With technology dominating all sectors in the world, AI has role in fundamental aspects of daily life. Even the simplest understandings such as how much time it will take to arrive at a destination or the next song in the queue can be known with AI and Machine Learning algorithms. These algorithms become man’s friend with their ability in data annotation. Data annotation is the process by which, a particular data is labelled so that the machine can understand the material that is inputted and come up with an accurate output.  

 

 Example for data annotation 

Source: Deigo Calvo 

The information is not new that AI runs on data. With its involvement in almost all sectors, enterprises need to have accurate data annotation techniques when they largely depend on the models for the results produced. Imagine an ML model with poor data set. There is no wonder that the prediction capacity of the particular model is not dependable. 

Types of data annotation: 

Data annotation techniques are of different kinds, depending on the ML applications. There are semantic annotations, which trains the machines to understand a particular text. Let it be long paragraphs or simple sentences, a properly labelled text can train models to provide useful solutions. For instance, an input statement “I want to chat with X”, can be understood by the ML model and they will classify them as request and approval. Concepts such as the name of a company, or a person are also labelled within a text so that the model can categorize it in the future.  

Similar are image annotations where images are labelled. Image annotations often involve imaginary boxes drawn on an image where the model recognizes the annotated areas as distinct objects. There are video annotations where bounding boxes form on videos on a frame-by-frame basis. Video annotation tools are also present to acknowledge movement. The data annotation techniques used by companies will be different from one another concerning their needs. 

 Source: appen 

In the above image, some enterprises might use the data annotation technique to understand the number of pedestrians while some might use it to understand the car models parked. 

Data annotation in India:

Amid COVID-19 when AI became a service essential to reopen the economy enterprises and government realized the significance of data assets and integrated systems. During this time there was an unexpected need to solve the AI-led decision-making process. Pandemic demanded the MSPs to alter their model of operation to ensure continuity in business. According to a Data Annotation Report published by NASSCOM in February 2021, the data annotation market serviced by India in FY20 valued at ~USD 250 Mn – with ~60% of the revenues derived from US clients. The market revenues according to the report are derived from multiple business models with managed services contributing 65%-75% to the overall market. 

Some of the major challenges that persist in the Indian market are issues concerning data privacy, cultural context and demand for non-English language in the process of labelling. 

Scope of data annotation in India: 

With the fast growth of AI in fields such as healthcare, education, automobiles, telecom, e-commerce etc., the development of data annotation is also inevitable. In the next few years, the country can witness an exponential increase in demand among establishments across the country despite their sector. The NASSCOM report states that the data annotation market can exceed up to USD 7 Bn by 2030. 

 

Want to publish your content?

Publish an article and share your insights to the world.

ALSO EXPLORE

DISCLAIMER

The information provided on this page has been procured through secondary sources. In case you would like to suggest any update, please write to us at support.ai@mail.nasscom.in