CIMUC

KWIC

Context

Frequency

Chart

Collocate

Cluster

N-Gram

Keyword

PoS

Utterance Detail

Wordcloud

About CIMUC

Total hits: — click a row to open its full multimodal record.

#	File / Time	Left Context	Hit	Right Context	Pragmatic Function	Status

Full-sentence view of the same hits as KWIC (respects active filters & search) — click a row to open its full multimodal record.

#	File / Time	Speaker (Region)	Sentence	Register

Occurrence counts per word, broken down by region and verification status.

Relative frequency per word, stacked by region. Illustrative counts from this sample dataset only.

Words co-occurring with a node word within its KWIC context window. Node word:

Collocate	Frequency	Left	Right

Common multi-word sequences around a node word. Node word:

Frequent word sequences across the whole sample dataset (not tied to a single word).

N-Gram	Frequency

Keyness of CIMUC words against the mocked baseline — higher ratio means more distinctively colloquial. Reference rates are illustrative, not measured.

Word	CIMUC count	CIMUC per-million	Reference per-million	Keyness ratio

Part-of-speech classification for each annotated word in this sample dataset.

Word	Tag	Description	Frequency

Frequency of annotated words in this sample dataset. Click a word to filter the KWIC view.

Colloquial Indonesian Multimodal Utterance Corpus

What is CIMUC? / Apa itu CIMUC?

CIMUC (Colloquial Indonesian Multimodal Utterance Corpus) is a free, public collection of real Indonesian conversations. It keeps the sound, the shape of the voice (from Praat), the exact words, the English translation, and the real meaning of each sentence. Native Indonesian linguists check every meaning by hand. CIMUC is not only about small words like "sih" or "dong" — it documents everything inside a colloquial Indonesian sentence: the grammar, the vocabulary, the sound, and the meaning.

CIMUC (Colloquial Indonesian Multimodal Utterance Corpus) adalah kumpulan data gratis dan terbuka berisi percakapan asli dalam bahasa Indonesia sehari-hari. Corpus ini menyimpan suara, bentuk nada bicara (dari Praat), kata-kata asli, terjemahan bahasa Inggris, dan makna sebenarnya dari setiap kalimat. Ahli bahasa Indonesia asli memeriksa setiap makna secara manual. CIMUC bukan hanya tentang kata kecil seperti "sih" atau "dong" — corpus ini mendokumentasikan semua hal dalam kalimat bahasa Indonesia sehari-hari: tata bahasa, kosakata, bunyi, dan makna.

What does CIMUC document? / Apa yang didokumentasikan CIMUC?

Audio recording of real, spontaneous speech
Pitch and intonation pattern (Praat-style pitch track)
Full, accurate transcript of the sentence
English translation
Pragmatic meaning — what the speaker really means, checked by native linguists
Speaker information — region, age, gender
Grammar and vocabulary used in real casual speech

Rekaman suara dari percakapan asli dan spontan
Pola nada dan intonasi (gaya pitch track Praat)
Transkrip lengkap dan akurat dari kalimat
Terjemahan bahasa Inggris
Makna pragmatis — maksud sebenarnya penutur, diperiksa oleh ahli bahasa asli
Informasi penutur — daerah asal, usia, jenis kelamin
Tata bahasa dan kosakata dalam percakapan santai sehari-hari

Who is it for? / Untuk siapa?

Researchers who study Indonesian linguistics, pragmatics, and phonetics. Teachers and learners who want to understand how Indonesian people really talk — not just textbook Indonesian.

Peneliti yang mempelajari linguistik, pragmatik, dan fonetik bahasa Indonesia. Guru dan pelajar yang ingin memahami cara orang Indonesia benar-benar berbicara — bukan hanya bahasa Indonesia versi buku pelajaran.

Founder & Team / Pendiri & Tim

Hening Dian Paramita

Founder & Developer

travelinguist.net linguist.co.id (Institution)

About this prototype build

This is a static front-end mockup with a small hand-authored sample of utterances, currently featuring frequent colloquial particles (sih, dong, kok, deh, nih, kan, toh) because they are short and easy to demonstrate — the corpus format itself covers full sentences, not only particles. Audio playback is a procedurally generated placeholder tone shaped to each utterance's pitch contour — real recordings are withheld pending speaker-consent clearance for public release. Pitch tracks are illustrative approximations of Praat output rather than measured acoustic data, except for entries whose detail view is labeled "PITCH TRACK (Praat-measured)" — those carry a real, timestamped pitch contour exported from Praat.

Data model (per utterance)

id, source file & timestamp, speaker metadata (gender, age, region), left/hit/right KWIC context, free translation, pragmatic-function annotation, verifying linguist, and verification status (Verified / Pending review). Each recording is a two-speaker conversation, so every target utterance also carries the surrounding turn(s) from its interlocutor — the Utterance Detail view shows the full exchange, not an isolated line. Two more fields track authenticity per entry rather than assuming it: contentSource ("fabricated" demo text vs. a real "fieldwork" transcript) and f0Source ("placeholder" illustrative contour vs. "measured" real Praat data) — the Utterance Detail view's warning banner and pitch-track label read directly from these.

Corpus analysis tabs

Context — full-sentence KWIC. Frequency / Chart — occurrence counts by region. Collocate — words co-occurring with a chosen word. Cluster — common word sequences around a word. N-Gram — frequent bigrams/trigrams corpus-wide. Keyword — keyness of each word against a mocked formal-register baseline. PoS — part-of-speech classification. All are computed live from the 12-utterance sample, so counts are small by design — the point is to demonstrate the interaction pattern, not to show statistically meaningful results.