Pinterest is increasingly working on building its eCommerce credentials. The overarching intent is to connect products with user interests, and prompt users to buy. At its core, Pinterest’s recommendation system is based on the content in each Pin. Therefore, for a better understanding of the content, the company relies on categorizing the Pins.
Pinterest categorizes Pins based on available data in the following order - taxonomy level, label, and score. Taxonomy level relates to popular concepts that appear in Pins. Each taxonomy level has around 10 levels of granularity, with 24 top-level concepts. For example, a top-level concept such as Home Decor would have a range of more specific sub-topics.
Label relates to the specific topics and/or subjects identified within the image, which comes primarily from the accompanying Pin text. A visual classifier can also infer the topics/subjects. Finally, score relates to how much of a match Pinterest's systems think the Pin is for each identified label, based on the information provided.
Some of the topics and matches are inferred based on Pinterest's overarching understanding. Pinterest's system matches contextually related elements in a Pin to ensure it measures for relevance when incorporating various factors. Therefore, relevant terms help Pinterest's system match content to relevant user queries.
How does this approach help a Pinterest user? By gaining, a better understanding of Pinterest's matching systems users can improve their marketing approach, and ensure their Pin content is displayed to interested users.
Click here to read the original article published in Social Media Today.
Please give your feedback on this article or share a similar story for publishing by clicking here.
