Datasets:
text stringlengths 1 8 |
|---|
أن |
أيضا |
إلى |
إن |
بل |
حتى |
على |
عن |
في |
قد |
كما |
لأن |
لأنه |
لا |
لكي |
ما |
من |
هذا |
هذه |
ولا |
ٱلذي |
أما |
أن |
أنا |
أيضا |
إذ |
إلا |
إلى |
إن |
التي |
الذي |
الذين |
بل |
به |
بها |
ثم |
حتى |
ذلك |
على |
عن |
عند |
غير |
فإن |
في |
فيه |
فيها |
قد |
كان |
كما |
لأن |
لا |
ما |
من |
هذا |
هذه |
ولا |
ولكن |
আছিল |
আছে |
আমি |
আৰু |
এই |
এজন |
এতিযা |
এনে |
ওপৰত |
কথা |
কাৰণ |
কাৰণে |
কি |
কিনতু |
কিযনো |
কৰক |
কৰা |
কৰি |
কৰিব |
কৰিবলৈ |
কৰে |
জনে |
তাৰ |
তেওঁ |
তেওঁক |
তেওঁলোকক |
তেওঁলোকে |
তেওঁলোকৰ |
তেওঁৰ |
তেতিযা |
থকা |
দৰে |
নিজ |
নিজৰ |
পৰা |
বাবে |
বিষযে |
বিৰুদধে |
বুলি |
মই |
মোৰ |
যদি |
যি |
target-stopwords
Per-language function-word lists, induced from that language's own Bible text — frequency +
dispersion (the classic corpus-linguistics stopword-induction recipe), then RESCUED against the
language's own alignment output + a source-anchored content signal so genuinely frequent CONTENT words
("God", "Lord") are never dropped. See the lexeme-aligner's target_stopwords.py for the mechanism: a
candidate is rescued if it concentrates most of its aligned mass on one prior-pack content lexeme (≥40%
share) — a true function word instead scatters thinly across dozens of distinct lexemes.
Many of the covered languages have no existing curated stopword list anywhere — this is a reusable resource for search, IR, topic modeling, or any NLP task needing one in these languages.
CC0-1.0 — derived word-frequency statistics, no source text redistributed. See manifest.json for
per-language stats + content hashes.
- Downloads last month
- 18