FAQ - Context Analytics

Was there a methodology change in 2016-2017?

In 2017, Context Analytics switched from scoring messages with cashtags only (ex. $AAPL) to a topic model based approach which allowed for additional terms (like company names and abbreviations) to be included in sentiment calculations. This also allowed for exclusionary terms on topics that have ticker conflicts with foreign equities or cryptocurrencies.

Did the sentiment scoring algorithm change?

Context Analytics’ NLP dictionary weights on words and phrases do change over time. This could be the inclusion of new terms or phrases or alterations in the magnitude of current weights. The dictionary is evaluated quarterly and small percentage of words/phrases in the dictionary are added/altered.

Have data sources (e.g. Twitter API access) change?

Twitter API access did not change.

Why does my historical sentiment data show a sharp drop in early 2016?

If you're observing a significant decline in sentiment values around March 2016, this is likely due to methodology updates rather than actual market sentiment changes.

Key Points:

Raw-S vs S-Score: Raw-S represents the sum of all sentiment scores from tweets mentioning a security over 24 hours, while S-Score is a normalized metric. If you're seeing a sharp drop in 2016, you're likely using Raw-S rather than S-Score.

March 2016 NLP Dictionary Update: Context Analytics implemented a major update to its sentiment dictionary on March 1, 2016. This update included:

• Addition of new sentiment terms and phrases
• Refinement of existing sentiment weights to be less extreme
• Expansion of the dictionary to capture more terms per tweet

Impact on Raw-S Values: The combination of reduced phrase sentiment magnitudes and increased term coverage resulted in lower Raw-S values across securities. This was a methodological change, not a reflection of actual sentiment shifts in the market.

Recommendation: For analysis spanning pre- and post-March 2016 periods, consider using S-Score (the normalized metric) or applying appropriate adjustments to account for the methodology change.