The 9 companies selected by the CDA to participate to the "Unstructured Data Challenge" presented yesterday in Aberdeen their results. The workshop organised conjointly by CDA and ECIM was very well attended with more that 70 representatives of operators and services companies actives in the North Sea.
It was very interesting to see the various approaches proposed by the innovative companies to extract values from the collection of 450,000 unstructured documents related to well and seismic surveys. Some have illustrate the potential of using Convolutional Neural Network (CNN) to classify the documents according to their graphic content, other adapted the last searches on taxonomy and ontology to measure the distance between geological description of the same interval between the well, other developed advanced QC tools to QC the well logs and create composites automatically from the unstructured data set.
It appears from this benchmark that the Agile Data Decisions' approach is quite unique. When comparing to our challengers, it could be noticed that AgileDD avoids the use of full text indexing to extract the useful information but prefers the application of text pattern detection driven by Machine Learning algorithms detect the targeted information and measure the confidence which could be associated to the detection.
Working this way, we limit the use of full text indexing post processing to avoid the extraction of false positive information and we think we could even extract more details.
It was very nice to have an insight on all the prototypes developed for the challenge, but as said by Malcolm Fleming, the CEO of the CDA, what it matters now is to see the various tools illustrated yesterday "at work". So he asked the representatives to reply now to the questions "What's next" by encouraging them to start pilots and feasibility studies using these new innovative tools within their data environment.
At AgileDD, we are ready to go!