"""Implements experimental features for formal concept analysis."""
from __future__ import annotations
from abc import ABC, abstractmethod
from warnings import warn
from toolz import memoize
import textnets as tn
[docs]
class FormalContext(ABC):
"""
Abstract base class providing experimental FCA features.
Textnets inherits methods from this class for treating its biadjacency
matrix as a formal context.
"""
@property
@abstractmethod
def m(self):
pass
@property
def context(self):
"""Return formal context of terms and documents."""
return self._formal_context(alpha=tn.params["ffca_cutoff"])
@memoize
def _formal_context(self, alpha: float):
# The biadjacency matrix is a "fuzzy formal context." We can binarize
# it by using a cutoff. This is known as an alpha-cut. This feature is
# experimental.
try:
from concepts import Context
except ImportError:
warn("Install textnets[fca] to use FCA features.")
raise
crisp = self.m.map(lambda x: x >= alpha)
reduced = crisp[crisp.any(axis=1)].loc[:, crisp.any(axis=0)]
objects = reduced.index.tolist()
properties = reduced.columns.tolist()
bools = reduced.to_numpy().tolist()
return Context(objects, properties, bools)
