AI-Powered Search Term Category Prediction Enhances Procurement Platform Accuracy

When users search for products on the procurement platform, they may either browse categories or type keywords into the search box. Previously, the system relied solely on keyword matching in product titles, leading to irrelevant results such as "computer headphones" when searching for "computer".

To address this, the technical team introduced an AI algorithm called Search Term Category Prediction . This feature enables the system to interpret the user's intent, consider multiple product attributes (category, brand, model, key features), and predict the most appropriate product category for the entered term.

The AI serves three main purposes: (1) helping the system understand the meaning of the search term so results are accurate; (2) filtering millions of items to those the user is likely to prefer; and (3) ranking the most desirable products at the top of the results page.

For example, when a user types "computer," the AI predicts a 47% probability for the "laptop" category, 31% for "handheld computer," and 12% for "desktop computer," while assigning near‑zero probability to unrelated categories like "computer desk" or "computer chair." Consequently, only items from the predicted categories are matched and weighted, greatly improving relevance.

After continuous tuning, the platform’s search accuracy has risen to nearly 97%, achieving the transition from "searchable" to "accurate" results. In addition to AI‑driven relevance, the platform also provides guidance for prohibited items and directs users to dedicated procurement zones.

Looking ahead, the team plans to explore further AI and big‑data applications such as intelligent product recommendation and user‑preference prediction to make procurement even simpler, more efficient, and enjoyable.