Content uploaded by David Miall
Author content
All content in this area was uploaded by David Miall
Content may be subject to copyright.
Personal
Librarian:
a Tool for the Literature
Classroom
D. S. MIALL
University of Alberta, Canada
Abstract
A text retrieval program for IBM-compatible microcomputers,
Personal
Librarian (formerly known as SIRE), is reviewed for
its relevance to undergraduate study of literary texts in the
classroom. In addition to supporting the standard text re-
trieval functions, with searches for words and collocates,
support of Boolean operators, and a proximity operator for
collocates, retrieved documents are rank ordered according to
their estimated relevance to the query.
Personal Librarian
also
allows searches that use a whole document as a query term
and an Expand command which produces a thesaurus of
collocates unique to the text base being studied. In this way
the program enables the user to progress from word occur-
rences to the level of themes and ideas. These facilities will
make the software particularly useful in supporting classroom
discussion of texts, where suggestiveness and a fairly
intui-
tive set of search pathways are more important than close
linguistic analysis.
In introducing students of English and other literatures
to the computer as a medium for the analysis of texts,
there is a need for computer software that will enable
rapid searching and retrieval of a text base. A range of
text retrieval packages has been created, and many are
now available for a microcomputer, but so far few have
found a regular place in the undergraduate curriculum.
For classroom use, where the focus of attention is on the
content of a text rather than on its linguistic or computa-
tional dimensions, there is need for a program that is
simple to understand and use, but powerful enough to
enable different search strategies to be implemented and
the results of searches to be viewed immediately on
screen. Personal Librarian, a program available for
IBM-compatible microcomputers, meets this basic need;
it also provides some facilities to support text retrieval
which are unique, and specially suited to work in the
humanities classroom.
Until recently the only program explicitly designed to
meet such a need was WordCruncher. a brief account of
the advantages of WordCruncher will help to set Per-
sonal Librarian in context. Almost no prior training is
required to make effective use of WordCruncher at a first
attempt. A student introduced to the program can
choose a text (or 'book') from the
'bookshelf,
shown
when the program is first called up, can go from there to
a screen containing an alphabetical list of all the words
in the text, and by choosing a word can call up all
instances of the use of the word in a window where the
search word is highlighted. Given that students ap-
proach a text with some understanding, with questions
they wish to ask, WordCruncher provides a highly
efficient tool for focusing on the vocabulary of a text.
Correspondence: David S. Miall, Department of English, University
of Alberta, Edmonton, Alberta, Canada, T6G 2E5.
Literary and Linguistic Computing, Vol. 5, No. 1, 1990
The program is flexible enough to support many differ-
ent types of enquiry: in addition to searching on single
words, words can be searched for in a variety of
combinations; the distribution of words across a text
base can be reviewed; and a concordance or index can
quickly be compiled and printed out. With these facili-
ties,
WordCruncher provides a tool offering immediate
support to classroom discussions of literature or to
students' work on their own projects.
At a basic level Personal Librarian offers the same
facility: the user can immediately find all occurrences of
a given word, or search for two or more words in
combination. Beyond this, however, Personal Librarian,
while lacking the word distribution and concordancing
facilities of WordCruncher, does offer two methods
which are particularly fruitful for the study of literary
texts.
First, on completing a search, the program will
rank order the sections of a text that it has retrieved;
and, secondly, the program has a routine for automati-
cally locating significant co-occurrences of
a
given search
word. These methods, which no other microcomputer
text retrieval package offers, can, if used judiciously, give
the user a better grasp of the conceptual structure of a
text: they enable searches to reach beyond the level of
words towards themes and ideas. In this respect Personal
Librarian puts the computer closer to the level at which
debate about a text normally takes place in the literature
classroom, and it gives the student a tool which will
stimulate exploration and discovery.
I now look in more detail at the operation of the
program, its rationale and history, and note some of its
limitations. I will deal primarily with the DOS-based
version of the program: more recently a version operat-
ing under MS-WINDOWS has been released which
offers some additional advantages. A version for the
•Macintosh is also said to be inv.y.inent.
