English loanword expansion — 2026-09-28

Historical expansion record. The later 2026-10-01 suffix update adds media, finite attached forms and the school alternative. This page retains the original 50-key count, held decisions and whole-word matching observations.

This dated record follows the 34-key coverage audit. That audit describes the engine before this change, at commit de2cb93; its company → चोम्पञ् observation is historical. The current change adds company → कम्पनी and 29 more exact English-spelling keys. Together with the original 20-word pilot, the built-in loanword catalogue now has 50 complete-word keys, each with one preferred output. The separate 12 Gregorian month names are not included in that count.

The preferred outputs below are source-assisted project choices for implementation. An institutional occurrence establishes that a Nepali spelling is used in its source context. It does not establish that the spelling is the only valid one, that the English key always has that sense, or that these keys are the most frequent loanwords. There are still no independently reviewed human loanword labels or population-wide coverage and accuracy estimates. The earlier audit and pilot retain their original dates, counts, and evidence scopes.

What was added

The 30 additions consist of one reported gap (company), 12 of the 14 proposals in the earlier research queue, and 17 further source-attested keys. Source IDs link to the access-scope notes below; L27 links to the 2026-09-27 ledger, where some observations were snippet-only. A source ID is evidence of usage for the proposed Nepali spelling, not independent approval of the English-to-Nepali mapping.

English key	New preferred output	Evidence	Scope or decision note
company	कम्पनी	E1	Official business and legal use; a complete base word only.
bus	बस	L27:S11, E9	The earlier PDF evidence was indexed only; a current insurer page directly uses बस टिकट.
cricket	क्रिकेट	L27:S23	Prior sports proposal; direct article.
football	फुटबल	L27:S23	फूटबल was observed separately; variant remains pending.
hotel	होटल	L27:S20, L27:S18	होटेल remains a research variant.
keyboard	किबोर्ड	L27:S02	The classroom page also shows spaced/hyphenated forms.
laptop	ल्यापटप	L27:S04, L27:S03	Prior device proposal.
microphone	माइक्रोफोन	L27:S02, L27:S04	Prior device proposal.
office	अफिस	L27:S16, L27:S17, E10	Prior observations were snippet-only; a current tax-department page directly uses अफिस.
password	पासवर्ड	L27:S04	Prior security-context proposal.
restaurant	रेस्टुरेन्ट	L27:S18, L27:S19	रेष्टुरेन्ट was observed separately; variant remains pending.
ticket	टिकट	L27:S12, L27:S11, E9	Prior source access was limited; a current insurer page directly uses बस टिकट.
wifi	वाइफाइ	L27:S04	वाइफाई remains a research variant.
ambulance	एम्बुलेन्स	E4	Police reporting on emergency response.
battery	ब्याट्री	E3	Police inventory of seized equipment.
car	कार	E2	Vehicle category; kar remains a different Roman Nepali key.
carpet	कार्पेट	E2	Household survey category.
connector	कनेक्टर	E3	Telecom equipment context.
courier	कुरियर	E2, E5	Survey expense category and industrial classification.
digital	डिजिटल	E4, E5	Police and statistical usage, each in a specific context.
drone	ड्रोन	E4, E6	Film Development Board directly pairs ड्रोन with “Drone.”
furniture	फर्निचर	E2	Household survey category.
inverter	इन्भर्टर	E3	Telecom equipment context.
motorcycle	मोटरसाइकल	E2	Vehicle category; spaced variants need separate review.
radio	रेडियो	E2	Household survey category.
router	राउटर	E3	Telecom equipment context.
sofa	सोफा	E2	Household furniture category.
telephone	टेलिफोन	E2	Survey category; phone → फोन already exists.
van	भ्यान	E2	Vehicle category.
website	वेबसाइट	E5	Government text also uses other spellings; the chosen form is a project preference.

Counts: prior 20 + this page's 30 = 50 built-in loanword keys. Of the wider research inventory, 12 of 14 old proposals, three of six examples in the audit (company, telephone, website), and 15 of the separate 46-candidate source scan are now built in. This is a count of project keys, not a percentage of Nepali loanword usage. The original 34-key source ledger is intentionally left unchanged.

Researched words still held

The two remaining old proposals are speaker → स्पिकर (L27:S02; स्पीकर also observed), and train → ट्रेन (L27:S25; the borrowed vehicle noun competes with other senses and with रेल). Neither is a new built-in loanword key.

