TICTAG'S CASE STUDY
Audio Transcription Annotation
Applicable Industries and teams: Conversational AI , Audio AI, Digital AI Assistance, Speech Recognition, Law Firms
Problem at hand
When it comes to annotating irregular shapes such as outlining the anatomy of a body part, detecting fruits and vegetables in the garden, detecting cars, the bounding box falls short compares to polygonal bounding because the latter can capture more lines and angles. But there are certain problems when it comes to polygonal bounding:
-
It's harder to reach maximum accuracy.
-
It's much more time consuming
-
Very little accessibility: Not many tools provide polygonal bounding with good controls.
Tictag's Solution
​Tictag leverages its crowdsourcing abilities and its gamified mobile platform to make data annotation accessible to more people while also maintaining high accuracy data even for complex polygonal bounding use cases.
-
Record speed of transcribing and cleansing of the audio
-
Crowdsourcing transcription abilities from various demographics in order to overcome the Limited vocabulary and accent bias. (Tictag's access to the uncommon SEA crowd allows clients to obtain data with much higher accuracy for projects that require such crowds)
-
Use of human intelligence to recognize and transcribe valuable data in low-quality audio files or ones with background noise.
-
Use of human intelligence to recognize multiple inputs (Example. When 2 people are speaking in one file, Tictag's solution allows the crowd to transcribe dialogues for the 2 people separately.)
-
Quality Control to sieve out contradictory and inaccurate data.
Results
Tictag leverages its crowdsourcing abilities and on-the-go mobile platform to get you highly accurate transcribed data to for your Audio AI Models.
Don't compromise on what your AI can do. Blow your audiences away with an Audio AI that can be truly intelligent, using Tictag's best-in-class data preparation solution.
Working on an Audio AI Model?
Speak with our experts and narrow down your use case/ project requirements. Obtain the accurate data that your model needs
​