Relevance Wall
The Relevance Wall in Kuroshio is designed to streamline and enhance counterparty matching by filtering incoming connection requests based on relevance to user-defined criteria. This system leverages the ChatGPT API and its Large Language Model (LLM) to dynamically analyze tags and preferences that clients specify, aligning these with incoming user requests and facilitating a more precise, targeted search experience via an intelligent chatbot interface.
How the Relevance Wall Operates
Profile Tagging and Initial Setup:
When users create or update profiles, they define key tags representing their expertise, interests, and requirements. Examples include tags like #DeFi, #GameFi, #SeriesA, #Developer for talents, or #Investor, #FinTech, #Gaming for VCs and KOLs.
These tags serve as the foundational elements of the Relevance Wall, providing a basis for assessing connection relevance by categorizing each user’s focus areas, preferences, and engagement requirements.
Request Processing with ChatGPT API and LLMs:
When a user initiates a connection request, Kuroshio uses the ChatGPT API to analyze the profile tags and any additional context in the request. This analysis is handled by a custom-trained LLM tailored to recognize and prioritize tags that closely match the target profile’s Relevance Wall settings.
The LLM processes the tags and contextual cues (e.g., project descriptions, interest in specific expertise areas) in the request, assessing compatibility based on both explicit tags and implicit contextual relevance.
Tag Matching and Demand Assessment:
The LLM-driven model evaluates primary tags (explicitly mentioned by users) and secondary tags (inferred based on contextual language in the profile and request).
Using ChatGPT’s capabilities, the system performs semantic matching between the initiator’s and target’s tags. This goes beyond direct tag matching, as the LLM can infer relevance by understanding phrases and keywords that imply compatibility even if they do not match the tags exactly.
For example, if an investor is tagged with #SeriesA, #DeFi, and #NFT, and an entrepreneur describes a "new DeFi platform seeking seed funding for blockchain interoperability," the model can assess the implicit relevance, even if “interoperability” isn’t a direct match to “DeFi.”
Relevance Scoring and Threshold:
Each connection request is assigned a Relevance Score based on how well it aligns with the target’s profile settings. The score reflects the closeness of the match between the tags, context, and user-defined requirements.
Users can set a Relevance Threshold to filter requests, allowing only those that meet or exceed this threshold to proceed. This enables users to avoid irrelevant or low-priority requests, ensuring time efficiency and enhancing interaction quality.
Chatbot-Driven Search and Interaction:
Kuroshio’s chatbot interacts with users in real-time, guiding them to refine their tags and criteria based on available data and trends from other similar profiles.
The chatbot can suggest adjustments to tags or recommend new tags based on market demand, optimizing the user’s profile for visibility and relevance in future searches.
Continuous Learning and Model Optimization:
Kuroshio’s relevance model benefits from continuous learning through user feedback and behavioral data (e.g., response rates, accepted/declined requests), refining its relevance scoring over time.
By leveraging the ChatGPT API’s LLM, Kuroshio continually improves the Relevance Wall’s accuracy and adaptability, providing users with increasingly optimized matches that reflect the evolving landscape of Web3 and crypto sectors.
Technical Benefits and Unique Value
Adaptive Semantic Matching: ChatGPT’s LLM enables the Relevance Wall to interpret tags and profile descriptions with a high degree of semantic depth, ensuring accurate matching even for nuanced or industry-specific terms.
Dynamic Filtering: With the Relevance Wall’s adaptive scoring and relevance threshold capabilities, Kuroshio maintains high-quality interactions by filtering out low-relevance requests, enhancing user experience.
Personalized Recommendations: The integration of a chatbot model creates a more interactive setup and adjustment experience for users, promoting profile optimization that aligns with real-world trends.
Last updated