TECHNOLOGY

Real-time processing, search, classification, identification... and more for all accessible information—in any format

Technology 2017-09-26T22:17:21+00:00

AiCURATE™ API and Product

Sigma Technologies Global has been working at the cutting-edge of machine intelligence for almost ten years.  There are three primary components to Sigma’s technologies: Signal Processing, Machine Learning/Intelligence, and Artificial Intelligence. Often these terms mean different things to various stakeholders. For the sake of clarity, we define them as:

Signal Processing is a set of techniques and algorithms that prepare signals in order to extract useful information from them. Signal processing includes filtering, signal enhancement, noise reduction, and format adaptation, among others.  

Machine Learning techniques provide computers with the capacity to learn from examples, in much the same, intuitive way in which a human-being learns to perform a new task. Sigma has developed a variety of fresh and innovative machine learning solutions based on a number of algorithms such as Hidden Markov Models, Support Vector Machines and Deep Neural Networks.   

Artificial Intelligence techniques imitate the thought process by which a human expert would solve a complex problem or perform a new task. So, the human expertise is transferred to the machine through the use of rule-based frameworks, which are still a key component to many solutions and are often used in tandem with machine learning.

Sigma’s API and Artificial Intelligence Product, AiCURATE™ examines and processes text, audio, image, video and biometric data. This broad technology coverage allows Sigma to build simple to extraordinarily complex and specialized solutions that combine a variety of sources of information.

AiCURATE™ API and Product Description

The AiCURATE™ API and Product is composed of three modules that can be used separately or in combination, depending on the use case. This allows for flexibility to use only those functionalities that are relevant to the problem at hand, a type of algorithmic tool kit if you will.  Through the use of web services, the integration of the AiCURATE™ API and Product with other platforms is straightforward.

Input Module

Pre-processes the text, audio, image, video or biometric data. Among other functions, it performs format adaptation, coding conversion, equalization, noise reduction, and filtering. It also provides web “crawlers” to look for all relevant information on the internet including Facebook (public), YouTube, Instagram, Snapchat, Twitter, Reddit, Vine, Pinterest, discussion forums, consumer review and social shopping networks, blogs, websites, streaming video and audio, radio, and television. These crawlers are proactive and utilized in a continuous manner, so data can be later processed and analyzed automatically.  With information and data sources being updated continuously 24/7, this module updates to the most current version of the data through asynchronous callback of the original data source. Essentially, in idle time, the module is checking whether the data is current and hasn’t been changed, ensuring the information used isn’t stale or obsolete.

Machine Learning and Artificial Intelligence Module

In the case of text processing, this module can work on over 60 different languages and dialects. All the European languages and dialects are covered, as well as most Asian languages, and the various dialects of Arabic. In the case of speech processing, the current module has an automatic speech recognizer (ASR) for English and Spanish. In the case of speech processing, the current module includes an automatic speech recognizer (ASR) for English and Spanish and Sigma has the ability to develop an ASR in another language within one to two months, depending on the data available and the complexity of the language.

  • Automatic Speech Recognition (immediate capability in English and Spanish)  

  • Transcription (conversion from voice to text)

  • Time alignment, consisting of indexing the transcription with the audio content to identify the instant of time in which each word was pronounced.  This provide audio search and analysis capabilities

  • Gender identification from voice, image and video

  • Sentiment/emotion/behavioral analysis from text, audio, image and video data

  • User authentication based on voice, handwritten signature and facial recognition  

  • Classification/Categorization of image and video through the use of customizable categories.  This allows the module to automate the classification of images and videos based on content.  It detects different objects, locations, activity, and events such as people, outdoor or indoor scenes, vehicles, car makes and models, flags, etc. Potentially any object, thing, person, or contextual situation/idea can be detected.

  • Object recognition in image and video. For example, for detecting logos, specific buildings or monuments, or any other object of interest, even if the object is semi-hidden. It can find anything given the client or problem demands needing to be found/identified.  The system is fully configurable.

  • Detection of people, as well as their faces in images and videos

  • Content-based text classification (the categories of course are customizable).  Clustering texts to group texts with similar content. For example, to organize information or to extract a subset of texts that are representative of a larger number of texts.

  • Topic identification (topics are also configurable

  • Language identification from text and audio

  • Text and audio analytics

  • Configurable intelligent agents and/or chatbots

Analysis Module

Uses the information obtained from the previous two modules and performs the following  data statistical analysis.

  • Data indexation (of audio and image/video)

  • Annotation and labeling (for audio, text and image/video)

  • Combining data and information extracted from unstructured data sources with information from structured data sources.

  • Information search and retrieval on unstructured sources of information (e.g. text, audio, image and video)

  • Alert generation

  • Summary, report and/or narrative generation

Sigma API Diagram

Conceptual Architecture of Sigma’s API