Boston College

- The web interface was developed on the FigureEight platform, which provides a standardized way to specify work interfaces in an HTML-like format.
- Researchers chose to use only standard interface elements which are typical of crowdsourcing tasks on FigureEight, to elicit normal interactions of workers with the web interface.
- Figure 1 illustrates the five different task types chosen for this experiment, and how they appear in a standard web interface vs. a conversational interface.

Chatterbox: Standard Web Interface

- Researchers designed and implemented a conversational interface using the Telegram messaging platform, comprised of two main modules: Conversational Management Module and Input Management Module.
- Microtask crowdsourcing user interfaces are typically designed to be minimalistic and easy to use, to enable fast and effective work execution.
- At the beginning of a task execution stage, a chatbot that drives the conversation prompts the worker with messages containing task instructions, including an explanation of the task at hand, and examples of how input could be provided.
- Once no more annotations are pending in the task, the chatbot prompts the next question to the worker and waits for the worker’s response.
- The answer provided then is validated, with positive feedback if acceptable, or a re-submission sequence if the answer is not valid.
- When no more annotations are pending, workers are shown their answers for review and can re-process a previously submitted answer.
- Figure 1 illustrates the five different task types chosen for this experiment, and how they appear in a standard web interface vs. a conversational interface.

Chatterbox: Conversational Interface

Mavridis, P., Huang, O., Qiu, S.. Gadiraju, U., & Bozzon, A. (2019) Chatterbox: conversational interfaces for microtask crowdsourcing. Association for Computing Machinery, 243–251. DOI:https://doi-org.ezproxy.lib.uconn.edu/10.1145/3320435.3320439

Chatterbox: conversational interfaces for microtask crowdsourcing

Chatterbox: Figure 1

- This module is responsible for aligning the status of the task execution with the status of a conversation, and for supporting navigation within the conversation.
- Figure 2 illustrates the 5 processing states of the conversation management module, and Figure 3 illustrates a brief example of the conversational flow in the chat interface.

Chatterbox: Conversational Management Model

This module is responsible for rendering the content associated to a task, and the UI elements required to allow and control user input.

Learn Before

Related