society of Hong Kong (LSHK)\nPOS scheme - Peita-Fujitsu-Renmin Ribao (PRF) corpus (Duan et al., 2000),\n             with extended tags for Cantonese-specific phenomena added by\n             Luke and Wang (see original paper for details).\n", "citation": "@article{luke2015hong,\n  author={Luke, Kang-Kwong and Wong, May LY},\n  title={The Hong Kong Cantonese corpus: design and uses},\n  journal={Journal of Chinese Linguistics},\n  year={2015},\n  pages={309-330},\n  month={12}\n}\n@misc{lee2020,\n  author = {Lee, Jackson},\n  title = {PyCantonese: Cantonese Linguistics and NLP in Python},\n  year = {2020},\n  publisher = {GitHub},\n  journal = {GitHub repository},\n  howpublished = {https://github.com/jacksonllee/pycantonese},\n  commit = {1d58f44e1cb097faa69de6b617e1d28903b84b98}\n}\n", "homepage": "http://compling.hss.ntu.edu.sg/hkcancor/", "license": "CC BY 4.0", "features": {"conversation_id": {"dtype": "string", "id": null, "_type": "Value"}, "speaker": {"dtype": "string", "id": null, "_type": "Value"}, "turn_number": {"dtype": "int16", "id": null, "_type": "Value"}, "tokens": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "transcriptions": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "pos_tags_prf": {"feature": {"num_classes": 120, "names": ["!", "\"", "#", "'", ",", "-", ".", "...", "?", "A", "AD", "AG", "AIRWAYS0", "AN", "AND", "B", "BG", "BEAN0", "C", "CENTRE0", "CG", "D", "D1", "DG", "E", "ECHO0", "F", "G", "G1", "G2", "H", "HILL0", "I", "IG", "J", "JB", "JM", "JN", "JNS", "JNT", "JNZ", "K", "KONG", "L", "L1", "LG", "M", "MG", "MONTY0", "MOUNTAIN0", "N", "N1", "NG", "NR", "NS", "NSG", "NT", "NX", "NZ", "O", "P", "PEPPER0", "Q", "QG", "R", "RG", "S", "SOUND0", "T", "TELECOM0", "TG", "TOUCH0", "U", "UG", "U0", "V", "V1", "VD", "VG", "VK", "VN", "VU", "VUG", "W", "X", "XA", "XB", "XC", "XD", "XE", "XJ", "XJB", "XJN", "XJNT", "XJNZ", "XJV", "XJA", "XL1", "XM", "XN", "XNG", "XNR", "XNS", "XNT", "XNX", "XNZ", "XO", "XP", "XQ", "XR", "XS", "XT", "XV", "XVG", "XVN", "XX", "Y", "YG", "Y1", "Z"], "names_file": null, "id": null, "_type": "ClassLabel"}, "length": -1, "id": null, "_type": "Sequence"}, "pos_tags_ud": {"feature": {"num_classes": 16, "names": ["DET", "PRON", "VERB", "NOUN", "ADJ", "PUNCT", "INTJ", "ADV", "V", "PART", "X", "NUM", "PROPN", "AUX", "CCONJ", "ADP"], "names_file": null, "id": null, "_type": "ClassLabel"}, "length": -1, "id": null, "_type": "Sequence"}}, "post_processed": null, "supervised_keys": null, "builder_name": "hkcancor", "config_name": "default", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 5746381, "num_examples": 10801, "dataset_name": "hkcancor"}}, "download_checksums": {"http://compling.hss.ntu.edu.sg/hkcancor/data/hkcancor-utf8.zip": {"num_bytes": 961514, "checksum": "09223963b8756254e15353cad843f8a4b0cbc4b9223dc8a8fa27fb1cf846057e"}}, "download_size": 961514, "post_processing_size": null, "dataset_size": 5746381, "size_in_bytes": 6707895}}

dummy/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac3b79bc1b7bb337ff108efb54a2d0c11cd9163eeed64680e511961e6d2df262
+size 36493

hkcancor.py

@@ -0,0 +1,315 @@
+# coding=utf-8
+# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Hong Kong Cantonese Corpus (HKCanCor)."""
