Below is a full-recap of the AMA, for the gem chasers who couldn’t make it, or for anyone that’s interested. Feel free to have a read!

Q: Tell us about your team

A: Thank you Sadaf, hello everyone, excited to join ENS twitter space,we are a startup team called DMGdata with highly skilled algorithm engineer and full stack developers. DMGdata is a web3 digital marketing and data service provider.

My name is kata, one of the co-founders of the team, mainly responsible for product design and data analytic.

Q: Introduce the product you developed

A: The product we build is called Dbrands, which is an AI-based ENS aggregator and trading platform independently developed by DMGdata.

Q: Issues that are expected to help the user solve via Dbrands

A: The basic function for V1 is search domains, go through hot collections, check your ens assets and customized share with friends. Besides this, we want to provide an easy to use and interesting user experience to the communities by using AI.

Get your web3 domain name and share your web3 identity with the world is our product slogan.

Q: What are the AI related functions you implement

A: We’ve developed some interesting functions, let me introduce them on behalf of the team.

ENS Associative Search - We detect your search input whether it’s a company or not, when you type in a full company name, system will return recommended abbreviation or brand name with .eth suffix.  In the future, we will also extend it into more interesting area, when you type in famous brand such as SONY, system return product list with .eth suffix or you type in a popular game, system return the game characters name list with .eth suffix. We believe this can facilitate the registration for ENS. Meanwhile, to meet the needs of different native language, we support multilingual search.

ENS Assets Appraisal - As we all know, most of ENS domains has no secondary marketplace trading history. This function is going to provide an explainable price evaluation which can help the traders and owners build up communication with some useful reference.

Label Generation for ENS - In this section, we are trying to discover and add more useful labels into ENS domains. Including club tags, language tags, value ranking tags, smart money tags etc.

Algorithm Recommendation - you may like is a very basic function in most of E-commerce platform such as ebay，amazon, shopify and steaming media. We add this function to enrich the choices for the user.

Q: Introduce the implementation of  ENS Associative Search

A: For user input, use a multilingual translation model to translate the input into English & Chinese

Use the English and Chinese translation results to obtain company information from relevant business information websites

Combining with Elasticsearch to build an index of multilingual text

Detect the language input by the user and display the enterprise text information in the corresponding language

Q: Introduce the implementation logic of  ENS Assets Appraisal 

A: For domain names of different lengths (three types - 3 characters, 4 characters, and 5+ characters), by using (n-gram) algorithm model to generate sub-word words of the domain name, and calculate the popularity of sub-word words (tf- weighted calculation of idf)

Use sub-word granular words to train the vector representation of each word (word2vec model, a shallow neural network model)

Combining the word popularity of Sub-word with word vectors, the vector representation of the domain name is finally constructed

Domain name vector characterization Combining domain name recent sales data, domain name features, domain name registration costs and other features, build a regression model to predict the short-term value of domain names

Q: Introduce the Label Generation for ENS

A: By monitoring the data on block chain, build domain name registration, wallet address and domain name correlation graph in the transaction process, and generate wallet address and domain name labels at the same time. Domain name tags include club tags, language tags, value ranking tags, smart money tags etc.

Q: Introduce the Algorithm Recommendation

A: For domain names of different lengths (three types - 3 characters, 4 characters, and 5+ characters), by using (n-gram) algorithm model to generate sub-word words of the domain name, and calculate the popularity of sub-word words (tf- weighted calculation of idf)

Use sub-word granular words to train the vector representation of each word (word2vec model, a shallow neural network model)

Combining the word popularity of Sub-word with word vectors, the vector representation of the domain name is finally constructed

Using Facebook's faiss vector retrieval technology to calculate the similarity between domain names

Q: Describe the technical architecture Dbrands used

A: Some of frameworks and components we are using are Flask、django、redis、MySQL、elasticsearch、docker、pytorch、kibana、vue ,etc

Q: What's your opinion on the future of ENS ecosystem

A: Vitalik has published a new book in 2022 <**Decentralized Society: Finding Web3’s Sou**l>.

We believe more and more web2 users will enter the decentralized society and everyone needs web3 identity to represent themselves and link with each other. The entire industry and ENS ecosystem are still early, as a startup team, let’s keep building. Thank you all.

Twitter:

https://twitter.com/dmgdatadao