Welcome to RETRIEVAL The on-line information retrieval performance evaluation tool.
Please sign in using your credentials or register in order to use this free-of-charge service. The Chrome browser is currently recommended.
- Retrieval offers several data input approaches such as binary ranked lists, unsorted floating point ranked lists and dissimilarity matrices.
- It comes with an increasing number of pre-installed retrieval benchmarks that are widely used by the scientific community in performance evaluation experiments.
- Retrieval offers interactive plots of various metrics such as precision vs recall, F-measure and E-measure.
- Retrieval comes with detailed help documentation and video tutorials that offer step-by-step guidelines for using different parts of the system.
- Plot content can be parametrised according to user's preferences and they can be downloaded as bitmap images or in a MS EXCEL compatible file type (CSV).
- Performance scalars such as NN, NN+1, 1st & 2nd tier, last place ranking, average precision, average dynamic call, cumulative gain, discount cumulative gain, fall-out rate, weighted f-measure and others are computed
- Metrics and scalars parametrisation according to user's preferences along with textual description of each parameter...
- A personalised file repository allows each user to have control over uploaded input data as well as performance reports files. Archiving reports into ZIP files is also supported.
- Performance results of pre-installed benchmarks can also be explored using the thumbnail-based ranking visualisation option.
- Batch performance evaluation jobs can be assigned for pre-installed benchmarks and computed at background while the user continues using other parts of Retrieval.
- Correlation Map visualisation of dissimilarity matrix with on-mouse over functionality to monitor performance of specific objects or outliers.
- Binary relevance ranking list scatterplot visualisation.