Roadmap » History » Version 7
Version 6 (Jeremy Gow, 2012-11-01 04:26 PM) → Version 7/34 (Jeremy Gow, 2012-11-02 10:10 AM)
h1. Roadmap
The immediate goal is to have a stable release version that is compatible with sbcl 1.1
* To fix: ground-truth-segmenter class missing during compilation
* To fix: non-terminating method calls in amuse interface.
* Minor fixes for boundary cases thrown up by Turkish Makam data.
* Ensure Conklin example etc. still work.
h2. Short-term
New A simple configuration script: remove the need to edit paths in source code.
Some basic viewpoints:
* cents - a higher resolution representation benchmarks to ensure stability of pitch. [Needed ASAP for Makam data.]
* comma (implemented?)
* metrical contour future development versions.
Remove dependancy on amuse, so the system is data agnostic.
Viewpoint selection:
* Adding viewpoint weights to output.
* Optionally specify: start point for search, min-links
* More flexible way for user to specify constraints on viewpoint search:
** Define labelled viewpoint classes
** Pairs/triples of labels/wildcards specify acceptable combinations
** User provides whitelist or blacklist spec
New basic viewpoints:
* cents - a higher resolution representation of pitch.
* comma
* metrical contour
Efficiency:
* Check/extend caching of models etc.
* Use sampling to estimate mean IC during VP selection.
Minor:
* Some basic benchmarks to ensure stability of future development versions.
* Have all viewpoints displayed in describe-dataset.
Remove dependancy on amuse, so the system is data agnostic.
h2. Mid-term
Viewpoint selection:
* Optimise based on match with existing IC values.
Allow user to specify structure of model.
* Determine order in which distributions are combined.
* Specify weights for particular combinations, e.g. weighted viewpoints, or weighted memory stores.
* Multiple memory stores.
* Specify alternative context strategies.
* Provide some prepackaged models, e.g. the current model structure.
Allow models to use predictive information (PI), expected PI and PI rate (as analogs to IC, entropy and entropy rate respectively).
Predict more than one basic viewpoint
h2. Long-term
Hierarchical structure.
Parallel implementation.
The immediate goal is to have a stable release version that is compatible with sbcl 1.1
* To fix: ground-truth-segmenter class missing during compilation
* To fix: non-terminating method calls in amuse interface.
* Minor fixes for boundary cases thrown up by Turkish Makam data.
* Ensure Conklin example etc. still work.
h2. Short-term
New A simple configuration script: remove the need to edit paths in source code.
Some basic viewpoints:
* cents - a higher resolution representation benchmarks to ensure stability of pitch. [Needed ASAP for Makam data.]
* comma (implemented?)
* metrical contour future development versions.
Remove dependancy on amuse, so the system is data agnostic.
Viewpoint selection:
* Adding viewpoint weights to output.
* Optionally specify: start point for search, min-links
* More flexible way for user to specify constraints on viewpoint search:
** Define labelled viewpoint classes
** Pairs/triples of labels/wildcards specify acceptable combinations
** User provides whitelist or blacklist spec
New basic viewpoints:
* cents - a higher resolution representation of pitch.
* comma
* metrical contour
Efficiency:
* Check/extend caching of models etc.
* Use sampling to estimate mean IC during VP selection.
Minor:
* Some basic benchmarks to ensure stability of future development versions.
* Have all viewpoints displayed in describe-dataset.
Remove dependancy on amuse, so the system is data agnostic.
h2. Mid-term
Viewpoint selection:
* Optimise based on match with existing IC values.
Allow user to specify structure of model.
* Determine order in which distributions are combined.
* Specify weights for particular combinations, e.g. weighted viewpoints, or weighted memory stores.
* Multiple memory stores.
* Specify alternative context strategies.
* Provide some prepackaged models, e.g. the current model structure.
Allow models to use predictive information (PI), expected PI and PI rate (as analogs to IC, entropy and entropy rate respectively).
Predict more than one basic viewpoint
h2. Long-term
Hierarchical structure.
Parallel implementation.