This repository was forked from Reykjavik University, [LVL]: https://github.com/cadia-lvl/regina_normalizer
regina_normalizer is a normalization system based on regular expressions. Its core functionality is to expand digits, symbols and abbreviations contained in an input text so that it can be converted to phoneme representation and then read by a speech synthesizer.
- Clone the repository
- Install requirements (preferably in a virtual environment)
- Run
pip install -e .from the root directory
To try the normalization out from the command line:
python3 regina_normalizer/main.py {sentence-to-be-normalized}
for example
python3 regina_normalizer/main.py "10.010.000 kr aukalega"
output:
[['tíu milljónir og tíu þúsund krónur aukalega']]
Two additional arguments can be given to the command line interface: -d for domain and -f for format.
The normalizer has a special processing for domain "sport", but the default domain is "other", i.e. no special treatment for the input. The default format argument is "plain", returning a list of normalized sentences from the input (see example above). Using -f "tokens" returns a list of tuples, where one can see how each token was normalized:
python3 regina_normalizer/main.py "Það eru 54 km eftir. Kostar þetta 1263 kr?" -f "tokens"
output:
[[('Það', 'það'), ('eru', 'eru'), ('54', ' fimmtíu og fjórir'), ('km', 'kílómetrar'), ('eftir', 'eftir'), ('.', '.')], [('kostar', 'kostar'), ('þetta', 'þetta'), ('1263', ' tólf hundruð sextíu og þrjár'), ('kr', 'krónur'), ('?', '?')]]
Note that all ouptut is lowercased.
- Anna Björk Nikulásdóttir email (this fork)
- Helga Svala Sigurðardóttir, Reykjavík University (original regina_normalizer)
This project, both the original regina_normalizer and this derivative, is funded by the Language Technology Programme for Icelandic 2019-2023. The programme, which is managed and coordinated by Almannarómur, is funded by the Icelandic Ministry of Education, Science and Culture.