What is Deception Analysis?
One of the most exciting features of Amenity Viewer is a natural language processing capability that we call the Deception Model. For earnings sentiment analysis, we define Deception as events in earnings call transcripts that highlight language corporate executives use to divert, pivot, or avoid providing commentary on a topic. Deception analysis is unique to text analytics since it’s virtually impossible for a person to uncover these valuable extractions without an NLP platform.
What exactly is our text analytics model looking for regarding Deception? The model has been trained to not only parse every word of every sentence for the underlying financial meaning but also to identify linguistic patterns that Amenity Analytics has determined as being consistent with deceptive language. Our text AI identifies linguistic trends that we call “Events” under the key driver, Deception. These Deception Events may indicate evasive language, attempting to spin a negative, tension with the analyst community, etc.
Deception Analysis in a Document
To highlight an application of Deception analysis, we analyze Lear Corporation, an automotive supplier.
Notice that Deception is highlighted as one of the Key Drivers behind the significant drop (-36%) in the Amenity Score for Q3 2018:
Selecting the latest earnings transcript, and sorting by the number of negatives in the Key Drivers tab, causes Deception to rise towards the top:
Next, in the navigation pane we expand Deception to view extractions of the deceptive events:
We see Detours - when a management team is trying to redirect and not answer the question directly. In this example the statement “Here's what I can tell you…” tries to spin a negative regarding a near-term headwind.
Scrolling through more deceptive events, we see Evasive commentary such as, “I don’t know exactly when that will occur,” again, following a positive statement. And you start to see a theme: the model is identifying hidden negatives surrounding seemingly positive comments:
Lastly, the phrase, “I can’t remember historically having this kind of downtime,” shows an example of Selective Memory:
These are the type of comments that the Deception Model can identify and cluster together, allowing you to spot patterns when management teams are being incrementally more evasive, less direct, or more confrontational with analysts relative to prior quarters. In your next big earnings call make sure to apply your analysis through the lens of the Deception Model to see what you've been missing.