Identification Software

[Mike Dallwitz]

                                                                 20 March 1995

> From: stephen gough <sgough at S850.MWC.EDU>
>
> To supplement traditional methods of identifying vascular plants (mostly
> trees and shrubs) in a Plant Ecology course, I would like to construct a
> software-based system for common items. I don't need the rigor of a
> DELTA-based system; rather I would like something that uses simple
> descriptive characteristics (perhaps a list of characteristics up front) and
> then determines one or more possible "hits," with the option to view line
> art or photographs on screen. Does anyone know what software I might use to
> construct something like this? It should be PC-compatible, and operation
> under Windows would be nice.

I'm not sure about the intended implications of the word `rigor' here. I am
flattered that it has been closely linked with DELTA, and indeed, many users
have commented to me that using DELTA has improved the quality of their work.
However, let me reassure you that it is quite possible to do sloppy work with
DELTA software (as with any database system), should you so wish.

I suppose that what you really want is a rigorous system that is easy for
students to use. There are four main aspects to ease of use for the end user
of an interactive identification system: (1) easily understood characters; (2)
accurate data; (3) the ability of the software to record and use effectively
the necessary data types and structures; and (4) the user interface. The first
two are the responsibility of the author of the data. The DELTA coding format
and our interactive identification program INTKEY are, I believe, unmatched in
the third area. However, some (perhaps most) potential INTKEY users seem to
feel that the user interface of the current MS-DOS version is too difficult.

In about a week we will be publicly releasing on the Internet a MS-Windows
version of INTKEY with a greatly improved user interface (and also some
increased functionality, particularly in relation to images). At program
startup, the user chooses from three modes of operation. The `simplified' mode
is strongly recommended for beginning or casual users of the program. We
believe that this mode can be used without reference to any documentation.
However, full instructions (190 words) for identifying specimens and
displaying descriptions are automatically displayed in a window, and the user
is discouraged from closing it.

Although `simplified' mode is extremely easy to use, it is still underpinned
by the powerful data-handling capabilities of INTKEY. A few months ago I
posted a list of features which I consider desirable or essential for an
interactive identification program. (There was no reply. Is this because of
apathy, because it was all obviously true, or because it was too ridiculous to
be dignified by a reply?) Here is the list again, with the features available
in `simplified' mode indicated by *.


    ----------------------------------------------------------------------


            Desirable Attributes for Interactive Identification and
                         Information Retrieval Systems

                       M. J. Dallwitz  20 December 1994


* Error tolerance. The ability to reach a correct identification after errors
have been made, or if there are errors in the data.

* Unrestricted character use. The absence of restrictions on the order in
which characters can be used (apart from restrictions imposed by character
dependencies - see below).

* `Best' characters. Whether the program can advise on the most suitable
characters for use at any stage of an identification. `Partial' indicates a
lack of flexibility in this area, usually owing to the recommendations being
built into the data, as in a key or rule-based expert system.

* Multiple state selection. Whether the user can specify uncertainty by
entering more than one state value, or a range of numeric values.

(*) Character deletion/changing. Whether characters used in an identification
can be removed, or their values changed. `Partial' indicates that removal is
possible only in the reverse order of use. (Deletion not available in
simplified mode.)

* Character weighting. Whether character weights can be used in the
calculation of `best' characters. `Partial' indicates that higher weights
always imply `better' characters, regardless of other considerations.

(*) Text characters. Whether free-text information about taxa can be stored
and searched. (Can be displayed but not searched in simplified mode.)

* Numeric characters. Whether numeric characters can be used directly (without
dividing them into ranges.

* Gaps for integer numeric characters. Whether recorded values for integer
numeric character can contain gaps, e.g. `5 or 10' distinguishable from `5 to
10'.

* Uncertainty ranges. Whether single numeric values in the original data can
be treated as ranges for identification purposes. `Partial' indicates that the
transformation is not under the control of the interactive user (as in the
ABSOLUTE/PERCENTAGE ERROR mechanisms in CONFOR/INTKEY).

* Inapplicable/unknown. Whether inapplicable values are distiguished from
unknown values.