+
+from __future__ import absolute_import, division, print_function
+
+import os
+import xml.etree.ElementTree as ET
+
+import datasets
+
+
+_CITATION = """\
+@article{luke2015hong,
+  author={Luke, Kang-Kwong and Wong, May LY},
+  title={The Hong Kong Cantonese corpus: design and uses},
+  journal={Journal of Chinese Linguistics},
+  year={2015},
+  pages={309-330},
+  month={12}
+}
+@misc{lee2020,
+  author = {Lee, Jackson},
+  title = {PyCantonese: Cantonese Linguistics and NLP in Python},
+  year = {2020},
+  publisher = {GitHub},
+  journal = {GitHub repository},
+  howpublished = {https://github.com/jacksonllee/pycantonese},
+  commit = {1d58f44e1cb097faa69de6b617e1d28903b84b98}
+}
+"""
+
+_DESCRIPTION = """\
+The Hong Kong Cantonese Corpus (HKCanCor) comprise transcribed conversations
+recorded between March 1997 and August 1998. It contains recordings of
+spontaneous speech (51 texts) and radio programmes (42 texts),
+which involve 2 to 4 speakers, with 1 text of monologue.
+
+In total, the corpus contains around 230,000 Chinese words.
+The text is word-segmented, annotated with part-of-speech (POS) tags and
+romanised Cantonese pronunciation.
+
+Romanisation scheme - Linguistic Society of Hong Kong (LSHK)
+POS scheme - Peita-Fujitsu-Renmin Ribao (PRF) corpus (Duan et al., 2000),
+             with extended tags for Cantonese-specific phenomena added by
+             Luke and Wang (see original paper for details).
+"""
+
+_HOMEPAGE = "http://compling.hss.ntu.edu.sg/hkcancor/"
+
+_LICENSE = "CC BY 4.0"
+
+_URL = "http://compling.hss.ntu.edu.sg/hkcancor/data/hkcancor-utf8.zip"
+
+
+class Hkcancor(datasets.GeneratorBasedBuilder):
+    """Hong Kong Cantonese Corpus (HKCanCor)."""
+
+    VERSION = datasets.Version("1.0.0")
+
+    # Original tagset has 110 + tags and includes fine-grained annotations,
+    # e.g., distinguish morphemes vs non-moprhemes. For practical purposes
+    # (usability, comparing across datasets), Lee 2020 mapped HKCanCor tags
+    # to the Universal Dependencies 2.0 scheme. The following is adapted from:
+    # https://github.com/jacksonllee/pycantonese/blob/master/pycantonese/pos_tagging/hkcancor_to_ud.py
+
+    pos_map = {
+        "!": "PUNCT",
+        '"': "PUNCT",
+        "#": "X",
+        "'": "PUNCT",
+        ",": "PUNCT",
+        "-": "PUNCT",
+        ".": "PUNCT",
+        "...": "PUNCT",
+        "?": "PUNCT",
+        "A": "ADJ",  # HKCanCor: Adjective
+        "AD": "ADV",  # HKCanCor: Adjective as Adverbial
+        "AG": "ADJ",  # HKCanCor: Adjective Morpheme
+        "AIRWAYS0": "PROPN",
+        "AN": "NOUN",  # HKCanCor: Adjective with Nominal Function
+        "AND": "PROPN",  # In one instance of "Chilli and Pepper"
+        "B": "ADJ",  # HKCanCor: Non-predicate Adjective
+        "BG": "ADJ",  # HKCanCor: Non-predicate Adjective Morpheme
+        "BEAN0": "PROPN",  # In one instance of "Mr Bean"
+        "C": "CCONJ",  # HKCanCor: Conjunction
+        "CENTRE0": "NOUN",  # In one instance of "career centre"
+        "CG": "CCONJ",
+        "D": "ADV",  # HKCanCor: Adverb
+        "D1": "ADV",  # Most instances are gwai2 "ghost".
+        "DG": "ADV",  # HKCanCor: Adverb Morpheme
+        "E": "INTJ",  # HKCanCor: Interjection
+        "ECHO0": "PROPN",  # In one instance of "Big Echo"
+        "F": "ADV",  # HKCanCor: Directional Locality
+        "G": "X",  # HKCanCor: Morpheme
+        "G1": "V",  # The first A in the "A-not-AB" pattern, where AB is a verb.
+        "G2": "ADJ",  # The first A in "A-not-AB", where AB is an adjective.
