File size: 2,263 Bytes
2eedd93 5c3a4f4 2eedd93 3d33521 2eedd93 5c3a4f4 2eedd93 3d33521 2eedd93 0470f26 705e6a7 0470f26 705e6a7 0470f26 5c3a4f4 2eedd93 3d33521 5c3a4f4 2eedd93 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 |
---
license: apache-2.0
---
# UK & Ireland Accent Classification Model
This is a model to classify and identify the accent of a UK or Ireland speaker among one of the following accents:
* Irish English
* Midlands English
* Northern English
* Scottish English
* Southern English
* Welsh English
The model implements transfer learning feature extraction using [Yamnet](https://tfhub.dev/google/yamnet/1) model in order to train a model.
### Yamnet Model
Yamnet is an audio event classifier trained on the AudioSet dataset to predict audio events from the AudioSet ontology. It is available on TensorFlow Hub.
Yamnet accepts a 1-D tensor of audio samples with a sample rate of 16 kHz.
As output, the model returns a 3-tuple:
- Scores of shape `(N, 521)` representing the scores of the 521 classes.
- Embeddings of shape `(N, 1024)`.
- The log-mel spectrogram of the entire audio frame.
We will use the embeddings, which are the features extracted from the audio samples, as the input to our dense model.
### Dense Model
The dense model that we used consists of:
- An input layer which is embedding output of the Yamnet classifier.
- 4 dense hidden layers and 4 dropout layers
- An output dense layer.
<details>
<summary>View Model Plot</summary>
</details>
### Results
The model achieved the following results:
Results | Training | Validation
-----------|-----------|------------
Accuracy | 55% | 51%
AUC | 0.9090 | 0.8911
d-prime | 1.887 | 1.743
And the confusion matrix for the validation set is:
---
## Dataset
The dataset used is the
[Crowdsourced high-quality UK and Ireland English Dialect speech data set](https://openslr.org/83/)
which consists of a total of 17,877 high-quality audio wav files.
This dataset includes over 31 hours of recording from 120 vounteers who self-identify as
native speakers of Southern England, Midlands, Northern England, Wales, Scotland and Ireland.
For more info, please refer to the above link or to the following paper:
[Open-source Multi-speaker Corpora of the English Accents in the British Isles](https://aclanthology.org/2020.lrec-1.804.pdf)
---
## Demo
A demo is available in HuggingFace Spaces ... |