pere commited on
Commit
050f09d
1 Parent(s): dcda64c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +22 -34
README.md CHANGED
@@ -48,7 +48,7 @@ task_ids:
48
  - **Point of Contact:** [Per Erik Solberg](mailto:per.solberg@nb.no)
49
 
50
  The Norwegian Parliament Speech Corpus (NPSC) is a corpus for training a Norwegian ASR (Automatic Speech Recognition) models.
51
- **!!!NEDS TO BE UPDATED!!! We need a real description here. About one paragraph summing up that is on the main web page, and telling how this dataset is the same but different - ie it is in a streaming format...**
52
 
53
  ## How to Use
54
  ```python
@@ -81,15 +81,15 @@ The NPSC dataset contains json lines with language training data. Here is an exa
81
 
82
 
83
  ```
84
- ## Data Fields
85
  |**id:** | String with id to source of line and a unique identifier|
86
  |:-----------|:------------|
87
  |**sentence_order** | String with order of sentence |
88
  |**speaker id** | Integer id of speaker |
89
  | **speaker_name** | String name of speaker |
90
  | **sentence_text** | String sentence text |
91
- | **sentence_language_code** | String sentence text |
92
- | **text** | String sentence text |
93
  | **start_time** | int start time |
94
  | **end_time** | int end time |
95
  | **normsentence_text** | String normalised sentence text |
@@ -101,11 +101,10 @@ The NPSC dataset contains json lines with language training data. Here is an exa
101
 
102
  ### Dataset Creation
103
  We are providing a **train** and a **validation** split. The standard size of the validation is a single 1GB file, while train is sharded in 1GB chunks.
104
- All files are gzipped. There is also a **test** split available for the dataset but this is hidden. Please contact [Per Erik Solberg](mailto:per.erik.solberg@nb.no) for access to the test set.
105
 
106
- Build date: 22012022
107
 
108
- #### Initial Data Collection and Curation
109
  The procedure for the dataset creation is described in detail in our paper.
110
 
111
 
@@ -117,43 +116,32 @@ The procedure for the dataset creation is described in detail in our paper.
117
  | Word count | 1,2 million |
118
  | Sentence count | 64.531 |
119
  | Language distribution | Nynorsk: 12,8%|
120
- | | Bokmål: 87,2%%|
121
  | Gender distribution | Female: 38,3% |
122
  | | Male: 61.7% |
123
 
124
- ## Considerations for Using the Data
125
- This corpus contains speech data and is allowed to be used outside the National Library of Norway for speech recognition technology purposes.
126
 
127
- ### Dataset Curators
128
- [Freddy Wetjen](mailto:Freddy.wetjen@nb.no) and [Andre Kaasen](mailto:andre.kasen@nb.no) **!!!NEDS TO BE UPDATED!!! Underline the origin of the data and the project. Add Javier de la Rosa and Per Egil Kummervold to the people having worked on this!!!**
 
 
129
 
130
- ### Licensing Information
131
- Licensed for use outside the National Library of Norway.
132
 
133
  ## License
134
- The dataset is released under the [CC-ZERO-license](https://creativecommons.org/publicdomain/zero/1.0/). The curation of the data
135
 
136
  ### Citation Information
137
- We are preparing an article with detailed information about this corpus. Until it is published, please cite out paper discussing the first version of this corpus:
138
  ```
139
 
140
- **!!!NEDS TO BE UPDATED!!!**
141
- @inproceedings{kummervold-etal-2021-operationalizing,
142
- title = {Operationalizing a National Digital Library: The Case for a {N}orwegian Transformer Model},
143
- author = {Kummervold, Per E and
144
- De la Rosa, Javier and
145
- Wetjen, Freddy and
146
- Brygfjeld, Svein Arne",
147
- booktitle = {Proceedings of the 23rd Nordic Conference on Computational Linguistics (NoDaLiDa)},
148
- year = "2021",
149
- address = "Reykjavik, Iceland (Online)",
150
- publisher = {Link{"o}ping University Electronic Press, Sweden},
151
- url = "https://aclanthology.org/2021.nodalida-main.3",
152
- pages = "20--29",
153
- abstract = "In this work, we show the process of building a large-scale training set from digital and digitized collections at a national library.
154
- The resulting Bidirectional Encoder Representations from Transformers (BERT)-based language model for Norwegian outperforms multilingual BERT (mBERT) models
155
- in several token and sequence classification tasks for both Norwegian Bokm{aa}l and Norwegian Nynorsk. Our model also improves the mBERT performance for other
156
- languages present in the corpus such as English, Swedish, and Danish. For languages not included in the corpus, the weights degrade moderately while keeping strong multilingual properties. Therefore,
157
- we show that building high-quality models within a memory institution using somewhat noisy optical character recognition (OCR) content is feasible, and we hope to pave the way for other memory institutions to follow.",
158
  }
159
  ```
 
