English Data Collection & Transcription

---

## Project Overview (For Vendors)

Suppliers are required to collect and deliver a dataset of **English-language meeting audio recordings** that meet the specifications below.

---

## Volume & Timeline

* **Total audio volume:** 2,000 hours
* **Delivery timeline:** 10 weeks
* **Participants per meeting:** 3 to 6
* **Maximum contribution per participant:** 5 hours

---

## Speaker Requirements

* **50% native English speakers**
* **50% non-native English speakers** (English as a second language)
* All recordings must be **in English**
* No restriction on country or locale

---

## Meeting & Recording Setup

* All participants must be **physically present in the same room**
* Each meeting must include:

* **One far-field recording** via Zoom using:

* Laptop or mobile phone (50% of meetings), or
* In-room microphone array (50% of meetings)
* **Individual close-talk/lapel microphone recordings** for each participant
* Total audio signals per meeting: **4–7 tracks**
* All audio tracks must be **time-synchronized**

---

## Recording Platform

* **Zoom** must be used to capture the far-field audio
* Close-talk/lapel microphones must be recorded **outside of Zoom**

---

## Conversation Types

Meetings may be guided using prompts and must include a mix of:

* Business – General
* Business – Technical
* Non-business / casual conversation

---

## Environmental & Acoustic Diversity

* Recordings must reflect **real-world meeting conditions**
* Variety in:

* Room size and layout
* Participant positioning
* Building materials and acoustics
* Audio must be **clear and intelligible**, with natural background noise and no excessive interference

---

## Metadata Requirements

Each meeting must be delivered with complete metadata, including:

* Number of participants
* Participant details:

* Gender
* Age range (e.g., 18–35, 35–50)
* Native language locale
* Device types used
* Room setup and environmental details
* Recording duration
* Country and locale

---

## Delivery Expectations

* Audio must be clean, properly labeled, and synchronized
* Metadata must be accurate and complete for every recording
* All requirements must be met to count toward the final accepted hours