Towards a World-English Language Model for On-Device Virtual Assistants
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910386845908992 |
|---|---|
| author | Jalota, Rricha Verwimp, Lyan Nussbaum-Thom, Markus Mousa, Amr Argueta, Arturo Oualil, Youssef |
| author_facet | Jalota, Rricha Verwimp, Lyan Nussbaum-Thom, Markus Mousa, Amr Argueta, Arturo Oualil, Youssef |
| contents | Neural Network Language Models (NNLMs) for Virtual Assistants (VAs) are generally language-, region-, and in some cases, device-dependent, which increases the effort to scale and maintain them. Combining NNLMs for one or more of the categories is one way to improve scalability. In this work, we combine regional variants of English to build a ``World English'' NNLM for on-device VAs. In particular, we investigate the application of adapter bottlenecks to model dialect-specific characteristics in our existing production NNLMs {and enhance the multi-dialect baselines}. We find that adapter modules are more effective in modeling dialects than specializing entire sub-networks. Based on this insight and leveraging the design of our production models, we introduce a new architecture for World English NNLM that meets the accuracy, latency, and memory constraints of our single-dialect models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_18783 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Towards a World-English Language Model for On-Device Virtual Assistants Jalota, Rricha Verwimp, Lyan Nussbaum-Thom, Markus Mousa, Amr Argueta, Arturo Oualil, Youssef Computation and Language Neural Network Language Models (NNLMs) for Virtual Assistants (VAs) are generally language-, region-, and in some cases, device-dependent, which increases the effort to scale and maintain them. Combining NNLMs for one or more of the categories is one way to improve scalability. In this work, we combine regional variants of English to build a ``World English'' NNLM for on-device VAs. In particular, we investigate the application of adapter bottlenecks to model dialect-specific characteristics in our existing production NNLMs {and enhance the multi-dialect baselines}. We find that adapter modules are more effective in modeling dialects than specializing entire sub-networks. Based on this insight and leveraging the design of our production models, we introduce a new architecture for World English NNLM that meets the accuracy, latency, and memory constraints of our single-dialect models. |
| title | Towards a World-English Language Model for On-Device Virtual Assistants |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2403.18783 |