OpenKey Notes from 2/23/03
Members present: Bryan Heidorn,
Steve Seiberling, Peter White, Alan Weakley, Evelyn Daneil, Sarah Reuning
Project
members reviewed the two lists of critical features for interactive keys and
discussed the current and possible future capabilities of BIBE. Notes appear in
the list texts in blue below.
Botany Features (DRAFT)
Critical, Very Important, and Desirable Attributes for the OpenKey Polyclave
This is an attempt to list
potential attributes for the OpenKey polyclave in categories of differing
importance. There are a number of other
useful attributes that could be incorporated into a polyclave that I’ve not
included here based on my sense of what is likely to be most useful for our
intended audience and what the grant proposal seems to call for.
This information is partially
derived from, M.J. Dallwitz, November 2, 2000, A Comparison of Interactive Identification Programs,
http://biodiversity.uno.edu/delta/www/comparison.htm
Critical Attributes (these features are necessary to have a reasonably functional polyclave and/or comply with basic provisions in the grant proposal)
Unrestricted character use – Characters from the full list may be
chosen by the user in any order. (One
possible limitation on this arises should we wish to incorporate character
dependencies [see below] in the polyclave, e.g., you couldn’t choose a cone
character if you’d already chosen flower characters.) BIBE
does this now. In the future, should BIBE suggest character order like the
“best characters” feature of other keys?
This question is addressed below.
Error tolerance – The program permits taxon matches even when some number of selected character states do not fit the values for the particular taxon. This allows the identification process to continue even when errors have been made by the user or there are errors in the data. BIBE does this now. The program can weight character values when entered. Confusing characters will not be weighted a high as straightforward ones. The program can be written so that the user can turn off this feature.
The team
also decided that to create two interfaces for the key, one technical, one
simplified, which would be easy to implement.
Choosing multiple character states – The user can specify more than one character state for a given character, due to user uncertainty or reflecting observed variation in a specimen. BIBE does this now.
Retaining unknowns – Taxa for which a character is not recorded are retained when that character is used (with any character state value) in an identification. BIBE can do this. The team agreed to put “unknown” in the XML files, so the user will see that the value is unknown, but the database will interpret “unknown” as all possible values are true. Thus, the other character states that the user enters will pull the taxon record.
Inapplicable versus unknown – The polyclave can distinguish inapplicable values from unknown values, e.g., the program can determine that none of the flower characters apply to conifer species in the database. The team decided not to code for inapplicable. BIBE will be silent here.
Character notes – Text providing interpretative information on each
character is conveniently available within the system. By character notes the team means both
definitions and instructions on how to determine character values. BIBE will
have these. We will add a new field
called “character notes,” which will be an attribute in DTD.
Character illustrations – The user can readily display
illustrations of characters and character states during the identification
process. These are not currently in BIBE, but team
members decided that these would be good to add and can be implemented easily.
Internet capability – The program is accessible and easy to use over the internet.
BIBE is
Internet accessible now.
Very Important Attributes (we should include these features if at all possible)
Character deletion and changing – Permits the user to remove characters used in an identification or change their values without having to restart the identification process.
BIBE has character deletion, but not character changing right now. The team would like to add character changing.
Locating errors – When error tolerance has been incorporated into the polyclave (see above), a taxon which differs by some number of selected character states can remain in contention in an identification. The program should be able to display which taxa do not perfectly match the selected character states and what the differences are. BIBE does this.
Alternative simplified terminology – The user can choose character state values from a simplified set of terms. The polyclave then finds approximate matches with the more detailed character states recorded for taxa in the database. BIBE can support this. The team decided to make two data sets. We will start with the more complete record and strip out characters to make a simplified record. This would be more work for the data enterer, but the database can handle it.
Numeric characters – Numeric characters can be used directly (without converting to multistate by dividing the values into ranges). BIBE will probably not be able to handle this. The team has already decided to use absolute minimum, usual minimum, usual maximum, absolute maximum. Can code to match anything with in the max-min range.
Text characters – The program permits storing and searching free-text information about taxa contained in the Special Diagnostic Characters field or in general taxon descriptions. BIBE does this.
