Background to the AURA Library

AURA (Advanced Uncertain Reasoning Architecture) is a generic family of techniques and implementations intended for high-speed approximate search and match operations on large unstructured datasets. AURA technology is fast, economical, and offers unique advantages for finding near-matches not available with other methods.

AURA is based on a high-performance binary neural network called Correlation Matrix Memory (CMM). Typically, several CMM elements are used in combination to solve soft or fuzzy pattern-matching problems. CMM has been the subject of research and development by Prof. Jim Austin and his group at York for more than 15 years, leading to a range of related methods now known collectively as AURA.

AURA takes large volumes of data and constructs a special type of compressed index. AURA can find exact and near-matches between indexed records and a given query, where the query itself may have omissions and errors. The degree of nearness required during matching can be varied easily. AURA implicitly supports powerful combinatorial query, which accepts a match between, for example, any 5 from 10 fields in the query against the stored records. The degree of index compression is preset, allowing a trade-off between storage efficiency and accuracy of recall. In practice, this means that AURA is guaranteed to find all genuine matches but will typically find additional false matches, depending on the degree of index compression used. The false matches are easily detected in the relatively small result set by conventional (but computationally slow) matching techniques.

The increasing range of applications for AURA includes:

• Postal address matching
• High-speed rule-matching systems
• High-speed classifiers (e.g. novel k-NN implementations),
• Structure-matching (e.g. 3D molecular structures), and
• Trademark-database searching.

Other applications under development include data analysis and case-based reasoning systems.

The core techniques used in AURA are derived from the earlier ADAM system, developed by Jim Austin for recognising image features in computer vision applications. The much wider application potential of these methods was recognised and led to the definition of the basic AURA system.

The goal of the original AURA research was to develop techniques better able to exploit the potential of CMM. The project developed a new class of architectures supporting high-performance knowledge based systems in real-time applications. The original AURA project ended in July 1997, and is succeeded by the AURA II project. AURA II aims to develop larger scale implementations of AURA, a version of the AURA software library with improved usability, and an extended range of applications. The project will also investigate methods to distribute pattern-matching tasks over a number of parallel CMM's, and new PCI-bus CMM hardware will be investigated in standard PC environments.