mict-zhaw commited on
Commit
eb870ed
1 Parent(s): 5464bfa

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +37 -11
README.md CHANGED
@@ -36,13 +36,15 @@
36
 
37
  ### Dataset Summary
38
 
39
- The dataset comprises audio recordings featuring spontaneous speech from young Swiss learners of English, specifically from students in grades 4 to 6. The audio data was collected during engaging language activities designed to elicit natural speech patterns and includes corresponding verbatim transcripts where learners' errors are annotated. The learners participated in a variety of tasks such as role plays, guessing games, and mock interviews, providing a rich source of authentic conversational data.
40
-
 
41
  The transcripts were prepared by a professional transcription service, and each recording was associated with detailed metadata, including school grade, recording conditions, and error annotations.
42
 
43
  Access to the dataset is regulated to ensure the confidentiality and ethical use of the sensitive data.
44
  To use the ChaLL dataset, you need to download it manually. Follow the instructions provided (**yet to be detailed**) for downloading the data. Once you have downloaded the files, please extract all files into a single folder.
45
 
 
46
  Examples in this dataset are generated using the `soundfile` library (for reading and chunking). To handle the audio data correctly, you need to install the soundfile library.
47
 
48
  ```shell
@@ -80,37 +82,51 @@ A typical data instance in this dataset include an audio file, its full transcri
80
 
81
  ### Data Splits
82
 
83
- [More Information Needed]
 
 
 
 
 
 
 
84
 
85
  ## Dataset Creation
86
 
87
  ### Curation Rationale
88
 
89
- The dataset was curated to address the lack of resources for developing ASR systems that can handle the unique characteristics of children's speech, especially non-native English speakers, and to facilitate the development of educational tools that provide corrective feedback on language learning.
 
 
90
 
91
  ### Source Data
92
 
93
  #### Initial Data Collection and Normalization
94
 
95
- Audio recordings were collected in controlled environments mimicking naturalistic settings to elicit genuine conversational speech from children. The recordings were then annotated professionally to include detailed error tags.
 
96
 
97
  #### Who are the source language producers?
98
 
99
- The language producers are Swiss children from grades 4 to 6, engaged in language learning activities designed to elicit natural speech.
100
 
101
  ### Annotations
102
 
103
  #### Annotation process
104
 
105
- [More Information Needed]
 
 
106
 
107
  #### Who are the annotators?
108
 
109
- Transcriptions and error annotations were performed by a professional service, following strict guidelines to ensure accuracy and consistency in error marking.
110
 
111
  ### Personal and Sensitive Information
112
 
113
- The dataset contains audio recordings of children, and all personal identifiers have been removed to protect privacy. Access to the data is restricted and requires approval to ensure ethical use.
 
 
114
 
115
  ## Considerations for Using the Data
116
 
@@ -124,7 +140,8 @@ Given the specific demographic (Swiss German-speaking schoolchildren), the datas
124
 
125
  ### Other Known Limitations
126
 
127
- [More Information Needed]
 
128
 
129
  ## Additional Information
130
 
@@ -138,7 +155,16 @@ The dataset was curated by researchers at PHZH, UZH and Zhaw, with collaboration
138
 
139
  ### Citation Information
140
 
141
- [More Information Needed]
 
 
 
 
 
 
 
 
 
142
 
143
  ### Contributions
144
 
 
36
 
37
  ### Dataset Summary
38
 
39
+ This dataset contains audio recordings of spontaneous speech by young learners of English in Switzerland.
40
+ The recordings capture various language learning tasks designed to elicit authentic communication from the students.
41
+ The dataset includes detailed verbatim transcriptions with annotations for errors made by the learners.
42
  The transcripts were prepared by a professional transcription service, and each recording was associated with detailed metadata, including school grade, recording conditions, and error annotations.
43
 
44
  Access to the dataset is regulated to ensure the confidentiality and ethical use of the sensitive data.
45
  To use the ChaLL dataset, you need to download it manually. Follow the instructions provided (**yet to be detailed**) for downloading the data. Once you have downloaded the files, please extract all files into a single folder.
46
 
47
+
48
  Examples in this dataset are generated using the `soundfile` library (for reading and chunking). To handle the audio data correctly, you need to install the soundfile library.
49
 
50
  ```shell
 
82
 
83
  ### Data Splits
84
 
85
+ #### `asr_acl`
86
+
87
+ For the experiments in this paper, we split the dataset into five distinct folds of similar duration
88
+ (about 16h each), where each class (and therefore also each speaker) occurs in only one fold.
89
+ To simulate the use case of the ASR system being confronted with a new class of learners, each fold
90
+ contains data from a mix of grades. The following figure visualises the duration and grade distribution of each fold.
91
+
92
+ ![Chall Folds](/chall_data_folds_v1.svg)
93
 
94
  ## Dataset Creation
95
 
96
  ### Curation Rationale
97
 
98
+ The dataset was created to address the need for ASR systems that can handle childrens spontaneous speech
99
+ and preserve their errors to provide effective corrective feedback in language learning environments.
100
+
101
 
102
  ### Source Data
103
 
104
  #### Initial Data Collection and Normalization
105
 
106
+ Audio data was collected from primary school students aged 9 to 14 years, performing language learning tasks in pairs, trios, or individually. The recordings were made at schools and universities, and detailed verbatim transcriptions were created by a transcription agency, following specific guidelines.
107
+
108
 
109
  #### Who are the source language producers?
110
 
111
+ The source data producers include primary school students from German-speaking Switzerland, aged 9 to 14 years, participating in language learning activities.
112
 
113
  ### Annotations
114
 
115
  #### Annotation process
116
 
117
+ The transcription and annotation process was outsourced to a transcription agency, following detailed guidelines for error annotation,
118
+ including symbols for grammatical, lexical, and pronunciation errors, as well as German word usage.
119
+
120
 
121
  #### Who are the annotators?
122
 
123
+ The annotators were professionals from a transcription agency, trained according to specific guidelines provided by the project team.
124
 
125
  ### Personal and Sensitive Information
126
 
127
+ The dataset contains audio recordings of minors.
128
+ All data was collected with informed consent from legal guardians, and recordings are anonymized to protect the identities of the participants.
129
+
130
 
131
  ## Considerations for Using the Data
132
 
 
140
 
141
  ### Other Known Limitations
142
 
143
+ The outsourcing of transcription and error annotations always poses a risk of yielding erroneous data, since most
144
+ transcribers are not trained in error annotation.
145
 
146
  ## Additional Information
147
 
 
155
 
156
  ### Citation Information
157
 
158
+ ```bibtex
159
+ @inproceedings{
160
+ anonymous2024errorpreserving,
161
+ title={Error-preserving Automatic Speech Recognition of Young English Learners' Language},
162
+ author={Anonymous},
163
+ booktitle={The 62nd Annual Meeting of the Association for Computational Linguistics},
164
+ year={2024},
165
+ url={https://openreview.net/forum?id=XPIwvlqIfI}
166
+ }
167
+ ```
168
 
169
  ### Contributions
170