Improve Asset Tagging with Image Recognition AI

• Background

  A leading provider of workplace products and business services to small, mid-sized, and large businesses in Europe provides solutions to over 9 million customers across 40 European countries each of which has its own E-commerce website.

• Information

  Use Case

  Metadata Generation

  Industry

  Retail

  Client

  Business Retailer

Challenge

To stay ahead of competitors, the company relies on SEO to optimize product pages and surface relevant products to potential customers. However, since the retailer spans across a geography that speaks a dozen different languages, attaching the right keyword metadata to products to improve SEO becomes 12 times as challenging and expensive. The retailer needed a fast, accurate, and inexpensive solution to adapt its E-commerce business for multiple languages and cultures and improve SEO for its enormous product catalog in every target language.

Solution

After initial testing, the company’s marketing team, lead by their SEO Specialist, selected Clarifai based on its impressive accuracy and easy platform setup. All it took to start the SEO project was an API key and an unofficial Clarifai Ruby client. Once the company had URLs for all six hundred of the images it wanted to initially tag, the SEO specialist wrote a script to loop through the images and call Clarifai’s API for each image and language.

Using Clarifai’s core visual recognition model, the retailer was able to automatically add relevant keywords to their product images. These keywords were then added to the product images’ ALT tags in twelve different languages in order to improve SEO and increase traffic to their online e-commerce product pages.

Results

With Clarifai’s image recognition technology, the retailer was able to complete the work of an outside translation agency in a single afternoon at a fraction of the time and cost and with the same level of accuracy.

Clarifai’s API translated over 11,000 English tags into other languages. Each concept in the computer vision model was trained and mapped to their respective words in different languages. Thus, tricky lingual concepts like “crane” (the bird) and “crane” (the machinery) or “fall” (the verb) and “fall” (the season) are always correct in Clarifai’s multi-language API.

• Clarifai’s results were just as accurate as results we’d seen from humans at agencies. Once I saw how easy Clarifai was to set up and how accurate the AI was, going forward was a no-brainer.”

  SEO Specialist