Wipi, a social dating application operated by Enrise, is transforming user experiences by integrating artificial intelligence (AI) throughout its services. The platform utilizes a recommendation AI that learns from actual user behavior data, a generative AI matchmaker named 'Mela,' and a safety AI that filters out inappropriate users, thereby enhancing its data-driven AI capabilities.
Wipi's strength lies not just in its generative AI model but in the real user data accumulated during service operations. The app has gathered data from 8.4 million registered users, 20 million matches, and over 100 million friend requests. This data, which includes information on which profiles users viewed, to whom they sent friend requests, and whether those interactions led to actual relationships, forms the foundation for advancing its AI.
Last August, Wipi introduced an AI prediction model to enhance personalized recommendations. Previously, matches were suggested based on predetermined criteria, but now the AI learns user patterns to predict potential matches that users are likely to be interested in.
Successful matches in dating services require mutual interest. Wipi considers this by analyzing not only a user's preferences but also the likelihood of a potential match's interest, implementing a two-way recommendation structure.
Crucially, the behavioral data accumulated from actual service use enhances recommendation accuracy. Various outcomes, such as users not responding to profiles, sending friend requests, or achieving successful matches, are utilized to refine the model. This creates a virtuous cycle of 'recommendation → user response → data accumulation → recommendation enhancement.'
Generative AI is also employed to improve the recommendation experience. Wipi's AI matchmaker, 'Mela,' goes beyond merely displaying potential matches; it engages users in conversation to understand their ideal partner preferences. It then recommends suitable matches and explains the reasoning behind those suggestions.
User feedback has been positive. Among applicants for the matchmaker feature, 67.3% are women in their 20s, which is about 21.5 percentage points higher than the proportion of women in that age group among all active female users.
Women who applied for the matchmaker on the day of registration showed a retention rate 1.4 times higher than those who did not apply. This indicates that users are responding positively to an AI-driven exploration experience that provides recommendations along with reasons tailored to their preferences, rather than simply browsing numerous profiles.
Wipi also leverages AI for service safety management. During the registration process, AI analyzes profile photos, self-introductions, keywords in chat rooms, and abnormal user behavior patterns to proactively detect potential violations of operational policies.
Key detection targets include users registering others' photos or using images that make facial identification difficult, entering phrases that violate operational policies, sharing contact information, and engaging in behaviors that promote inappropriate encounters. The AI detects anomalies first, allowing operators to conduct final reviews, thereby improving response speed and consistency.
Wipi plans to further enhance recommendation accuracy and expand the application of AI across its services through the advancement of its AI systems.
A Wipi representative stated, "We will continuously learn from users' actual behavior and preference data to improve recommendation accuracy and reflect individual preferences, such as appearance preferences, in greater detail." They added, "We plan to expand the use of AI across recommendations, exploration, and overall operations by combining machine learning with generative AI."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.