CryptoTheorem parses through 850,000,000 social media discussions and news every day using proprietary machine learning and natural language processing technology to derive sentiment on cryptocurrencies.
Learn about the power of CryptoTheorem's sentiment data and patented data collection, cleaning, and scoring process that drives it. Check our articles and research reports below.
STEP 2 - Extraction
Each conversation is scanned individually to extract meaning, sentiment and ultimately the price impact related to particular cryptocurrencies. For example, a tweet mentioning LTC.X would be bucketed into a group of tweets on Litecoin. Using patented technology CryptoTheorem is able to combat problems related to overlapping tickers across asset classes. For example, identifying context whether an ETH tweet is on Ethereum or Ethan Allen, the publicly traded furniture company.
STEP 3 - Evaluation
A filter is used to assess the relevance and novelty of the data and to remove spam and fake news. Sources and accounts are reviewed and reported bots are removed from some datasets.
After the evaluation process has been completed over 90% of content on a particular asset class are filtered out of the stream.
Proprietary algorithms are used to identify and rate professional investors and traders.
STEP 4 - Calculation
Each expression in each conversation is individually scored using CryptoTheorem's proprietary natural language processor using machine learning technology developed over the last 7 years. The scores of each expression are aggregated and each individual tweet or news is given a score.
STEP 5 - Actionable data
CryptoTheorem produces raw sentiment scores which it converts into actionable insights and signals for cryptocurrency traders.
Among those insights are real-time sentiment scores which measures the total sentiment of an asset class on a 24 hour moving average sentiment score.
Sentiment is measured on a per asset basis to ensure that scores reflect how positive or negative conversations on social media platforms are for a particular asset relative to itself on other days. This helps remove bias of one asset generally having more positive or negative conversations that another.
The entire sentiment scoring process is done in less than 1/3r of a second and scores are updated on CryptoTheorem in real-time.