48
  - **Point of Contact:** [Per Erik Solberg](mailto:per.solberg@nb.no)
49
 
50
  The Norwegian Parliament Speech Corpus (NPSC) is a corpus for training a Norwegian ASR (Automatic Speech Recognition) models.
51
+ **!!!PER ERIK - COULD YOU PLEASE UPDATE THIS? 1-2 paragraphs. FOR PEOPLE NOT BOTHERING TO READ THE ARTICLE!!! We need a real description here. About one paragraph summing up that is on the main web page, and telling how this dataset is the same but different - ie it is in a streaming format... This is X days of transcripts of parliamentary speeches. Each day XX hours. Most without any manuscript. They are split into XXXXX based on.... The id allows you to combine this into single sound files....**
52
 
53
  ## How to Use
54
  ```python
 
81
 
82
 
83
  ```
84
+ ## Data Fields (!!!!PER ERIK - We really need a descrption here - and maybe some cleaning!!!)
85
  |**id:** | String with id to source of line and a unique identifier|
86
  |:-----------|:------------|
87
  |**sentence_order** | String with order of sentence |
88
  |**speaker id** | Integer id of speaker |
89
  | **speaker_name** | String name of speaker |
90
  | **sentence_text** | String sentence text |
91
+ | **sentence_language_code** | String sentence text (!!!PER ERIK - LIST ALL ALTERNATIVES!!!)|
92
+ | **text** | String sentence text. This is a copy of "sentence_text". It is included here to make it more convenient to interleave with other datasets.|
93
  | **start_time** | int start time |
94
  | **end_time** | int end time |
95
  | **normsentence_text** | String normalised sentence text |
 
101
 
102
  ### Dataset Creation
103
  We are providing a **train** and a **validation** split. The standard size of the validation is a single 1GB file, while train is sharded in 1GB chunks.
104
+ All files are gzipped. There is also a **test** split available for the dataset but this is hidden. Please contact [Per Erik Solberg](mailto:per.erik.solberg@nb.no) for access to the test set. !!!!!PER ERIK - Verify that this is correct!!!!!!!!!!!!!!!
105
 
 
106
 
107
+ #### Initial Data Collection
108
  The procedure for the dataset creation is described in detail in our paper.
109
 
110
 
 
116
  | Word count | 1,2 million |
117
  | Sentence count | 64.531 |
118
  | Language distribution | Nynorsk: 12,8%|
119
+ | | Bokmål: 87,2%|
120
  | Gender distribution | Female: 38,3% |
121
  | | Male: 61.7% |
122
 
 
 
123
 
124
+ !!!PER ERIK - It would be great to know how many hours of Bokmål/Nynorsk/English(?) here!!!!!
125
+
126
+ ## Considerations for Using the Data
127
+ This corpus contains speech data. All recordings are of parliament members in a public setting, and can be distributed without any restrains.
128
 
129
+ ### Dataset Creators and Curators
130
+ The content of the dataset was done by Språkbanken. !!!!!FILL IN HERE PER ERIK!!!!!!!!!!!!!!!. [Javier de la Rosa](mailto:javier.rosa@nb.no), [Freddy Wetjen](mailto:Freddy.wetjen@nb.no), [Per Egil Kummervold](mailto:per.kummervold@nb.no), and [Andre Kaasen](mailto:andre.kasen@nb.no) all contributed in making this into a HuggingFace Dataset. Thanks to the HuggingFace team for assistance.
131
 
132
  ## License
133
+ The sound and the transcriptions are released under the [CC-ZERO-license](https://creativecommons.org/publicdomain/zero/1.0/). The curation of the HuggingFace Dataset is released under [CC-BY-SA-3-license](https://creativecommons.org/licenses/by-sa/3.0/).
134
 
135
  ### Citation Information
136
+ The following article gives detailed information about the corpus. Please refer to the article and this page if you are using this dataset:
137
  ```
138
 
139
+ @misc{solberg2022norwegian,
140
+ title={The Norwegian Parliamentary Speech Corpus},
141
+ author={Per Erik Solberg and Pablo Ortiz},
142
+ year={2022},
143
+ eprint={2201.10881},
144
+ archivePrefix={arXiv},
145
+ primaryClass={cs.CL}
 
 
 
 
 
 
 
 
 
 
 
146
  }
147
  ```