+        "H": "PROPN",  # HKCanCor: Prefix (aa3 阿 followed by a person name)
+        "HILL0": "PROPN",  # In "Benny Hill"
+        "I": "X",  # HKCanCor: Idiom
+        "IG": "X",
+        "J": "NOUN",  # HKCanCor: Abbreviation
+        "JB": "ADJ",
+        "JM": "NOUN",
+        "JN": "NOUN",
+        "JNS": "PROPN",
+        "JNT": "PROPN",
+        "JNZ": "PROPN",
+        "K": "X",  # HKCanCor: Suffix (sing3 性 for nouns; dei6 地 for adverbs)
+        "KONG": "PROPN",  # In "Hong Kong"
+        "L": "X",  # Fixed Expression
+        "L1": "X",
+        "LG": "X",
+        "M": "NUM",  # HKCanCor: Numeral
+        "MG": "X",
+        "MONTY0": "PROPN",  # In "Full Monty"
+        "MOUNTAIN0": "PROPN",  # In "Blue Mountain"
+        "N": "NOUN",  # Common Noun
+        "N1": "DET",  # HKCanCor: only used for ne1 呢; determiner
+        "NG": "NOUN",
+        "NR": "PROPN",  # HKCanCor: Personal Name
+        "NS": "PROPN",  # HKCanCor: Place Name
+        "NSG": "PROPN",
+        "NT": "PROPN",  # HKCanCor: Organization Name
+        "NX": "NOUN",  # HKCanCor: Nominal Character String
+        "NZ": "PROPN",  # HKCanCor: Other Proper Noun
+        "O": "X",  # HKCanCor: Onomatopoeia
+        "P": "ADP",  # HKCanCor: Preposition
+        "PEPPER0": "PROPN",  # In "Chilli and Pepper"
+        "Q": "NOUN",  # HKCanCor: Classifier
+        "QG": "NOUN",  # HKCanCor: Classifier Morpheme
+        "R": "PRON",  # HKCanCor: Pronoun
+        "RG": "PRON",  # HKCanCor: Pronoun Morpheme
+        "S": "NOUN",  # HKCanCor: Space Word
+        "SOUND0": "PROPN",  # In "Manchester's Sound"
+        "T": "ADV",  # HKCanCor: Time Word
+        "TELECOM0": "PROPN",  # In "Hong Kong Telecom"
+        "TG": "ADV",  # HKCanCor: Time Word Morpheme
+        "TOUCH0": "PROPN",  # In "Don't Touch" (a magazine)
+        "U": "PART",  # HKCanCor: Auxiliary (e.g., ge3 嘅 after an attributive adj)
+        "UG": "PART",  # HKCanCor: Auxiliary Morpheme
+        "U0": "PROPN",  # U as in "Hong Kong U" (= The University of Hong Kong)
+        "V": "VERB",  # HKCanCor: Verb
+        "V1": "VERB",
+        "VD": "ADV",  # HKCanCor: Verb as Adverbial
+        "VG": "VERB",
+        "VK": "VERB",
+        "VN": "NOUN",  # HKCanCor: Verb with Nominal Function
+        "VU": "AUX",
+        "VUG": "AUX",
+        "W": "PUNCT",  # HKCanCor: Punctuation
+        "X": "X",  # HKCanCor: Unclassified Item
+        "XA": "ADJ",
+        "XB": "ADJ",
+        "XC": "CCONJ",
+        "XD": "ADV",
+        "XE": "INTJ",
+        "XJ": "X",
+        "XJB": "PROPN",
+        "XJN": "NOUN",
+        "XJNT": "PROPN",
+        "XJNZ": "PROPN",
+        "XJV": "VERB",
+        "XJA": "X",
+        "XL1": "INTJ",
+        "XM": "NUM",
+        "XN": "NOUN",
+        "XNG": "NOUN",
+        "XNR": "PROPN",
+        "XNS": "PROPN",
+        "XNT": "PROPN",
+        "XNX": "NOUN",
+        "XNZ": "PROPN",
+        "XO": "X",
+        "XP": "ADP",
+        "XQ": "NOUN",
+        "XR": "PRON",
+        "XS": "PROPN",
+        "XT": "NOUN",
+        "XV": "VERB",
+        "XVG": "VERB",
+        "XVN": "NOUN",
+        "XX": "X",
+        "Y": "PART",  # HKCanCor: Modal Particle
+        "YG": "PART",  # HKCanCor: Modal Particle Morpheme
+        "Y1": "PART",
+        "Z": "ADJ",  # HKCanCor: Descriptive
+    }
+
+    def _info(self):
+
+        pos_tags_prf = datasets.Sequence(datasets.features.ClassLabel(names=[tag for tag in self.pos_map.keys()]))
+
+        pos_tags_ud = datasets.Sequence(
+            datasets.features.ClassLabel(names=[tag for tag in set(self.pos_map.values())])
+        )
+
+        features = datasets.Features(
+            {
+                "conversation_id": datasets.Value("string"),
+                "speaker": datasets.Value("string"),
+                "turn_number": datasets.Value("int16"),
+                "tokens": datasets.Sequence(datasets.Value("string")),
+                "transcriptions": datasets.Sequence(datasets.Value("string")),
+                "pos_tags_prf": pos_tags_prf,
+                "pos_tags_ud": pos_tags_ud,
+            }
+        )
+
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=features,
+            supervised_keys=None,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        """Returns SplitGenerators."""
