This question type is chosen when the answer to the question can be provided in a single line with few words. Suited best for those insights, which summarizes the data qualitatively such as translation in NLP (applies for Indic languages), short transcription in audio, captions for image and video.
Paragraph question types are intended for more in-depth analysis of the input data, where the responses span for multiple sentences. Example projects with these question type can be “Story completion” in NLP, multiple captions/descriptions in image/videos and multi sentence speech to text transcription in audio. There will be no pre defined word limit for either of the question type, unless specifically indicated by the user.
In certain cases, the user would like the human annotator to select one option from the multi choices. When multiple such insights are needed for a single data, the user will simply repeat these question types. Some of the examples of such annotations are emotion analysis in NLP, presence/absence of object in images/videos and question answering in audio.
The question type is similar to checkbox, where the user expects the human annotator to pick from the pre defined list of options except here the annotator is allowed to select more than one option for a single question. Selection of available objects from list in image/video, picking all of the emotional states for a given sentence or paragraph are some of the examples for this question type.
Likert scale allows the annotators rate the intensity of a particular label for the given data. This question type provides exactly that facility for the user, and above that the scale can be customized by the user to show the right qualification of the scale. Some of such examples could be scale between 1-7 the quality of a image or video, scale between most-intense to most-subtle emotion expressed in an audio or sentence.
A dropdown provides the same functionality as that of Multiple choice question type, except that in the case of drop down question type the user has moment to think about the response for the question prior to looking to the choices; there by removing the inherent bias. This question type can also be used to save space, when multiple questions are needed to be shown in a single screen.
A question type that serves as a consolidation of multiple checkbox questions, where the annotators can see all the choices available for each of individual categories for a given data. In this question type, the rows represents the qualitative parameter and the columns represents potential values these parameter can take. An example in video could be “quality”, “relevance” etc as rows, “bad”, “average”, “good” as column. Annotators will be able to select one value for each of the parameter.
Multiple Choice Grid
This question type is very similar in its application to the checkbox grid, but allows the annotators to select more than one value for each row. An example in video annotation could have “Person talking”, “Argumentative” and “speaker 1”, “speaker 2” as column.