hhl is a library for Hidden Markov Models (HMM) and Hierarchical HMMs (HHMMs) in Ocaml. Both are probabilistic sequence models. Please refer to [Rabiner, 1989] and [Fine, 1998] for more details about these models.
hhd is a designer that allows to design by hand models and save them quickly. You can zoom the screenshot on your right by clicking on it.
The requirements below should be met by several linux or BSD recent distribution, as packages. For instance under debian or ubuntu, type as root: apt-get install graphivz liblablgtk2-ocaml-dev libgnomecanvas2-dev.
This software should work under Windows, but this was not tested.
Please consult the generated documentation.
Usage should be easy. The only important thing to do is to set the cardinal (i.e. the size of the alphabet on which observed and generated sequences are written) before constructing the model. Below is the list of buttons and entries:
| Card: [entry] | Sets the cardinal (size) of the alphabet to entry |
| New | Creates a new model |
| Open / Save | Open/save a model from/to file |
| Add state to: [entry1] pi: [entry2] | Parameters for a new state as child of node entry1 with transition probability entry2 |
| Internal / Normal | Create the new state with parameters above, and of type Internal/Normal |
| Save Dot (HHMM) | Save the model in .dot file as HHMM |
| Transition from: [entry1] to: [entry2] probability [entry3] Make | Add or modify a transition (vertical or horizontal) from entry1 to entry2 with probability entry3 |
| Convert internal: [entry] Perform | Make an equivalent model, in which internal state entry is removed |
| Flatten | Convert the HHMM to a HMM |
| Save Dot (HMM) | Save the model in .dot file as HMM |
hhl and hhd are copyrighted by Henri Binsztok
both hhl and hhd are licensed under the GNU General Public License.
New version released Fri May 23 11:03:42 CEST 2008: Corrected a bug in matrix.ml, thanks to Benjamin Snyder. If you used previous version, you need to update.