3nesdeniz/english-daily-dialogues-10k

English Daily Dialogues 10K

A general-purpose, open dataset of 10,000 synthetic multi-turn English conversations spanning ten everyday-life domains. Built as a clean NLP resource for dialogue modeling, response generation, intent understanding, and conversational evaluation. This is a general language resource — not a safety or security benchmark.

Curated by Enes Deniz (ORCID 0009-0006-9491-3565), Co-Founder at AltaySec. It is the English companion to the Turkish Daily Dialogues 5K resource.

Why this dataset

Most open dialogue corpora are either scraped (noisy, licensing-murky, PII-laden) or narrowly task-oriented. This set is deterministically generated from a large bank of hand-designed, human-reviewed scenario templates with wide slot-based variation, so it is:

• —Clean & reproducible — a fixed seed reproduces the exact corpus; no scraping, no PII.
• —Domain-balanced — ten daily-life domains, each with many distinct scenarios.
• —Split-isolated — every scenario template lives in exactly one split; no leakage between train / validation / test (verified: 0 shared transcripts).
• —Deduplicated — every transcript is unique.

Dataset structure

Each row is one complete conversation between two speakers (A and B).

Splits

Average turns per conversation: 7.94.

Domains

educationandlearning, familyandhome, foodandcooking, healthandwellness, hobbiesandleisure, shoppinganderrands, smalltalkandsocial, technologyhelp, travelandcommute, workand_office

Example

A: Can we talk about chores? I feel like I've been doing the bathroom every single Tuesday lately.
B: That's fair, honestly. I've been slacking on that one.
A: I'm not trying to make it a big thing, just want it more even.
B: How about I take the kitchen and bathroom, you keep the laundry?
A: That actually works for me. Let's put it on the whiteboard so we don't forget.
B: Deal. And if either of us drops the ball, coffee's on them next week.
A: Now that's motivation.

Intended uses

• —Fine-tuning / evaluating dialogue and response-generation models
• —Intent and topic classification, dialogue-act tagging
• —Data augmentation for conversational assistants
• —Teaching and prototyping where clean, license-clear dialogue is needed

Generation methodology

Human-designed scenario templates (12–15 per domain) were authored across ten domains, each parameterized with slot variables (names, places, times, foods, devices, etc.). A deterministic generator expands templates over their slot space with scattered mixed-radix sampling for even coverage, deduplicates transcripts, and isolates scenario templates across splits. The pipeline is fully reproducible.

Limitations & ethical considerations

• —Synthetic: dialogues are generated, not collected from real people. They approximate natural conversation but do not carry real-world distributional nuance.
• —Template-derived variety: while widely varied via slots, conversations descend from a finite scenario bank.
• —English, contemporary, general-audience: no dialectal or historical coverage guarantees.
• —No PII: all names/places are fictional slot fillers.

License

Released under CC-BY-4.0. You may share and adapt with attribution.

Citation

@dataset{deniz_2026_english_daily_dialogues_10k,
  author    = {Deniz, Enes},
  title     = {English Daily Dialogues 10K},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/3nesdeniz/english-daily-dialogues-10k},
  note      = {ORCID: 0009-0006-9491-3565}
}