Deep Web and Social Analytics


The Deep Web (DW) consists of a massive amount of mostly unstructured data – primarily textual information but also visual, like images, videos and interactive components. AGT Deep Web & Social Analytics transform this unstructured data into usable information by harvesting the data, normalizing, cleaning it and using Big Data analytics to extract meaningful insights.

Below are several key examples of AGT’s Deep web and social analytics software:

Entity Extraction – Take any type of unstructured content, detect entities and classify them into meaningful types and categories such as people, places, organizations, companies and more. AGT uses a platform-agnostic approach that can analyze any textual data source. Analysts can define entities in any language and organize them into hierarchal groups to facilitate complex queries

Social Network Analysis – Investigate and explore social structures (both physical and virtual) through the use of network and graph theory. AGT analytics measure centrality, density, cliques, distance and other attributes that reveal underlying structures, relationships and roles so that you can quantify the level of influence or importance of a particular node within one or more networks.

Video & Image Analytics – Identify specific objects or behaviors in video clips and assemble summaries based on textual information. AGT analytics work on existing videos as well as real-time video cameras or streaming services. They can detect faces and facial expressions, vehicles or crowds and can register changes such as a reduction in crowd density. 

Profiling Analytics – Extract relevant profile parameters from huge deep web data stores and create meaningful clusters and insights to guide personalization. AGT supports clustering, entity resolution across different networks, identification of behavioral patterns (e.g., every other Tuesday afternoon she goes to visit her grandmother, every home football game five friends meet and head over to the stadium), and predictive analytics that forecast how an individual or group will behave.

Geospatial analysis – understand how information, behavior, and attitudes are distributed over a geographical terrain and how they change. AGT geospatial analytics use smart devices’ location information to analyze data based on dynamically reconfigurable geographical areas in real time. Analysts can tap into an array of analysis capabilities over an interactive map to generate a multi-perspective view of an event or topic from varied aspects and generate insights such as leading chatter topics, key contributors, concentrations, trends or routes. AGT can correlate deep web data with sensorial IoT data from the same geographical area and sometimes from the same device.

Temporal analysis – Understand how entities, topic, and overall chatter change over time as reflected in what people are talking or commenting about and identify differences between evolving situations and similar past events. AGT uses machine learning capabilities to examine past data, generate models and predict trends, volumes, peaks of activity and turnarounds.

Anomaly detection – Understand what’s actually an anomaly within social and deep web data. Originally developed for IoT analytics, the AGT analytics engine automatically constructs a model for the norm for situations, behaviors or attitudes and identifies significant deviations from this norm

Sentiment analysis – determine the attitude or “tone” expressed in textual data with respect to some topic or entity. AGT analyzes sentiment both by identifying keywords with an inherent sentiment (happy, bad, annoying, etc.) and by applying sophisticated machine learning classification schemes to data ranging from tweets to lengthy documents. Analysts can fuse this data with IoT data such as facial expressions (using video analytics) and cheering or booing (using audio analytics) to produce even more accurate sentiment analysis.

Visual Link analysis – Search for known relationship patterns as well otherwise hidden, indirect relationships and abnormal patterns by viewing a linked network visually. AGT analytics display the related entities (nodes) and their connections in various layouts, applying a vast set of filters, finding shortest paths, and attributing colors, size and other advanced properties to nodes and edges. Visual link analysis has been proven to be very effective in law enforcement and intelligence settings.

Metadata analytics – Extract and analyze metadata from sources ranging from the camera that took a particular picture to the follower count of a person who sent a particular tweet. Metadata analytics add an additional layer of intelligence to deep web data, intelligence that’s usually invisible to the analyst. AGT analytics allow for compliance with privacy policies and related regulations set by different countries and companies

Share this: