This file contains a history of changes made to CoMIRVA's releases.

version 0.36:
- included retrieval from Twitter API
- new packages for text-based similarity calculation and normalization (comirva.web.text.similarity and comirva.web.text.similarity.normalization)
- new classes for indexing retrieved (Web) documents using lucene (comirva.web.indexing)
- additional implementation of the Fluctuation Patterns extractor (using the Cent scale)

version 0.3:
- restructured and reorganized the packages
- added various information extractors for music information systems (last.fm, allmusic.com, Yahoo! Artist Directory)
- implemented various approach to detect the country of origin of a person (comirva.web.ie.countryoforigin); some of them will require Lucene Java libraries
- added an option to Web retrieval-related dialogs: "Force Use of Phrase Search for Query Terms" (automatically quote all search terms)

version 0.2.N3:
- completely redeveloped GUI

version 0.11:
- implementation of the Growing Hierarchical Self-Organizing Map (GHSOM)
- enhanced ETP creator (index audio, image, video files)
- new class GoldenRetriever for high-performance Web page fetching with load balancing

version 0.10:
- added new colormap "Ocean"
- modified visualizations to run under Java 6 (optimized thread-based processing)
- added EPS output functionality for visualizations ("Visualization" -> "Export to EPS..."); for this function to work, "Enable EPS-Output" in "Visualization" -> "Preferences..." must be enables
- Principal Components Analysis (PCA) for data projection implemented ("Data" -> "Data Matrix" -> "Principal Components Analysis...")
- modified term extraction functionalities to be able to index multimedia content (audio, image, video files) in web pages
- implemented "Music Description Map" ("Visualization" -> "SOM" -> "Create Music Description Map (MDM) Labels" and "Show MDM-Grid")
- export SOM- and MDM-grid visualizations to HTML ("Visualization" -> "SOM" -> "Export SOM to HTML..." and "Export MDM to HTML...")
- given a list of MP3 files as meta-data vector, the new function ("Data" -> "Meta-Data" -> "Extract ID3-Tags from File List") extracts artist and song title of each of the MP3-files in the list and adds them to new meta-data vectors
- some minor bugs in various visualization methods were resolved
- resolved error in SOM initialization method due to incorrect use of Double.MIN_VALUE

version 0.09:
- improved apprearance of SDH, borders are smoothed now
- removed bug creating NullPointerExceptions in comirva.util.FileUtils.getAllFilesRecursively
- implemented audio feature calculation of GMM-MEs according to Mandel and Ellis (e.g. http://www.music-ir.org/evaluation/mirex-results/articles/audio_genre/mandel.pdf)
- split visualization classes in different thread-based subclasses - one for every visualization - for better modular design
- changed default values of SOM and SDH creation (now optimized for speed rather than quality)
- optimized calculation time of MFCCs

version 0.08:
- resolved a bug: saving a meta-data instance was not possible unless at least one data matrix was present
- new function to extract terms from retrieved HTML-documents: "Data" -> "Web Mining" -> "Term Profile Creation" -> "Extract Terms from Retrieved Documents"
- resolved same key-shortcut for "Data" -> "Meta-Data" -> "Sort List" and "Visualization" -> "Similarity Matrix" -> "Continuous Similarity Ring (CSR)"
- implemented audio feature calculation of MFCCs according to Aucouturier and Pachet and Fluctuation Patterns according to Pampalk
- improved performance of linear init of SOMs