Personal Librarian derives from an earlier research
project at Syracuse University on text data bases, in
which the primary aim was to develop a method for
automatically assessing the relevance of documents to a
search query. In standard information retrieval, as in
most data bases, a document is retrieved if it matches the
terms in the search query, and documents will be dis-
played in the order in which they occur in the raw data
base.
In Personal Librarian a weight is given to the
search terms in a retrieved document, based on the
number of times the terms occur in the document and on
the length of the document. The documents are then
rank ordered, so that those documents that are most
relevant to the query will be seen first. The algorithms
for this were developed during the 1970s in an earlier
version of the program called SIRE.
A study carried out to help validate the ranking
method, involved asking physicists to frame statements
© Oxford University Press 1990
based on their research interests; search terms derived
from the statements were then used to retrieve citations
and abstracts of articles. The physicists rated the re-
trieved articles on how relevant they were to their
interests; the resulting rank orders correlated at a highly
significant level with the rankings produced by the
program (Noreault et ai, 1977). The search procedure
thus provides a more efficient and rapid identification of
relevant documents.
The ranking of documents is the most immediate and
attractive feature of the program: it requires no action
from the user. On starting the program the user is
presented with a screen showing a menu of options on
the top line, and a bar at the bottom of the screen where
the user enters commands. There are no other menus.
All commands can be entered by a single letter, followed
by the relevant arguments. The most recent search query
is shown below the command line. The normal mode of
operation of Personal Librarian, however, is the search,
and this requires only the entry of search terms. Thus a
single word entered on the command line is enough to
initiate a search. For more complex searches Boolean
operators are supported, multiple levels of brackets,
right-hand truncation (called 'stemming' in the language
of Personal Librarian), and wildcards. If the user is
searching for collocates the distance of the words can be
controlled by the proximity operator: thus 'sleep w/20
dreams' will locate occurrences of the target words
within a span of twenty words.
The standard result of a search is a list of the
documents occurring first in rank order: the first two
lines of one or more of the fields of each document is
shown, together with a note of the total number of
documents retrieved at the foot of the screen. An
example list is shown in Fig. 1, derived from the poetry
of Coleridge. Following the list each document itself can
be viewed on screen: nineteen lines are shown at once,
with the search word or words highlighted (see Fig. 2);
paging through the document, if it consists of more than
nineteen lines, is done by pressing the Enter key. Press-
ing the down or up arrow keys will jump to the next or
previous document in rank order. An alternative to the
rank order list is the bar chart (see Fig. 3): a histogram of
estimated relevance is shown, in which the first docu-
ment in the rank order scores 100% and subsequent
documents are scaled accordingly. This provides a useful
impression of the effect of the search command. The bar
chart shows immediately if the search command was too
broad or too narrow and should be issued again in
modified form.
Searching is extremely rapid since, like WordCruncher,
texts are pre-indexed. The size of the index can be
controlled by use of a stopword list, which can be written
by the user (a default list is provided with the program,
but this will sometimes be too extensive for literary
purposes). Personal Librarian enables documents to be
organized by fields, as in a data base: up to 256 fields per
document (or record) can be specified, and field lengths
are unlimited. This makes it suitable for a number of the
requirements of literary text analysis, although for a
straight literary text base one field per document may be
all that is required (for example, each poem, or section of
a longer poem, in a collection of poetry would form one
document). The field names are inserted as ordinary text,
taking up one line and enclosed in hyphens, for instance:
'-Author-'. The end of a document is marked by '-end-'.
The use of the hyphen as field name delimiter is a minor
irritation, since literary texts often include a dash or
hyphen at line endings; but a word processor can be used
to find these and add another character outside each
hyphen to 'protect' it.
Taking a specific example, I found that indexing a file
665K in length, the complete poetry of Coleridge, on an
80286 computer with a hard disk took 12 minutes (the
poetry was divided into documents consisting of one
field). By contrast, WordCruncher, which admittedly
requires three levels of coding, took 55 minutes to index
the same material. Retrieval is almost immediate: within
4 seconds of entering a one word search term the first
nine document titles are listed on screen. A more com-
plex search adds another 1 or 2 seconds. Obtaining the
full text of the opening of the first document then takes
Get Display Jump
> <
Quit Help
Set
List
Bar
Alpha Expand Past SOrt EDit Write
Rank
1) title:
|L145
2) title:
Christabel SECTION
7
Outside
her
kennel,
the
mastiff
old
Dejection:
An Ode
SECTION
7
But hush! there
is a
pause
of
deepest silence!
