Patented Artificial Intelligence

​patented artificial intelligence

Dr. Anna Maria Di Sciullo and her team of Ph.Ds. and Post-Docs have developed differentiated technology over the better part of 10 years. With expertise…

INDUSTRY LEADING SENTIMENT ACCURACY: ~85% vs. ~60% industry average

iS​ENTIUM’s patented Natural Language processing AI technology has been developed by a world class team of linguistic thought leaders led by Dr. Anna Maria Di Sciullo.

Core technology is unique in utilizing linguistic algorithms to provide precise and efficient processing of social media messages.

iS​ENTIUM ~85% accuracy vs. ~60% achieved by Neural Network systems and Machine Learning methods.

​AI can be compared to the human brain and data to the oxygen that feeds it. If you feed the brain carbon monoxide, it will die. There’s a lot of carbon monoxide in the internet’s atmosphere. The digital ecosystem is plagued with bots, trolls, agents of provocation, pornographers, propagandists, etc. Considering that data is bigger and more compromised than ever before, how does iS​ENTIUM determine the difference between authentic intelligence and damaging noise?

The answer is borne out by our patent as inventors of the chat bot, uniquely enabling us to better detect, categorize and suppress malicious content. iS​ENTIUM’s capacity was recognized by the Department of Justice with the issuance of a permit to identify instances of child exploitation, a targeted priority on its agenda.

​Filtration module detects novel spam methods
Suppression of ads, marketing and misinformation
Adaptive text normalization

​Identification of linguistic constituents
(e.g., adjective, noun, verb)
Emoji and emoticon detection
String-linear categorical representation

​Computation of hierarchical structure
No use of tree banks or reference to previous observations
Novel expressions are not problematic
​Recovery of covert elements
(i.e., filling the gaps in short messages)

​Lexical sentiment mapped to parts of speech
Ticker and domain-specific information is referenced
(e.g., ticker, company, product, etc.)

Sentiment score is calculated on the basis of structural relationship between
domain-specific and common sentiment bearing constituents​

white papers

(click to see the full document)

  • Natural Language Processing AI Technology Overview
  • Sentiment Analysis Methods
  • iSENTIUM Deterministic Computation vs. Stanford Deep Learning
  • Universal Sentiment Computation
  • Universal Grammar - Modeling human language and its diversity


(click to see the full document)