Get real dialogues, the many ways people ask for the same thing and live testing of your bot from everyday users in the countries and languages you choose.
Text and voiceMany languagesReview every dialogue190+
Countries
13,487
Tasks completed
12,000+
Registered workers
~4 Hours
Avg time to results
Launch a dialogue or bot testing task yourself and start receiving real conversations, without long negotiations.
Explain the situation, the roles and what the user wants to achieve. For bot testing, add a link to your bot and a list of things to try.
Scenarios and rolesLink to your bot for testingExample of a good dialogue
Select the countries your contributors should come from and the language they should use. See how real users in each market talk to your assistant.
Any number of countries in one taskNative speakers by countryYou set the total volume
Your task becomes available to workers in the selected countries. They write dialogues, record voice requests or talk to your bot and report what happened.
Text or voiceNatural everyday phrasingNew person in every submission
Read or listen to every submission, approve the ones that follow your scenario and reject the rest. Keep your dataset clean from the start.
Check every submissionApprove or reject each oneClean data for trainingSet up a task for the exact conversations your assistant needs. Contributors follow your scenario and use their own words.
Give contributors a goal, such as "cancel my order", and they write or say how they would ask for it. Teach your assistant to understand real wording, not just one phrase.
Your list of intentsText or voice requestsDozens of variations per intent
Contributors write full conversations where the user clarifies, changes their mind or asks again. Train assistants to keep context across several turns.
Several turns per dialogueClarifications and correctionsScenarios you choose
Workers play a customer with a problem and write or record the conversation with a support agent. Use them to train and test support bots.
Realistic customer problemsFictional personal details onlyText or recorded with a partner
Share a link to your chatbot or voice assistant. Contributors try the tasks you list and report where the bot failed, misunderstood or gave a wrong answer.
Link to your bot in the taskReport of every failureRating of the conversation
Contributors record themselves asking an assistant for help, the way they would at home or in the car. Get real accents, pauses and phrasing.
Natural spoken requestsAccents by countryTranscript on request
Native speakers write requests with everyday slang, mixed languages and regional expressions. Make your assistant sound local, not translated.
Slang and informal wordingMixed languages like HinglishRegional expressions
Ask contributors to name what the user wanted next to every dialogue. You get conversations with labels, ready for training.
Your setup
Choose Text or Audio collection, describe the scenario and roles, and ask workers to name the user's goal at the end.
Text or audio collectionScenario and roles in instructionsIntent label in every submissionYour view
Each submission arrives with the dialogue or recording and the intent label, ready to review and approve.
Dialogue and label in one submissionWorker country on every submissionApprove or reject each oneChoose where your contributors come from and which language they use. Every submission comes from a different person.
Accept workers from all countries or pick specific ones by continent or individually. Hear how users in each market really talk.
Describe who should take part in your instructions, such as language level, age group or experience with chatbots. Reject submissions that do not match.
Add a sample conversation that shows the length, tone and format you expect, so contributors get it right the first time.
Each worker can complete your task only once. You get as many different voices and writing styles as the submissions you order.
You decide what counts as a good dialogue and review every submission before it joins your dataset.
Describe the roles, the goal and how many turns each dialogue should have.
Show good and bad examples so contributors know exactly what you expect.
Read or listen to each submission and accept only dialogues that follow your scenario.
Launch a separate task where other workers check dialogues and labels for you.
Every dialogue comes from real people who chose to take your task and knew what it was for.
Your task description explains what people will write or record and that it will be used to train or test AI.
When two people record a conversation together, both of them agree to take part. Role-plays use fictional names and details only.
Each submission comes from a specific worker in the country you selected, created specifically for your task.
You set the pay per task. Adjust the numbers to see what your project will cost, with the 15% platform fee included.
How long it takes a worker to complete one task.
Minimum pay is $0.10 per task. Need more dialogues later? Extend the same task instead of creating a new one.
Conversational AI data is the dialogues, requests and feedback used to train and test chatbots and voice assistants. It shows how real people ask for things, clarify and react to answers.
Here every task is built around a conversation with an assistant: intents, dialogues and bot testing. For general voice recordings see speech data collection, and for other writing tasks see text data collection.
Yes. Add a link to your bot and a list of things to try in the instructions. Workers talk to it and report failures, wrong answers and their overall rating. Do not ask them to share real personal data with the bot.
Yes. Create a text collection task for written dialogues and an audio collection task for spoken requests, or contact us for a combined setup.
Any language spoken by workers in the countries you select. Choose the countries and state the required language in your instructions.
You set the pay for each task, starting from $0.10, plus a 15% platform fee. Long dialogues and bot testing usually need a higher pay rate. A campaign starts from $0.30.
Ask contributors to use fictional names, addresses and numbers in role-plays and tests. Reject any submission that contains real personal data.
You review every submission yourself and approve or reject it. For larger projects, you can launch a separate task where other workers check dialogues and labels.
Yes. Extend your existing task instead of creating a new one. Workers who already took part still cannot submit again.
If you need a special format, a review flow of your own or help writing scenarios, contact us through the form below and we will set it up with you.
Need many languages, bot testing at scale or help writing scenarios? Send a message and we will set it up with you.