+        data_dir = os.path.join(dl_manager.download_and_extract(_URL), "utf8")
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "data_dir": data_dir,
+                    "split": "train",
+                },
+            )
+        ]
+
+    def _generate_examples(self, data_dir, split):
+        """ Yields examples. """
+
+        downloaded_files = [os.path.join(data_dir, fn) for fn in sorted(os.listdir(data_dir))]
+        for filepath in downloaded_files:
+            # Each file in the corpus contains one conversation
+            with open(filepath, encoding="utf-8") as f:
+                xml = f.read()
+                # Add dummy root node to form valid tree
+                xml = "<root>" + xml + "</root>"
+                tree = ET.fromstring(xml)
+
+            # Extract dialogue metadata
+            info = [line.strip() for line in tree.find("info").text.split("\n") if line and not line.endswith("END")]
+            tape_number = "".join(info[0].split("-")[1:])
+            date_recorded = "".join(info[1].split("-")[1:])
+
+            turn_number = -1
+            for sent in tree.findall("sent"):
+                for child in sent.iter():
+                    if child.tag == "sent_head":
+                        current_speaker = child.text.strip()[:-1]
+                        turn_number += 1
+                    elif child.tag == "sent_tag":
+                        tokens = []
+                        pos_prf = []
+                        pos_ud = []
+                        transcriptions = []
+                        current_sentence = [w.strip() for w in child.text.split("\n") if w and not w.isspace()]
+                        for w in current_sentence:
+                            token_data = w.split("/")
+                            tokens.append(token_data[0])
+                            transcriptions.append(token_data[2])
+
+                            prf_tag = token_data[1].upper()
+                            ud_tag = self.pos_map.get(prf_tag, "X")
+                            pos_prf.append(prf_tag)
+                            pos_ud.append(ud_tag)
+
+                        num_tokens = len(tokens)
+                        num_pos_tags = len(pos_prf)
+                        num_transcriptions = len(transcriptions)
+
+                        assert len(tokens) == len(
+                            pos_prf
+                        ), "Sizes do not match: {nw} vs {np} for tokens vs pos-tags in {fp}".format(
+                            nw=num_tokens, np=num_pos_tags, fp=filepath
+                        )
+                        assert len(pos_prf) == len(
+                            transcriptions
+                        ), "Sizes do not match: {np} vs {nt} for tokens vs pos-tags in {fp}".format(
+                            np=num_pos_tags, nt=num_transcriptions, fp=filepath
+                        )
+
+                        # Corpus doesn't come with conversation-level ids, and
+                        # multiple texts can correspond to the same tape number,
+                        # date, and speakers.
+                        # The following workaround prepends metadata with the
+                        # first few transcriptions in the conversation
+                        # to create an identifier.
+                        id_from_transcriptions = "".join(transcriptions[:5])[:5].upper()
+                        id_ = "{tn}-{rd}-{it}".format(tn=tape_number, rd=date_recorded, it=id_from_transcriptions)
+                        yield id_, {
+                            "conversation_id": id_,
+                            "speaker": current_speaker,
+                            "turn_number": turn_number,
+                            "tokens": tokens,
+                            "transcriptions": transcriptions,
+                            "pos_tags_prf": pos_prf,
+                            "pos_tags_ud": pos_ud,
+                        }