“We use different neural net models to determine the intent and whether to show a suggestion.”
Facebook has introduced new AI processes into the back-end of Facebook Marketplace and its Messenger application.
Marketplace was first introduced in 2016, originally it was used by Facebook users to sell and buy items in their local communities. Now one in three of the social media platform’s US base uses the market.
The market already uses AI when someone is looking for an item to buy. A content retrieval system and product index is powered by AI-based computer vision and natural language processing platforms.
The computer vision is employed to match items of similar design into categories as well as suggesting the product to a user who has bought a similar item.
Lu Zhyeng, Rui Kui and David Kim wrote in a Facebook blog that to make this new function possible “we built a multimodal ranking system that incorporates Lumos, our image understanding platform, and DeepText, our text understanding engine.”
“As a result, the index for each product includes layers for both the text and the images. To model the word sequence, we map each word from the title and description to an embedding and then feed them into the convolutional neural networks.”
Facebook has now included into the market an AI feature that auto-suggests the best price for your product. Say you take a photo of a mug and you are posting it on to the market, the AI will compare your item with similar products in the database and will offer a price and category suggestion.
The Facebook blog notes that: “Before the auto-suggest feature for categories was enabled, friction was high: 7 to 9 percent of sellers abandoned the process without completing the listing, and the seller-selected categories were sometimes less effective than those recommended by our model.”
The company has also introduced new AI text suggestions into its messenger application. If you are contacted by a buyer through the app in relation to an item in the market place it will auto-suggest quick responses based on their message, e.g. “Is that mug still for sale?” will get an auto-suggestion of No, Yes or I think so.
From the buyer’s perspective if the answer is no, then messenger will prompt the buyer with a “Find more like this on marketplace” suggestion; facilities online retailers like eBay already use widely.
Lu Zhyeng, Rui Kui and David Kim commented that: “All these suggestions are automatically generated by an AI system that was trained via machine learning to recognize intent. We use different neural net models to determine the intent and whether to show a suggestion.”
“We also use entity extractors like Duckling to identify things like dates, times, and locations. For some suggestions, we use additional neural net models to rank different suggestions to ensure the most relevant options are surfaced (like translation, stickers, or replies). M also learns and becomes more personalized over time, so if you interact with one suggestion more regularity, M might surface it more.”