3) title:
An
Effusion
at
Evening SECTION
2
Propitious Fancy hears
the
votive sigh
—
Elegy: Imitatated from
one of
Akenside's Blank-verse Inscriptions4) title:
IPg.69
5) title: Songs
of the
Pixies SECTION
4
When Evening's dusky
car
6) title: Happiness SECTION 4
'Tis thine
the
converse deep
to
hold
Pain
. ._. .
7) title:
!Pg.l7
8) title:
9) title:
[Pg.89
Lines:
On an
Autumnal Evening SECTION
2
O dear Deceit!
I see the
Maiden
rise,
To Lord Stanhope
on
Reading
his
Late Protest
in the
House
of
Lord
Query
6
Retrieved
67
Enter command>
SLEEP*
OR
ASLEEP
DB:
\pl\coleridg\coleridg
Fig. 1. Rank ordered list of documents retrieved.
20Literary and Linguistic Computing, Vol. 5, No. 1, 1990
get Display Jimp > < Quit Help Set List Bar Alpha Expand Past 90rt EDit Write
-title-
To a Friend [Charles Lamb] together with an Unfinished Poem
|Pg.78
JL1 Thus far my scanty brain hath built the rhyme
Elaborate and swelling: yet the heart
Not owns it. From thy spirit-breathing powers
I ask not now, my friend! the aiding verse,
|LS Tedious to thee, and from thy anxious thought
Of dissonant mood. In fancy (well I know)
From business wandering far and local cares,
Thou creepest round a dear-lov'd Sister's bed
With noiseless step, and watchest the faint look,
|L10 Soothing each pang with fond solicitude,
And tenderest tones medicinal of love.
I too a Sister "had, an only Sister —
She lov'd me dearly, and I doted on her!
To her I pour'd forth all my puny sorrows
|L15 (As a sick Patient in a Nurse's arms)
Hit Enter for more To a Friend [Charles Lamb] together with an Unfinished Po
Query 8 Retrieved 31 Doc. # 113 Rank 2
Enter comnand>
SISTER DB: \pl\coleridg\coleridg
Fig. 2. Displaying retrieved text, with search words highlighted.
get Display Jimp > < Quit Help Set List Bar Alpha Expand Past SOrt EDit Write
Estimated Relevance
100!
75
50
25
1 5 10 15 20 25
Query 6 Retrieved 67
Enter conmand>
SLEEP* OR ASLEEP •
30 35 40 45 5055Document'8
Rank
DB:\pl\coleridg\coleridg
Fig. 3. Bar chart showing relevance of retrieved documents.
only another second, and paging of additional docu-
ments takes place with no perceptible pause. During
retrieval the list can be recovered by pressing
'1',
and the
bar chart can be seen by pressing 'b'. A document can
also be seen out of rank order by entering 'd' and the
rank order number, eg. 'd 23'. It is also possible to page
through documents outside the retrieval list by using the
left or right arrow keys to move backwards or forwards
in the text base from a given document.
More sophisticated searches can be performed by
using a document itself as a search term. This has the
effect of making all the significant words in the docu-
ment into search words. Documents similar to the target
document will be retrieved and rank ordered; each will
show some degree of commonality in its themes and
concerns. This search takes longer to complete: using a
Coleridge poem of twenty-three lines as the search term,
Personal Librarian required 30 seconds before producing
the rank order list. As a classroom technique for widen-
ing the range of poems or other documents students will
consider, however, this facility has considerable poten-
tial.
