Development-oriented todo list.
- Update demo notebook to use functions now integrated into
feature_vector.py
:symbol_to_feature_vector
feature_vector_to_symbols
extension_to_symbols
symbol_to_feature_dict
pfv_to_fd
- Update demo notebook with whatever rewrite-rule related functionality is there.
- Replace notes pdf with something more recent (slides?)
- Add more unit tests for
feature_vector.py
functions (in general). - In particular, add unit tests as multiple implementations of core functions in
feature_vector.py
are added.
- update/trim dependencies.
requirements
file.- add license.
- Split up the overgrown feature_vector.py module into a few natural submodules drawn from, eg.
- feature dictionary <-> pfv
- pfv creation
- feature-vectors-as-functions-ops
- (semi)lattice operations
- priority union/functions/operators
- property/structure/morphism checking
- maybe very high level functions related to identifying natural classes and processes
- Refactor the following functions for stack-compatibility if possible:
left_inv_priority_union
right_inv_priority_union
pseudolinear_inverse_possible
pseudolinear_inverse
pseudolinear_decomposition
normalize
- Add functionality for solving the function analogues of the three problems described in the
readme
.
- Add support for more memory-efficient and memory-safe implementations from existing code.
- Add support for
joblib
from existing code. - Add support for
PySpark
and a library (e.g.TileDB
) that will efficiently support concurrent reads/writes. - Add support for
PyTorch
+ gpus from existing code.
- Better support for creating and being able to use an association between symbols (or object dictionaries) and associated ternary ndarray feature vectors.
- Add support for converting between conventional feature vector strings, feature dictionaries, and ternary vectors.
- Add support for taking and creating projections of a loaded inventory.
- Add CLI support.
- Add default support/nice integration with
PanPhon
-exportable data structures/formats. - Add default support/nice integration with
Phonological Corpus Tools
-exportable data structures/formats.