Over the last while, we at Hive Discover have been working on similarity searches and have made breakthrough improvements to our algorithm. Test it now at: https://www.hive-discover.tech

1. HIVE own text corpus

All posts from this December were stored in the database and analyzed regarding their language. This was important to get as large a word pool as possible from the HIVE space. Because now our text model also knows words like "splinterlands" and does not connect "hive" with bees anymore! So we can compare texts more precisely and find similar articles based on the text.
Available at the moment are only English, Spanish posts (also mixed bilingual ones). But other languages can be added if needed.

2. image analysis

As a second point, the images used in the posts are also compared. This gives us a second clue to find similar content, even if it is written in different languages!

As soon as a user requests similar content, the text and images are weighted and compared so that the most similar posts are displayed. This is particularly important for our content recommendation system, as we are not dependent on the users and their tagging skills. We have often had good articles suggested to us in this way, but they have received very little attention due to poor tags and community choosing.
You can try it out on our site by clicking on a post, or by calling the API manually. How to do that will be shown later.

Similar content can be seen on the right under the author profile. Example post by @aswita

A semantic search engine has also been created in this way. In contrast to HiveSearcher, it does not search for keywords, but only for similar content based on text and images. It's not perfect yet, but it's already going in the right direction (currently only the last 7 days are indexed):

If you don't want to use our website to find the content or you want to integrate the results into your website yourself, you can also call our API manually. Here's how to do that:

Find similar posts:

The parent_permlinks and tags fields are optional and can be omitted.

Request Body
{
    "permlink" : "bar",
    "author": "foo",
    "parent_permlinks" : ["hive-foo"],
    “tags” : [“bar”],
    "amount" : 7
}

curl --location --request POST 'https://api.hive-discover.tech/v1/search/similar-post'  \
--header 'Content-Type: application/json'  \
--data-raw '{
    "permlink" : "bar",
    "author": "foo",
    "parent_permlinks" : ["hive-foo"],
    "tags" : ["bar"],
    "amount" : 7
}'

Semantic search

Request Body
{
    "query":  "foo",
    "amount": 10,
    "type" : "smart" | “text” or “images”
}

curl --location --request POST 'https://api.hive-discover.tech/v1/search/posts'  \
--header 'Content-Type: application/json'  \
--data-raw '{
    "query":  "foo",
    "amount": 10,
    "type" : "smart"
}'