The LinkedIn AI is Capable of Adding Description for The Photo Text

As technology is developing, tech companies are now moving towards automation and adding AI for better response, assisting work and making everything rapid. Developers and technical scientists at LinkedIn have been long trying to figure out if they have improved AI technology and incorporate it, to enhance the performance of the platform. In a recent attempt to improve the efficiency of the platform, the technical engineers at the platform have introduced a tool to add suggested alternative text in the description of the image that has been uploaded on LinkedIn. Just like Facebook, the tool will automatically add an alternative text under the image and will help the user in saving time. This tool has been designed with the help of Microsoft’s cognitive services platform in combination with the LinkedIn derived data set.

This newly designed tool has not yet been incorporated which means that currently, the user can only add alternative text in description manually. This means that on adding the image users are required to add something in the description, however, most of the users do not add a description which is the reason LinkedIn has introduced this feature. Since the idea of this tool is under discussion, more people are now trying to evaluate the kind of description that this tool is adding. These captions are more subjective which means that they might describe the picture correctly but this is not adding any value which is the reason it is suggested that this technology must be improved by adding objective knowledge as well as time-based information that will further help in adding a better description.

To recognize how a normal reader perceives these descriptions and what must be suggested as an alternative a team of researchers tapped Cognitive Services’ Analyze API to design a feature for suggesting alternative text descriptions for the picture that has been ranked by a confidence score. As a second suggestion, these researchers invited human evaluators to evaluate the descriptions by integrating the scores. These scores provided information about the text, description as well as tags. To improve these descriptions, the development team later solved the discrepancies by vaulting the feedback of both descriptions.

Photo: OhmZ via Getty Images

Read next: LinkedIn Prevented Millions of Fake Accounts from Being Created in 2019
Previous Post Next Post