Heimdall

Kevin Voigt
Heimdall Research
Published in
5 min readNov 3, 2020

Heimdall is a powerful social listening tool that filters and analyzes Twitter data in real-time, mapping and measuring all social interactions related to cryptocurrencies.

Developed by traders, physicists, and mathematicians: Heimdall is a gem hunter tool to help you watch cryptocurrencies as they grow.

Social Score

Our social score is a key metric measuring the social relevance of users, tweets, and whole communities. It’s heavily based on graph theory and was inspired by Google’s search engine algorithm.

Our model

Graph theory is a branch of mathematics that studies the relationships between objects in a given set. Centrality is a measure of the importance of a vertex in a graph and is related to the number of times that a node acts as a bridge along the shortest path between two other nodes.

Twitter can be interpreted as a graph, where we have users and tweets as vertices and the relationships between two vertices, such as when a user tweets a tweet, represented by the edges of the graph.

PageRank ™ is an algorithm, among many, used to calculate the centrality of a vertex. It is used by the Google search engine to position websites among your search results. PageRank ™ measures the importance of a page by counting the quantity and quality of links pointing to it.

PageRank ™ considers that whenever you point to another, you are giving a vote for the relevance of that, however, the value of your vote depends on who points to you.

If you are interested in a better understanding of how the algorithm works, we recommend this reading.

At Heimdall, we build an algorithm inspired by PageRank™ that dynamically measures users centrality for the cryptocurrency Twitter subgraph. We also extend that concept propagating the user's centrality score — a.k.a relevance, to tweets and communities.

Key Aspects

To measure a user's relevance to the crypto community, an extremely simpler solution would be to use their number of followers as a relevance metric. However this simple model, in fact, hardly ever presents good results, because it opens loopholes for bots and non-engaged users to have the same network-weight as influential users. In contrast, for our algorithm more important than how many follow you, it’s who follows you. Quality over quantity.

To score the relevance of a community we compose the relevance of all users who interacted with the community, in that way communities gain much more social score when highly socially engaged users interact with them, like @binance, than with newly created accounts interactions. In fact, newly created accounts have extremely low social scores, as they have never received any social engagement. Our algorithm also penalizes spam, users who talk a lot and receive little engagement have their social relevance drastically reduced.

To summarize, our community social score measures how truly socially active a cryptocurrency is, user by user, interaction by interaction, all done in a very smart way, and in real-time.

Sentiment

Another key metric available at Heimdall is our sentiment analysis. Using our centrality score it becomes easier to identify highly active communities, but without a sentiment analysis, it would still be unclear if a community is bearish or bullish. Our sentiment algorithm identifies the sentiment in every tweet, classifying them as negative, neutral, or positive.

Our model

Machine learning algorithms require data sets as a way to learn and recognize patterns, which means that if we want to build a sentiment classifier, we would need a large number of tweets with sentiment classified by humans.

There are a lot of open data sets and already trained sentiment classifiers, but none of them perform well for cryptocurrency sentiment analysis. The most common definitions of positive/negative sentiment are very associated with happiness, angriness, etc., which is not exactly what we’re looking for. When talking about cryptocurrency sentiment, we’re looking for market-related sentiment, most commonly known as market optimism/pessimism or bullish/bearish sentiment. Also, cryptocurrencies have a very unique vocabulary when expressing sentiment, like expressions: “pump”, “moon”, “rekt”, “scam”, “shitcoin”, etc…

After failing in achieving good models with public data sets, we decided to classify our own: we’ve classified thousands of tweets using our own definition of sentiment, every tweet was classified by at least three people with deep cryptocurrency knowledge, to ensure data quality.

Using our own data set we trained the most known and battle-tested machine learning and artificial intelligence algorithms, selecting and composing those which performed the best.

Key Aspects

Our sentiment algorithm can detect Twitter and cryptocurrency’s specific way of communication, truly identifying tweets that have buying or selling intentions related to cryptocurrencies. Different from other implementations, our sentiment analysis is very correlated with price change — as is expected.

Sometimes sentiment analysis is an overvalued feature, people tend to believe in a short path to alpha. In real life, sentiment analysis is not a magic way to earn money, but an easy way to synthesize information. It’s expected that when markets are rising people would behave and express optimistically, but it doesn't mean that when sentiment rises price will necessarily follow.

Due to the large number of tweets related to cryptocurrencies, it would be practically impossible to keep track of everything. Our sentiment score should help you save time when trying to identify hype and fear, without needing to dive deep in Twitter itself — or in Heimdall Relevant Tweets section.

Don’t miss a thing

In Norse mythology, Heimdall is a god with exceptional senses, he can see and hear the grass as it grows. We’re compromised on bringing reliable and transparent data to the cryptocurrency decision-making process as soon as they begin to grow. With our algorithms we automatically detect and start tracking all social interactions even for recently born projects related to cryptocurrencies, making sure you’re never missing a thing.

Heimdall, a bunch of crypto enthusiasts seeking for the future.

Available at: https://heimdall.land/

--

--

Kevin Voigt
Heimdall Research

Physicist, data scientist, blockchain enthusiast and CTO at heimdall.land