Impact Of Lexicon Based Detection On Surveillance Efficiency

 The lexicon-based detection system is used to detect risks in electronic communication. Most of the time it gets the job done, but sometimes lexicon-based analysis creates huge inefficiency and potential risks. The lexicon-based system is developed around alerts, These alerts are triggered whenever a certain keyword is used in conversation or in online chat. The biggest problem of this system is the return of false-positive results. Approximately 3,000 alerts may be produced daily, from a total of 75,000 messages collected and analyzed. Therefore, The large number of alerts produces a lot of burden on compliance surveillance teams, Especially when we know that the current market norm of a false-positive rate is 95 to 99.9%.

Overall, a lexicon-based detection system is not just an ineffective solution to combat market manipulation and abuse, but it also puts more workload on analysts who have to achieve regulatory compliance through these alerts. Change in communication platforms and reinvention of languages in an organization creates difficulty to monitor triggers. In the era of SMS, Whatsapp, Skype, etc. it is difficult to monitor with an outdated system like lexicon-based because it only adds more false-positive results. When it comes to compliance it develops a huge risk if we fail in detection, many of the offenders are charged with millions of dollars in fines.

By keeping this in mind, most financial institutions have increased their expedition on compliance than before. According to a survey by the risk management association, 50% of the organization replied that they spent 6 to 10% of revenue if it is related to compliance. Therefore, if the financial institution is spending this much amount of money then the organization must ensure that surveillance efforts are not more challenging and compliance is using the right tool for the job.

The communication compliance platform Shield approached its focus on the three C's i.e. characteristics, content, and context. Shield uses various text analyzing tools like fuzzy matching, AI algorithm, tokenization, and other expert-driven rules. It also uses machine learning and advanced natural language processing (NLP) tools to detect the context of a particular comment with a high level of accuracy.

The complete process takes place in a hybrid surveillance system, which has the capability to link the data with various surveillance systems to detect the market manipulation and abuse scene, data handling concerns, and many other aspects with complete perfection. This process reduces the false positive by 80% in comparison to lexicon-driven alerts.