Three examples from the coverage audit also remain unshipped: fax → फ्याक्स needs a second direct source because its cited PDF evidence was screenshot/index based; wallet → वालेट is directly attested by Nepal Rastra Bank in payment use but has other English meanings; and share → शेयर/सेयर has competing finance spellings and an ordinary English verb sense. No unconditional default or dropdown alternate is inferred from those observations.

The separate 46-candidate source scan supplied 15 of the newly shipped keys. Its other 31 keys are listed below to prevent the research list from being mistaken for implemented coverage. “Later exact-key candidate” means the form looked plausible but was deliberately left for a later batch; “hold” means a concrete spelling, meaning, segmentation, or evidence decision remains. All outputs in this table are proposals, not built-in choices.

Key	Proposed form	Source	Why held
jeep	जीप	E2	Hold: जिप in a government insurer listing also occurs; choose spelling.
scooter	स्कुटर	E2	Hold: स्कूटर in a postal procurement notice also occurs; choose spelling.
casino	क्यासिनो	E2	Later exact-key candidate; review context before another batch.
lighter	लाइटर	E2	Hold: English adjective sense collides.
lipstick	लिपिस्टिक	E2	Later exact-key candidate.
makeup	मेकअप	E2	Later exact-key candidate.
shampoo	स्याम्पू	E2	Hold: corroboration and spelling variants needed.
gas	ग्याँस	E2	Hold: ग्यास in a Law Journal record also occurs; choose spelling.
package	प्याकेज	E2	Hold: holiday package context does not settle other meanings.
machine	मेशिन	E2	Hold: मेसिन/मेशिन variation.
postcard	पोष्टकार्ड	E2	Hold: पोस्टकार्ड also occurs in the indexed Law Commission dictionary PDF; choose spelling.
beautyparlour	ब्युटिपार्लर	E2	Hold: English is commonly two words; key segmentation needs review.
pendrive	पेनड्राइभ	E3	Hold: also typed as pen drive; spacing and aliases need review.
port	पोर्ट	E3	Hold: hardware, harbor, and software senses.
bill	बिल	E3	Hold: common personal name “Bill” needs case behavior.
platform	प्लेटफर्म	E4	Hold: observed in a digital-platform context only.
link	लिंक	E4	Hold: avoid converting actual URLs and literal technical text.
code	कोड	E4	Hold: legal and technical contexts require review.
card	कार्ड	E4, E7	Later exact-key candidate; source contexts differ.
scam	स्क्याम	E4	Hold: one police usage; corroborate.
logo	लोगो	E4	Later exact-key candidate.
page	पेज	E4	Hold: web page and other meanings.
message	मेसेज	E4	Hold: म्यासेज also used; choose spelling.
traffic	ट्राफिक	E4	Later exact-key candidate.
postmortem	पोष्टमार्टम	E4	Hold: spelling decision needed.
prepaid	प्रिपेड	E7	Hold: attested as part of “prepaid card”; review standalone use.
credit	क्रेडिट	E7	Hold: attested as part of “credit card”; review standalone use.
agent	एजेन्ट	E7	Hold: financial-agent context only.
juice	जुस	E8	Later exact-key candidate.
biscuit	बिस्कुट	E8	Later exact-key candidate; विस्कुट also occurs on source page.
chocolate	चकलेट	E8	Hold: चक्लेट also occurs elsewhere; choose spelling.

Thus the recorded research pool is 86 distinct keys: 34 in the prior catalogue, six new probes in the audit, and 46 in the further scan. Of these, 50 are built-in and 36 remain held (two old, three audit, 31 further). This arithmetic describes only the collected pool; it is not a frequency-weighted or exhaustive dictionary. The 31 rows above preserve the scan's research decisions without silently promoting them to accepted mappings.

What typing these words means

• The lookup is for a complete Roman key. convertWord('company') now returns { text: 'कम्पनी', candidates: ['कम्पनी'], ambiguous: false }; convertText('company') and a Nepali-enabled demo field use the same preference. companyma and companyharu still use the phonetic fallback; this batch does not infer Nepali suffixes from an English stem. The old company bug is addressed for the exact key, not for all attached forms.
• Company and COMPANY currently reach कम्पनी because their capitalization does not contain a reserved Shift sound. Other capitalized words may intentionally differ. T, D, S, R and contextual H retain the sound controls described in the typing reference. The expansion does not broadly lowercase those keys.
• The default convertText policy keeps recognizable URL, domain, and email spans literal, such as company.com and company@example.com. An English-only field should remain unattached or have conversion disabled; SahajLipi does not detect arbitrary English prose. See the API guide.
• Each new key has one candidate. Source-observed variants are research notes, not automatically offered dropdown alternatives. A developer may provide ordered alternatives or a different preference with createEngine({ entries }) as described in the API reference.

