Please describe your proposed solution.
Overview:
TrustLevel has already created a POC that uses our NLP model to analyze content, map it across a knowledge graph and use our analytics and algorithms to identify valuable interactions within a system, which can then be converted into reputation scores. This proposal is designed to create the MVP (ready for Testnet) for the blockchain integration to manage and store reputation scores of the TrustLevel system.
Extended Problem Statement:
In today's interconnected world, it is nearly impossible for individuals to make judgments about the trustworthiness of online sources in a reasonable amount of time. The massive improvement in AI in recent months and the resulting rapid growth in the use of AI for content creation is further intensifying this issue. In the following are presented some of today’s main challenges:
- Information Overload: The internet provides an overwhelming amount of information, making it difficult to differentiate between reliable and unreliable sources.
- Ease of Publication: Anyone can create and publish content online without rigorous fact-checking or editorial oversight.
- Confirmation Bias: Online platforms often use algorithms that personalize content based on users' preferences and browsing history.
- Malicious Actors: There are individuals or groups with the intent to deceive or manipulate others by disseminating false or misleading information.
- Lack of Accountability: The anonymity and pseudonymity allowed by the internet can make it difficult to hold individuals or organizations accountable for the information they publish.
- Limited Media Literacy: Many individuals lack sufficient media literacy skills to critically evaluate online sources and distinguish between credible and unreliable information..
Extended Solution Rationale:
The challenge is to create a system that not only captures various reputation attributes, but also ensures transparency, security and immutability, all of which are compatible with the principles of blockchain technology:
- Trust and credibility are essential elements for collaboration, transactions and engagement, and a decentralized and interoperable reputation protocol can solve these problems.
- Our approach is to develop a Universal Reputation Protocol that, by clearly defining reputation attributes such as reliability, expertise and ethical behavior, can provide a comprehensive and nuanced representation of a person's or company's reputation.
- Tokenization on the Cardano blockchain was chosen to ensure a decentralized and immutable ledger of reputation scores. The use of smart contracts provides an additional layer of automation and predefined rules, making the reputation system self-regulating and efficient.
Target Group:
Trustlevel engages individuals seeking trustworthy online content, content creators, and publishers. By addressing reputation challenges, it benefits users striving for credible information, while content creators and publishers gain a transparent and decentralized reputation system.
Reputation Scoring Process (POC):
To demonstrate impact, the existing POC employs a three-step reputation scoring process. Content Quality Score (CQS), Content Categorization, and Reputation Calculation are showcased on Github. The POC is currently being further developed for the SingularityNET ecosystem. Feel free to reach out on our website for a demo.
As of December 6th 2023 the POC includes the following three step to calculate a reputation score:
<u>1. Content Quality Score (CQS)</u>
We use our NLP-Model to create a CQS for each content.
- Currently a mix of Sentiment, Objectivity, Clickbait and Metadata analysis
- The demonstration can be found on Github: <https://github.com/TrustLevel/Content-Quality-Score>
- Public launch on SingularityNET in December 2023 (as of writing this proposal, it’s already on the Testnet) as part of the following funded proposal: <https://deepfunding.ai/proposal/trustlevel-integration-into-snet-ai-marketplace/>
<u>2. Content Categorization </u>
We use different AI Classifiers to tag and categorize content and a Knowledge Graph to visualize and represent the data.
- We differentiate between 20 different news categories (IPTC Classifier) , 7 different content tags (news report, opinion, tweet, etc.) and a ‘Named Entity Classifier' to extract mentioned locations as part of this proposal: <https://deepfunding.ai/proposal/unique-knowledge-graph-for-source-and-content-reliability/>
- The Knowledge Graph demonstration and sample dataset can be found on Github: <https://github.com/TrustLevel/Knowledge-Graph-for-Source-and-Content-Reliability>
<u>3. Reputation Calculation </u>
In the POC this step is done manually.
- We use the CQS to calculate domain specific reputation scores (depending on category, tags, named entities) for content creators and publishers
- A demonstration can be found here as well: <https://github.com/TrustLevel/Knowledge-Graph-for-Source-and-Content-Reliability>
Smart Contract Design (MVP):
The MVP of the blockchain integration will utilize the Cardano Blockchain as the foundational layer for recording reputation scores. The smart contracts define rules and logic for the reputation protocol to facilitate the scoring mechanism and attribute updates.
- Storing reputation scores: the smart contracts store the reputation values associated with users or categories.
- Updatable reputation scores: Smart contracts allow reputation scores to be updated.
- Integration with TrustLevel backend: The connection of the smart contracts and the TrustLevel backend enables reputation scores to be retrieved and displayed to users, and feedback or other relevant data to be submitted that may affect reputation scores.
- Event logging and transparency: The smart contracts log relevant events, such as reputation score updates or user interactions. This ensures transparency and traceability of reputation-related activities on the blockchain.