seetm 1.1.0

Converts English tokens into the equivalent Sinhala representation using IPA (International Phonetic Alphabet)

SEETM (Sinhala-English Equivalent Token Mapper) allows creating equivalent token maps and replace them with a base token to avoid OOV tokens and generate a single feature for all equivalent tokens in a Sinhala-English code-switching dataset in rasa-based conversational AIs.

Features

• Allows mapping multiple equivalent tokens into a base token
• Fully supports rasa 2.8.x projects
• Provides an easy-to-use CLI
• Provides an efficient server-based GUI
• Provides a fully-functional custom whitespace tokenizer
• Fully-supports Sinhala in the GUI

What's Cooking?

• Mapping suggestions in the SEETM server GUI
• Automatically generated mappings

Limitations and Known Issues

• Should manually add the SEETM tokenizer to the rasa pipeline or else the token maps are not taking any effect
• IPA-based suggestions could contain slight changes based on th IPA mapping origin. (SEETM uses CMU)

Resources and References

• CMU Pronunciation Dictionary
• eng-to-ipa pip package (GitHub)

📒 Docs: https://seetm.github.io
📦 PyPi: https://pypi.org/project/seetm/1.1.0/
🪵 Full Changelog: https://github.com/SEETM-NLP/seetm/blob/main/CHANGELOG.md