Broker Algorithm Sample Clauses

Broker Algorithm. The algorithm which is currently implemented and deployed works in the following manner: 1. For every immediate user requirement: Collect the key­value pair from relevant taxonomies (Exhibit 5) 2. Create data­packet in memory and send to a search and retrieve component. 3. Search and retrieval component searches for these key value pair combinations within the Software Tools Catalogue and relevant standards requirements (if any). This components makes extensive use of regular expressions to make sure all possible matches are retrieved initially. 4. A full list of potential matches are retrieved as a result. 5. Based on the preferences set by the user for each field (Mandatory, Optional or Unknown) and the range of TRL level: Add weights to each result. 6. Create a final rank for each result object by aggerating the weights for each field within the result object. 7. Sort results as per final ranking 8. Send results back to user interface for display purposes. On­going and future work on the algorithm includes the parsing of Needs to further refine the ranking. This will involve the use of Natural Language Processing techniques to perform automated entity extraction from free text entered by users. For example, we could create classifiers for semantic extraction of concepts like desired features, funding (which may not be expressed in numerical form or embedded somewhere within the text), barriers (examples might include what the tool should not contain e.g. a commercial licence as it is outside the organisations policy). Similar techniques when applied to the generic organisational profile will provide further insight into the broad needs (e.g. a post­production house aiming to make the transition from HD video production to 8K may have changing AV preservation needs in terms of storage). Future work will also include methods on scaling up the algorithm so that it can deal with large datasets of requirements and tools. The current size of these databases in PrestoCentre is quite small and hence the current implementation of the broker works fine. In the future, when the amount of data is vast, the algorithm will need to implement reverse indexing scheme as a way to retrieve results quickly and efficiently.