Lightweight character embedding framework using the GDk9 hollow-vector model. Maps any character to a 3-element numeric vector — normalised code point, glyph hole count, and alphabetic flag — then aggregates across strings for analysis, comparison, and ML integration.

python demo.py "Hello World!"
# Character vectors:
#   'H' -> [0.573, 0.0, 1.0]
#   'o' -> [0.610, 1.0, 1.0]   # 'o' has one enclosed loop
# Mean vector: [0.5886, 0.4167, 0.8333]
# String value (magnitude): 1.1203

The GDk9 hollow vector captures three features per character:

These simple features preserve structural character properties while remaining fully numeric — making them directly usable in clustering, anomaly detection, and classification pipelines.

git clone https://github.com/ao3575911/gdk9-alphabet.git
cd gdk9-alphabet
python demo.py "Example"

Requires Python ≥ 3.11. Zero external dependencies.

from gdk9_alphabet import get_char_vector

get_char_vector("A")   # [0.504, 1.0, 1.0]  — code, 1 hole, is alpha
get_char_vector("B")   # [0.511, 2.0, 1.0]  — 2 holes
get_char_vector("5")   # [0.417, 0.0, 0.0]  — digit, no holes

from vector_ops import string_vector, string_value

vec = string_vector("Hello")    # component-wise mean across all chars
val = string_value("Hello")     # Euclidean magnitude — scalar metric

# Compare strings
string_value("AAAA") < string_value("BBBB")   # quantitative ordering

from dcg import build_dcg, format_dcg

graph = build_dcg("LOLLOL")
print(format_dcg(graph))
# L -> O (2)
# O -> L (2)
# L -> L (1)

The DCG captures transition patterns: edge weight = number of consecutive occurrences. Useful for sequence analysis, Markov models, and visualisation.

• Custom alphabets — update SUPPORTED_CHARACTERS with hole counts for your symbol set
• Additional features — add stroke counts, phonetic properties, or frequency statistics
• Graph analytics — feed the DCG into networkx for PageRank, centrality, shortest-path
• ML pipelines — use string_vector outputs as feature vectors for classifiers

MIT — © 2025 Adam Grange