Our Specific NLP Approach
NLP is at the core of Stockpule’s know-how, which differentiates us from our peers. Apart from collecting editorial content and key events, Social Media posts and comments are swift and short. Their length is mostly limited between 100-300 characters and represents around 90 percent of all collected information. To evaluate and classify these “informal” comments, NLP must be adjusted to social media to accurately achieve processable results. Stockpulse developed a unique system for financial markets. Supported languages are English, German, Chinese (traditional and simplified), Swedish, Finnish, Danish, and Icelandic.
Below we illustrate the methodology of our specifically adapted NLP and ticker matching for a sample social media message.
“So my TLRY Yolo 42C 2/12 is gonna pay is what you’re telling me?”
Our NLP and sentiment analysis detects TLRY (Tilray Inc., a Canadian cannabis company) with a neutral sentiment. The term “42C 2/12” is probably referring to a call option of the user, which would indicate a positive meaning. However, the question form of the sentence leaves some uncertainty and that is the reason why Stockpulse’s NLP labels this message with a neutral score.
Google‘s NLP API detects:
Google’s NLP detects a negative sentiment of -0.4 on a scale of -1 to +1
This example demonstrates the different NLP approach between Google and Stockpulse explicitly. The adapted use case and domain have not been correctly addressed, although Google is widely seen as the industry’s benchmark.
“$TLRY $APHA calls since 1/5/2021 got me to over $500k today. Still think it has much upside“
Our NLP detects the ticker symbols TLRY (Tilray Inc.) and APHA (Aphria Inc., also a cannabis company). The text receives a positive sentiment. The words “calls” and “upside” are the major factors. Especially the word “call” or “calls” is a very specific term in financial markets and normally means to buy an option on a stock, which can be associated with a positive meaning.
Google‘s NLP API detects:
Google labeled it negatively (-0.3).