version 0.07:
- new fonts for UI components
- modified "About"-dialog: added contributors
- new function "Data" -> "Web Mining" -> "Term Profile Creation" -> "Retrieve Meta-Data-Related Web Pages...": crawls web pages by first querying a search engine for values in meta-data instance and then retrieving the resulting pages and storing them locally (wget is needed to perform this function!)
- new function "Data" -> "Web Mining" -> "Term Profile Creation" -> "Create ETP: Entity Term Profile(s) from Retrieved Documents...": creates term profiles from the retrieved web pages (or any other textual documents) and stores them as XML-files (a meta-data vector containing a list of terms must be loaded)
- new function "Data" -> "Web Mining" -> "Term Profile Creation" -> "Load Term Occurrence Matrix from ETP-XML-File(s)...": loads the term occurrence matrix together with the corresponding term list from an XML-file containing an Entity Term Profile (ETP)
- implemented "batch SOM" training algorithm, modified SOM-dialog: user-selectable training method and training length
- directory selection bug in file dialogs solved: no longer possible to open a directory when a file is needed (e.g. in "Load Matrix Data File..." and "Load Meta-Data File..." dialogs)
- new function to sort list of data matrices w.r.t. to their name ("Data" -> "Data Matrix" -> "Sort List")
- new function to sort list of meta-data w.r.t. to their name ("Data" -> "Meta-Data" -> "Sort List")
- new visualization "Sunburst" for term co-occurrence matrices ("Visualization" -> "Term Occurrence Matrix" -> "Sunburst")
- modified the colormap "Sun" (old: from yellow to orange, new: from yellow to red), the new version can be used well for "Sunburst" visualizations
- new function to store meta-data instances ("File" -> "Save Meta-Data File")
- new function for updating paths in XML-serialized ETPs ("Data" -> "Web Mining" -> "Term Profile Creation" -> "Update Paths in ETP-XML-File(s)..."): since the paths in ETP-XML-files must be absolute, this function updates the paths automatically after the corresponding files have been moved to another location
- new initialization methods for SOM: linear init according to Kohonen, gradient init, SLC (according to Su, Liu, Chang: "Improving the Self-Organizing Feature Map Algorithm Using an Efficient Initialization Scheme")
- improved placement of labels in SOM and SDH visualizations (avoiding roundoff errors)
- resolved a bug in "Probabilistic Network" visualization: using the predefined label size by "Visualization" -> "Preferences..." for the mouse-over labels now works
- resolved same key-shortcut for "Data" -> "Meta-Data" -> "Sort List" and "Visualization" -> "SOM" -> "Create SOM"

version 0.06:
- additional option in "Co-Occurrence Analysis" -> "Retrieve Page Counts", new radio button "Usage of Meta-Data" inserted, "query for all pairs" performs the queries exactly as in previous versions (each pair of elements in the selected meta-data-instance is queried), "query for single items" uses the selected meta-data-instance as list and queries each of the element in it
- new "Continuous Similarity Ring"-visualization
- new function to invert colormaps
- changed ColorMap_Islands in order to avoid too smooth transitions between "ocean" and "sand" and "sand" and "meadow"
- new function "Data" -> "Co-Occurrence Analysis" -> "Requery Invalid Entries in Page-Count-Matrix" searches the loaded page-count-matrix for -1 values (no page-count-retrieval was possible) and raises the appropriate query again
- new functions to normalize data matrices ("Data" -> "Data Matrix" -> "Normalize...")
- redesigned data matrix and meta-data lists in user interface
- created new package comirva.audio containing audio player and a playlist classes
- all file chooser dialogs remember the path of the latest opened file and automatically access it the next time a file chooser is opened
- implemented a simply audio player (accessible via menu "Audio" -> "Show Audio Player")
- file open/save dialogs now have the Java look&feel instead of the platform-dependent L&F
- solved a bug in the meta-data loader (if a non-valid file was specified in the user dialog, an item was added to the meta-data list nevertheless)
- user can now set global preferences for the visualization area ("Visualization" -> "Preferences...")
- now also for "Save SOM..." and "Save SDH..." (and not only for "Save Matrix Data File...") it is checked whether a file with the given name already exists (if yes, the user is asked whether s/he wants to overwrite it)
- new function to store the content of the visualization area ("Visualization" -> "Save Visualization...")
- new functions to load/save workspaces (data matrices and meta-data) in menu "File"

version 0.05:
- inserted separation lines between arcs in the "Circled Bars"-visualization: with these lines the individual arcs can be distinguished more easily
- modified the "About"-dialog: inserted a CoMIRVA-logo and additional information about affiliation and e-mail address of the author
- new "Circled Bars Advanced"-visualization where the user can specify some parameters (number of data items to be shown - displays only the n data items with lowest distances, sort by distance/meta-data alphabetically)
- restructured menu "Data"
- new function to rename (meta-)data matrices (Data/Data Matrix/Rename... and Data/Meta-Data/Rename...)
- the main window is now centered after CoMIRVA was started
- vectorizing a data matrix by row or by column automatically names the newly created data vectors according to the values of a meta-data instance if a meta-data instance with the same number of items is selected
- key shortcuts are available for all menu entries now (mnemonics)
- new network visualization based on a random graph which uses a probabilisitc model
- the last settings of configuration dialogs are remembered and reloaded the next time the dialog is opened
- bug in SOM-grid fixed: now the size of the SOM-grid coincides with the size of the SDH in the SDH-visualization (independently of the window size)
- the co-occurrence/page count retriever does not use the Google Web API anymore since the results were quite poor; now it provides a general framework for querying arbitrary search engines directly via HTTP