Page 19 - THE LIVE STREAM MANUAL ONLINE PRODUCT SALES

must be overcome in order to adapt recommendation algorithms. First, content availability on live-

streaming platforms is dynamic, which limits users to selecting only a subset of items at any given
time; during training and inference, we must exercise caution. This factor must be considered in

order to properly account for such signals, where non-interactions reflect both availability and
implicit preference. Streamers are also fundamentally different from 'items' in traditional settings

in that repeat consumption of specific channels is important, despite the fact that the content itself

is ephemeral.

Video streaming platforms like Twitch and YouTube Live are becoming an increasingly

important part of people's daily lives. Twitch reported 3 million monthly broadcasters as of

February 2020. There are 15 million active users every day. The growing number of concurrent
broadcasts, an expanding audience, and the long tail of niche content all imply the need for systems

specifically designed for such platforms. Content creators broadcast video on live-streaming
platforms on their respective channels in real time. The transmission of real-time content implies

that videos can only be viewed at specific time intervals. This dynamically changing stream
availability means traditional methods face difficulties. Methods of implicit feedback that have

been trained to capture preference signals from positive interactions.

Because users repeatedly consume content produced by the same streamers, a live-streaming

model should account for repeat consumption, which accounts for a large portion of observed
interactions. This is in contrast to typical recommendation domains (for example, movies), (e-

commerce) that is based on one-time user-item interactions. At the same time, it contrasts with
ongoing lines of work consumption because channel content is dynamic and varies with each new

interaction. (J Rappaz, J McAuley & K Aberer 2021).

Live-streamers are seen as an important node connecting the brand and consumers.
Professionalism is about a seller's ability to provide accurate information about the product. Unlike

in a brick-and-mortar shop where a consumer can make a purchase decision based on the sense of

touch and smell, providing accurate information is crucial in a live-streaming shopping
environment that supports consumers in framing their choices. "Buy intention" is the behavioral

possibility of consumers to purchase products or services in marketing activity.

Interaction is the process of information exchange and mutual communication between the e

18