Technologies

To facilitate research translation, CATOS consolidates a collection of technologies named the “OTS Toolkit”, which are at a mid-to-high readiness stage. All these mid-to-high readiness technologies are ready for evaluation and adoption if they are deemed useful.

The OTS Toolkit™ is being applied in the prevention, early detection, and mitigation of broad areas of harmful online content, in particular on Area A. Deception and manipulation and Area B. Toxicity, online harassment & bullying, and hate.

You may reach us via connect@catos.sg to request a demo or trial for any or all of these tools.

Early Detection
1. Audio-Visual Deepfake Detector (SLEUTH™ DF Detector)
2. Multimodal Automatic Fact Checker (SLEUTH™ AFC)
3. Multidimensional Emotion Analysis Engines from Natural Language (CrystalFeel™)
4. Multimodal Hateful and Social Divisive Content Detector (CrystalFeel™ Plus)
5. Neural Machine Translation Engines
6. Virality Predictor

Mitigation
7. Integrative Social Listening and Harmful Content Alert Platform (HarmWatch™, a.k.a. OTS MasterApp)

Prevention
8. Seamless Interface for Content Publication with Credentials (PROVO™)

SLEUTH™ Audio-Visual Deepfake Detector

Automatic detection of manipulations in audio and video deep fakes that involve human speech and faces

Multimodal Context-aware Automatic Fact Checker (mcAFC) Image

SLEUTH™ Multimodal Automatic Fact Checker

Discern false from authentic news, despite human or machine generated

Background Image of Multimodal Emotion Analysis Engines

Multimodal Emotion Analysis Engines is developed by A*STAR Institute of High Performance Computing (IHPC)to enable early sensing content and communities with high-intensity fear, anger, sad and joy emotions.

CrystalFeel™

Accurately analyse the explicit and implicit expressions of emotions—fear, anger, sadness, and joy—from written or spoken text, including intensity scoring and sarcasm detection

CrystalFeel™ Plus

High-sensitivity detection of socially divisive, hateful content associated with protected identities, with severity markers (Beta)

Neural Machine Translation Engines

Automatic translations of Mandarin, Malay, and Bahasa Indonesian content to English for multi-language analysis

Virality Predictor

Predictive scoring versus monitoring of digital content popularity and virality using network-based and content-based features

HarmWatch™

OTS Maser App for social listening and harmful content alert—with integrated backend AI engines, live data pipeline and frontend user interface—and capable of flagging content with heated emotions, non-factual claims, mid-high severity toxicity and hatefulness on trending topics or events

PROVO™

A seamless interface for publishing an image and video with verifiable credentials, enabling content creators and publishers to publish with “trust by design”