Cognitive Text Mining
The cognitive text mining algorithms developed by Conscious Machines have been internationally acknowledged for their speed and accuracy in high volume indexing and search. The platform combines a range of techniques to deliver better results than commercially available search tools in a much shorter time frame. These techniques embody innovations that are the outcome of over 50 person years of research and include:
- GSOM based clustering and visualization
- Hierarchical clustering with visualization
- WFSGSOM – specialized GSOM for faster text clustering
- Adaptive Suffix Tree (AST) for capturing variable length sequences from data
- Parallel implementation of the GSOM
- Incremental clustering techniques with knowledge accumulation.
The platform has been incorporated into the Genix iBPMS and currently supports:
- Dynamic Federated Search and Visualisation
- Social Media Analytics
- Automated Marking of short-answer and essay-type questions in the PROTRACK Examination Management platform
- Search across all of the Genix iBPMS platforms.
The use of the Conscious Machines tools has resulted in a range of benefits when related to search outcomes including:
- Improved search efficiency and time savings
- Significantly higher quality of results
- Delivery of results based on the most current content
- Aggregation of results across multiple content sources
- Improved ranking of results drawn from multiple content sources
- De-duplication of content (rules based).
Where the platform has been trialed in environments requiring automated marking or short-answer questions, the solution has delivered 92% accuracy when compared to human markers and in some cases has actually highlighted errors made by human markers. The solution is being extended to address the marking of essay-type questions and preliminary results are encouraging.
The Conscious Machines platform is capable of being combined with traditional data mining tools to support the analysis of structured and unstructured data in a single pass. This capability has been demonstrated successfully in research environments and is currently being embedded in the iBPMS platform as part of a research program funded by the Data to Decisions Co-operative Research Centre in Australia. The integrated solution is being designed for use by a range of industry sectors including intelligence and law enforcement agencies.