Glossaries – Readily accessible definitions for botanical terms (characters, character states and other used terms) are available. At a minimum the user should have access to definitions of character states when choosing among them. (This approach could be incorporated into character notes, as described above.) Preferably, the user could access definitions of terms through a simple means wherever the terms occur, including from within the definitions themselves. BIBE does this. The team decided to make all glossary terms that appear in definitions clickable. We will also have a link to the full glossary on the site.
State selection from illustrations – The user can select character state values from illustration screens during identification. BIBE can do this. We will present metadata with picture and present a way to chose values.
Taxon illustrations – The user can display illustrations of taxa. BIBE does this as part of the taxon description. We will need extra fields for herbarium sheets and field/seasonal photos. We will list seasons under habit, “habit_spring”
Flexible display of illustrations – Illustrations of any size can be scaled, scrolled, repositioned, displayed simultaneously, and tiled or cascaded. BIBE does not do this. Right now imageas are a fixed size at 150x150 pixels
Text with illustrations – Text can be associated with illustrations (instead of being built into the illustrations). BIBE does this.
Searching the taxon names – The program is able to finding text strings in the taxon names, synonyms and common names. BIBE does this.
Descriptions from the data – The program displays descriptions generated directly from the data used in identification. [Not completely sure about this one.] This seems to be the same thing as the taxon sheet.
Text files attached to taxa – The polyclave includes text files containing displayable information about the taxa. BIBE does this.
Online help – The program includes context-sensitive, built-in help. Bryan has a team of three writers, but they are not Botanists. There is currently a video tutorial, but the English needs to be redubbed.
Desirable Attributes (these features would be valuable to incorporate into the polyclave, but it would probably be O.K. if they had to wait.)
Best characters – The program provides advice on the most helpful
characters for use at any stage of an identification. The
team decided to break best characters into two features: 1) what order should
the characters be displayed in? 2) what features should be dynamically
reordered after the user selects a character value? The second option cannot be
implemented in BIBE now, but could be added in some form later.;l
Removing redundant characters – The program removes from the list of available characters those that cannot separate the remaining taxa in an identification. This feature depends on the best character feature being present.
Character dependencies – The polyclave contains information
specifying that some characters are inapplicable when other characters take
certain values. The inapplicable
characters then can no longer be selected by the user. BIBE
not now do this; it must still allow for errors.
Probabilistic identification (fuzzy logic) – The program associates given character states with other related states for the same character, based on assigned probabilities (e.g., serrulate and serrate would be associated with a high probability). This information is used by the program to make probabilistic or partial matches with the appropriate taxa. BIBE can do this via a similarity matrix. The team decided to create a “present by misinterpretation” tag.
Subsets of Characters – The user can specify subsets of characters to which subsequent operations will be restricted. BIBE does this by grouping. The user can look just at leaf attributes if they only have a leaf.
Taxon illustrations by subject – The user can select illustrations of taxa by subject (e.g., growth habit, flowers, leaves, etc.).
The team discussed whether we want to display results in one or two dimensions. We can give the user the choice between viewing results in a ranked list or viewing them in a graphical 2 dimensional mode.
The team was only able to briefly look at the critical system features, not the very important.
System Attributes 1.1
Critical, Very Important, and Desirable System Attributes for the OpenKey Polyclave
This document lists potential
system attributes for the OpenKey polyclave in categories of differing
importance. Definitions of Critical,
Very Important, and Desirable system attributes are adapted from
Steve’s Botany Attributes document.
Critical Attributes (these features are necessary to have a functional polyclave. Without these features, the system ceases to function or no longer meet the goals set out in the grant application.)
Data storage – the system must have a way to store data. Data includes the plant character states associated with each tree species and images of tree species BIBE does this.
Data Entry Interface – an interface that provides a way to add data to the system
BIBE does this via webforms.
User Interface – an interface that provides access to a mechanism that will search the stored data
Search mechanism – something that links user input to the stored data and retrieves the appropriate information BIBE does this.
Public Access – The key should be freely, easily, and publicly available via the Internet. BIBE does this.