pyinfopedia

Typed Python client for Infopédia — the European-Portuguese dictionary by Porto Editora.

Each word page is parsed into a typed Entry: headword, IPA pronunciation(s), syllabification, etymology, grammatical categories with numbered senses, set phrases, inflected forms, and the sidebar related-word lists (synonyms, rhymes, neighbours…).

Built for Portuguese NLP/lexicon work — it correctly separates heterophonic homographs (same spelling, different pronunciation per reading), which is what makes it useful for grapheme-to-phoneme and disambiguation tasks.

Install

pip install pyinfopedia            # or: uv pip install pyinfopedia
pip install pyinfopedia[stealth]   # + curl_cffi for Cloudflare bypass

Depends on unblock_requests for transport.

Quick start

import pyinfopedia

entry = pyinfopedia.get_word("casa")
print(entry.pronunciation)              # ˈkazɐ
print(entry.categories[0].pos)          # nome feminino
print(entry.categories[0].senses[0].definition)

for r in pyinfopedia.search("cas"):     # prefix autocomplete
    print(r.word, r.url)

Heterophonic homographs

Infopédia lists one entry block per pronunciation; pyinfopedia keeps them separate, tying each grammatical category (and its senses) to the reading it belongs to:

entry = pyinfopedia.get_word("sede")
for cat in entry.categories:
    print(cat.pronunciation, cat.pos, "->", cat.senses[0].definition)
# ˈsɛdɨ nome feminino -> lugar onde alguém se pode sentar ou fixar   (seat / HQ)
# ˈsedɨ nome feminino -> sensação causada pela necessidade de beber  (thirst)

The two readings carry disjoint senses — corte (cut ˈkɔɾtɨ / court ˈkoɾtɨ), molho (sauce ˈmoʎu / bundle ˈmɔʎu), forma (mould ˈfoɾmɐ / shape ˈfɔɾmɐ) all behave the same way. See examples/heterophones.py.

Transport / Cloudflare

All HTTP goes through Transport, a wrapper over unblock_requests.CloudflareSession. Pick a mode when the default is blocked:

from pyinfopedia import Infopedia
client = Infopedia(mode="curl_cffi")                                   # impersonate a browser
client = Infopedia(mode="flaresolverr",
                   flaresolverr_url="http://192.168.1.116:8191")        # FlareSolverr

Modes: requests · curl_cffi · flaresolverr · wayback.

Verbs

from pyinfopedia import get_verb
conj = get_verb("jogar")
print(conj.first_person_singular())     # jogo
print(conj.present_indicative())

Datasets

pyinfopedia.dataset exports JSONL/CSV for a word list — see examples/build_dataset.py.

Development

pytest -m "not live"            # offline parser/model tests (HTML fixtures)
PYINFOPEDIA_FLARESOLVERR=http://host:8191 pytest -m live    # hit the live site

Apache-2.0 · JarbasAi <jarbasai@mailfence.com>. Data belongs to Porto Editora / Infopédia; this is an unofficial client — respect their terms and rate limits.