src/morph_hrr/morphemes.py

"""Dependency-free English morphological segmentation.

Splits a word into ``(prefix, root, suffix)`` using curated affix lists with
longest-match and minimum-root guards. Deliberately simple and predictable:
no dictionary, no learned model, no external deps. Imperfect (it won't undo
consonant doubling like ``running -> run``) — documented as future work — but it
handles the common derivational/inflectional cases the tokenizer relies on.
"""
from __future__ import annotations

# Conservative, high-precision derivational prefixes (min root kept >= MIN_ROOT).
PREFIXES: tuple[str, ...] = (
    "anti", "auto", "circum", "contra", "counter", "dis", "extra", "fore",
    "hyper", "hypo", "in", "im", "il", "ir", "inter", "intra", "mal", "mid",
    "mis", "multi", "non", "out", "over", "poly", "post", "pre", "proto",
    "pseudo", "re", "retro", "semi", "sub", "super", "supra", "sur", "trans",
    "tri", "ultra", "un", "under",
)

# Inflectional + common derivational suffixes, tried longest-first.
SUFFIXES: tuple[str, ...] = (
    # 5+ chars
    "ation", "ation", "ically", "iosity", "itious", "aceous", "acious",
    "ality", "ative", "ator", "fully", "ially", "ables", "ibles", "iness",
    # 4 chars
    "able", "ably", "ance", "ence", "ency", "hood", "ible", "ical", "iest",
    "isms", "ists", "ment", "ness", "ship", "tion", "wise",
    # 3 chars
    "acy", "age", "al", "ant", "ate", "dom", "ees", "ers", "est", "ful",
    "ial", "ies", "ily", "ing", "ion", "ish", "ism", "ist", "ity", "ive",
    "ize", "ity", "less", "ous", "ity",
    # 2 chars
    "al", "ed", "en", "er", "es", "ic", "ly", "or", "ty",
    # 1 char (high-precision only)
    "s",
)

# Keep roots at least this long after affix removal (avoids eating short words).
MIN_ROOT = 3

_SUFFIXES_BY_LEN = tuple(sorted({s for s in SUFFIXES if s}, key=len, reverse=True))


def segment(word: str) -> tuple[str, str, str]:
    """Return ``(prefix, root, suffix)`` for ``word``; affixes are "" when absent."""
    w = (word or "").strip().lower()
    if not w.isalpha():
        return ("", w, "")

    prefix = ""
    for p in PREFIXES:
        if w.startswith(p) and len(w) - len(p) >= MIN_ROOT:
            prefix = p
            break
    root = w[len(prefix):]

    suffix = ""
    for s in _SUFFIXES_BY_LEN:
        if root.endswith(s) and len(root) - len(s) >= MIN_ROOT:
            suffix = s
            root = root[: len(root) - len(s)]
            break

    return (prefix, root, suffix)