Control of value matching. Whether the user has control over whether
overlapping, unknown, and inapplicable values are deemed to match other
values. `Partial' indicates limited control, e.g. `identification' vs.
`information retrieval' settings.

* Character dependencies. Whether the program is aware of character
dependencies, i.e., characters which are inapplicable when other characters
take certain values.

* No dependency restrictions. Whether there are restrictions on the order in
which dependent/controlling characters may be used.

Keywords. Whether there is a mechanism for referring to subsets of the
characters and taxa. `Partial' indicates that such subsets cannot be defined
by the user (i.e. are built into the system).

* Character notes or glossaries. Whether extensive text to aid interpretation
of characters can be conveniently available within the system.

(*) Information retrieval. Whether the system can be used for information
retrieval (e.g. displaying descriptions, finding all taxa which have certain
combinations of attributes).

Differences between taxa. Whether the program can find the differences between
members of a set of taxa. `Partial' indicates that the set must contain only 2
taxa.

Similarities between taxa. Whether the program can find the similarities
between members of a set of taxa.

Diagnostic descriptions. Whether the program can find diagnostic descriptions.
`Partial' indicates inability to distiguish between taxon and specimen
diagnostic descriptions, or inability to restrict the choice of characters to
those not used in the current identification.

Character-value distributions. Whether the program can display the
distribution of character values within a set of taxa.

(*) Global restriction to subsets. Whether it is possible to specify subsets
of characters and taxa to which all subsequent operations will be restricted.
(In simplified mode, possible in initialization file, not interactively.)

* Local restriction to subsets. Whether it is possible to specify subsets of
characters and taxa for the operation of a single command.

Searching the character list. Finding text strings in the character list.

* Illustrations. Whether illustrations of character and taxa can be displayed.

* Flexible display of illustrations. Whether illustrations of any size can be
scaled, scrolled, repositioned, and displayed simultaneously.

* State selection from character illustrations. Whether character state values
can be selected from illustration screens during identification.

* Text on illustrations. Whether text can be superimposed on illustrations
(instead of being built into the illustrations).

* Missing illustrations. Whether a package containing illustrations can be
used without them.

* Import DELTA format. Whether DELTA-format data can be used to create the
interactive system.

Export DELTA format. Whether DELTA-format data can be exported from the
interactive system.

* Links with description writing. Whether publication-quality descriptions can
be generated from the same data that are used to construct the identification
system.

* Links with key generation. Whether conventional keys can be generated from
the same data that are used to construct the identification system.

* Links with classification. Whether cladistic and phenetic analyses can be
carried out from the same data that are used to construct the identification
system.

(*) Command files or macros. Whether there is a mechanism for storing and
repeating a series of operations. (In simplified mode, execution only.)

Log files. Whether it is possible to create a file showing the history (input
and output) of a session.

Data output. Whether it is possible to output program results in forms
suitable for input to other programs.

* Online help. Whether the program has complete, built-in help. `Partial'
indicates that the help is not context sensitive.

* External program text. Whether the program text (commands, help, messages,
etc.) is external to the program, allowing easy creation and use of different
language versions.

* Maximum field lengths. Limits (if any) on lengths of text and other fields
(e.g. taxon names, text of characters, character notes, number of character
states).

* Maximum size of data. Maximum number of characters and taxa.

* Memory requirements. Program memory requirements (including dependence on
data size, if applicable).

* Execution speed. Execution times of representative operations on a
reasonably large data set (e.g. 200 characters, 400 taxa).


    ----------------------------------------------------------------------


I claim that, in simplified mode, the Windows INTKEY will outperform any other
identification software both in ease of use and accuracy of results. If, after
testing the program, you think otherwise, please give your reasons on this
list. The program will be released with a toy set of data on butterflies and
moths. These data were created for our open days, and can be used successfully
by small children.

Mike Dallwitz                                  Internet md at ento.csiro.au
CSIRO Division of Entomology                   Fax +61 6 246 4000
GPO Box 1700, Canberra ACT 2601, Australia     Phone +61 6 246 4075