This is a coverage change for selected exact keys, not an English pronunciation model or a general spelling correction. More mappings can also create false automatic conversions in mixed Nepali/English fields. The seed benchmark and Node tests are software regression checks against project-authored expectations. They are not independent human language labels or evidence of improved typing accuracy. A future evaluation should use a frozen, separately reviewed set with denominators for preferred output, acceptable candidates, and false conversions across ordinary Roman Nepali, loanword, and mixed-text slices.

Source groups and access scopes

Sources were checked on 2026-09-28. We record short observations and links, not bulk source content. The old L27 links retain their own 2026-09-27 access dates and scopes; especially do not upgrade a search snippet into a directly read PDF. The source groups below establish attested usage in context. Even a direct bilingual parenthesis is not an independent review of every Roman key or sense.

E1

Office of the Company Registrar — Registration uses कम्पनी in directly opened Unicode HTML (full_page_unicode_html). The Law Commission legal dictionary PDF indexes the explicit “कम्पनी (Company)” entry (search_result_indexed_pdf_excerpt; its large PDF was not fully read here). These support official business/legal usage, not a ranked everyday frequency claim.

E2

National Statistics Office — Nepal Living Standards Survey category metadata was opened as Unicode HTML (full_page_unicode_html). Its category labels contain the cited forms, sometimes within longer or truncated labels. The numeric “Cases” beside categories describe survey-file records and are not word counts or a usage-frequency ranking.

E3

Nepal Police — equipment seizure report was opened as Unicode HTML (full_page_unicode_html). One inventory line contains राउटर, ब्याट्री, इन्भर्टर and कनेक्टर. It documents a telecom equipment context, not a normative spelling standard.

E4

Nepal Police — miscellaneous news archive was opened as Unicode HTML (full_page_unicode_html). The archive contains the cited forms in individual news entries. It is a changing listing rather than a frozen dataset; source usage and sense must be checked in the article context before admitting more keys.

E5

National Statistics Office — National Industrial Classification brief was opened as a PDF with extracted Nepali text (fetched_pdf_extracted_text). Its economic-activity examples use कुरियर सेवा, डिजिटल प्लेटफर्म and वेबसाइट. It is not a dictionary or frequency survey of borrowed words.

E6

Film Development Board — drone camera workshop notice has an indexed HTML search excerpt (search_result_indexed_html_excerpt) whose title directly pairs ड्रोन with “Drone” in a photography context. The full page was not retrievable in this check; the Nepal Police archive separately displays ड्रोन in directly opened Unicode HTML.

E7

Nepal Rastra Bank — electronic payment FAQ was opened as Unicode HTML (full_page_unicode_html). It attests वालेट, कार्ड, प्रिपेड, क्रेडिट and एजेन्ट in payment contexts; it does not settle unrelated English senses.

E8

Armed Police Force Women's Association — activities was opened as Unicode HTML (full_page_unicode_html). Its activity descriptions attest जुस, बिस्कुट and चकलेट, and a separate entry uses विस्कुट. This is usage evidence for food terms, not a spelling ranking.

E9

Rastriya Jeewan Beema Company Limited — vendor listing was opened as Unicode HTML (full_page_unicode_html). The travel-service list directly uses बस टिकट. This corroborates the borrowed transport words in a procurement context, not their rank among all possible Nepali equivalents.

E10

Inland Revenue Department — office layout was opened as Unicode HTML (full_page_unicode_html). The page title uses अफिस लेआउट. This supports the spelling in an office context, while कार्यालय also remains an ordinary Nepali alternative by meaning.

Reproduce the code observations

From the repository root after this expansion, run:

node --input-type=module - <<'JS'
import { convertWord, convertText } from './src/index.js';
for (const key of ['company', 'Company', 'COMPANY', 'companyma', 'ambulance', 'router', 'website', 'speaker', 'train', 'fax', 'wallet', 'share']) {
  console.log(key, convertWord(key));
}
console.log(convertText('company company.com company@example.com'));
JS
npm test
npm run benchmark -- --check

Use the historical audit reproduction at its pinned starting commit to inspect the before output. Do not run that old command on a later engine and present it as the 2026-09-28 baseline.