Smart NFT Search

Proposals Round 2
#1

Name of Project: Smart NFT Search smartnftsearch.xyz

Proposal in one sentence: Search for anything in NFT space with text. Just like you use Google search.

Description of the project and what problem it is solving:
NFTs (non-fungible tokens) have exploded in popularity since 2020. OpenSea, the largest NFT marketplace processes more than 2 billion ETH in volume on a monthly basis, with an ever-expanding number of NFT collections added every day. Currently, it is hard for beginners to get started in this space due the sheer volume of the information out there. It is very easy to buy a JPEG from any collection just based on the FOMO and then get disappointed when you lose your ETH.

That’s where Smart NFT Search comes in. The end goal is to make anything in NFT space searchable with text, just like you use Google Search. Whether it is general information about a collection you are looking for (what are doodles, who are people behind it), general market data information (what is trending today, what is the floor of pudgy penguins), attribute search for a particular collection that you like (azuki with a face mask). Anything you are curious NFT-wise about could be searchable on such platform.

We currently have a proof of concept running for attribute search for two blue chip collections smartnftsearch.xyz . This is already much more convenient than searching for that particular attribute in the marketplace. In a marketplace like OpenSea you need to search for a collection first, then search for attributes in the drop-down box; whereas in our platform you accomplish the same with a simple text search. The response from the community has been very positive and we are continuously building on it!

screen1
screen11920Ă—1287 129 KB

Grant Deliverables

We are taking a step-by-step approach in delivering the end goal of Smart NFT Search

Things we promise to deliver in a very short-term

  • Attribute search for 20+ popular collections (that had at least 5k eth in volume) with text
  • General market data search (example volume, floor; top collections; recent sales) with text
  • Enabling new kind of searches not possible with current market place interfaces
  • Cross-collection search being one of them (example sappy seals and pudgy penguins with blue background)

This raises a question whether it is possible to have a very natural interaction with NFT space.

Squad

Hal https://twitter.com/bowtiedhal

2 Likes
Wat.to - Smart NFT Search Round 2
#2

Initial Thoughts

Socially
I really like this design and thought of Google when using the interface! I also think this interface makes NFTs more like a shopping experience and really enjoyed the simplicity.

Data Oriented
On a technical point, I like how when I took out “wearing” in the “cool cat wearing sunglasses on the face and apple on the head” example, the search returned cats without sunglasses, but still with an apple on the head, so that was interesting. Is there a grammatical component that the search knows about in the english language? I see this as a great app to create data sources for future data science projects and I think it would greatly simplify creating a NFT dataset.

Deliverable Questions

  1. Is the data retrieval directly from the chain or will there be a database that holds the images and metadata?
  2. I agree, cross collection searching would be really cool, I wonder if there are arbitrage opportunities?

Possible Metadata to also include
I think a detail that would be great to know is the IP rights for each collection, might be tough data source but could be good value add.

Excited to see this app!

#3

Hi @vintagegold

Thanks a lot for kind words and enthusiasm for the project. Below answer to your questions:

when I took out “wearing” in the “cool cat wearing sunglasses on the face and apple on the head” example, the search returned cats without sunglasses, but still with an apple on the head, so that was interesting.

Yeah, the ML model that interprets the meaning of the search query is not perfect and is bound to mistakes :slight_smile: Part of it is due to the synthetic data generation and part of it is due to the number of iterations I trained the model for. Thanks for noticing this, expect better performance in future iterations.

Is there a grammatical component that the search knows about in the english language?

No, we take a raw query and “translate” it to the semantic meaning representation. The arguments in the semantic meaning representation are further passed into the right API call.

I see this as a great app to create data sources for future data science projects and I think it would greatly simplify creating a NFT dataset.

We might release the scripts for generating synthetic search query data in the future.

  1. Is the data retrieval directly from the chain or will there be a database that holds the images and metadata?

From the chain. We use Module API (recommend) to retrieve the listings About

  1. I agree, cross collection searching would be really cool, I wonder if there are arbitrage opportunities?

Maybe, we will find out :slight_smile:

Excited to see this app!
Thank you!