Content Understanding is the ability of computers to ingest, relate, and process information contained in data and to perform tasks using that information. Content Understanding considers all types of data, including text, speech, images, sensor data, and static or streaming data.
The question of whether computers can truly be said to "understand" is endlessly debatable. In the field of Content Understanding, understanding is demonstrated by acceptable performance on one or more tasks. Content Understanding tasks are those which would usually be said to require or demonstrate understanding, if performed by a human. Content Understanding tasks include the following:
Interpret, Categorize, Prioritize, Answer, Sort, Correlate, Select, Find, Translate, Summarize, Discover, Explain, Relate, Alert, Infer, Predict, Explore, Decide, Respond, Recommend, Visualize, Synthesize, …
Research in Content Understanding is pushing technology toward automating these tasks in increasingly complex applications. Two major dimensions of this complexity include data size and incorporating context.
Content Understanding and Big Data
Today, the amount of data relevant to many Content Understanding tasks exceeds the amount of data that can be effectively used by the current methods of performing those tasks. This is the "Big Data" problem, and formulating a solution involves understanding the tasks, modeling the large volumes of data, and matching effective automated technologies appropriate for the data sizes. Effective automation of Content Understanding tasks that are currently performed manually is necessary to match capabilities to modern data sizes.
Content Understanding and Context
Work in Content Understanding recognizes that relationships between information contained in one piece of data and information contained in other pieces of data are often essential for effectively using that information. An important research goal is identifying relationships and creating context-aware solutions.