The most unusual aspect of Personal Librarian is its
'Expand' facility. The program will search for significant
collocates to a given word or set of
words,
and produce a
list of forty-eight words on screen. Again, this search
takes time: length of time appears to depend in part on
how many occurrences of the target word or words
Personal Librarian has to take into account. The Expand
command on one word, that occurs seventeen times in
the Coleridge file, took 117 seconds to produce the list of
collocates. Expanding on this word and one other word,
a co-occurrence of which there are only two examples,
took only 25 seconds. The results of the expand com-
mand, however, can often be illuminating. The effect is a
type of thesaurus entry for the target word or words,
Literary and Linguistic Computing, Vol. 5, No. 1, 199021
unique to the text base being consulted. As such it can
often reveal special properties of the vocabulary and
assocations of the author under study. For example,
Coleridge lost his only sister at an early age: this event is
referred to in several poems, and in other poems the idea
of sisterhood holds a special place for Coleridge. The
Expand command with 'sister' as the key word indicated
typical Coleridgean concerns: gloomy, distress, woes,
pain,
and agony, but also a number of occurrences of
love,
shown here in its 'stem' form, lov (see Fig. 4).
Again, the Expand facility has considerable potential
for focusing and extending students' researches on a text
base.
Unfortunately, the authors of Personal Librarian
provide no information about the Expand algorithm. It
is not clear on what basis the collocates are selected, and
the list shown on screen does not seem to occur in a
systematic order; a collocate occurring sixteen times may
be listed next to one occurring only twice. Thus the
results of an Expand command should be treated with
some caution.
In the DOS version an Expand list will disappear from
screen as soon as another command is issued. In the
WINDOWS version the list can be kept on screen or
recalled as a reference point while the collocates it shows
are examined in more detail (this is also true of the other
displays offered by Personal Librarian, such as the bar
chart).
Several other options are available from the
menu at the top of the screen. The Alpha command,
issued with a target word, shows nineteen lines of the
dictionary of words in the text, centring on the target
word and including the number of occurrences (it is
similar to the 'Select Word' screen of WordCruncher).
The Set command shows a summary of the current
default options, for example whether the list or the bar
chart is shown first following a search. Previous queries
are remembered by Personal Librarian, and these can be
listed on screen by issuing the Past command. This is
helpful if the user wants to repeat a previous query, add
another term to a query, or combine two queries, since
previous queries can be run again merely by typing 'Q_'
and the number of the query. A set of retrieved docu-
ments can be sorted by entering the Sort command,
together with the name of the field on which the
documents are to be sorted.
Personal Librarian also provides several methods for
extracting documents if the user wants to print out the
results of a search or edit the material with a word
processor. Specified documents can be dumped to a file,
called DOCOUT, which can be edited through a word
processor after leaving Personal Librarian. Alternatively
the Write command, which is toggled on and off by
pressing 'w', will dump the current document on screen
to the DOCOUT file. In addition, an Edit command is
provided which suspends Personal Librarian and calls up
whatever word processor has been specified in the
installation routine. On exiting the word processor,
Personal Librarian resumes at the point where it was
left.
In summary, Personal Librarian provides an efficient
context for searching large text bases, with a fairly
intuitive set of pathways to assist exploration following
an initial search. Simple searches can be carried out with
little knowledge or training. An on-line help facility
provides some guidance for the more elaborate searches
and other commands. My version came with a small
sample file called Movies, which was used as the training
example in the manual. The manual
itself,
provided in an
A4 ring back binder, is divided into two main sections: a
beginner's guide and reference, and an 'administrator's
guide' which shows how to prepare a file for Personal
Librarian and how to index it. The tutorial section
assumes no knowledge of data bases and provides a
useful introduction to the program. I found some inac-
curacies in the manual: the syntax of some of the
commands appears to have changed since the manual
was last printed. Like most commercially produced
manuals, however, this one is inappropriate for use with
humanities students, who require a tutorial built around
a more relevant set of examples.
The Coleridge file I installed with Personal Librarian,
which was 665K in size, took a total of 560K for the
twelve other index and system files needed by Personal
get
Display
SIblfcK
POISON
ROLL
NATIVE
LOV
TEAR
SHRIEKS
INDIGNATION
THRILLING
MOURN
THANKS
HELD
ANGUISH
BOWL
MONODY
MERCY
Jump
> <
Quit
BADE
COT
DASH
MARK
BROW
CHALICE
SLAU
KEEN
BARDS
DEATH
MELT
INSULT
RAIS
WOE
SAME
AGONY
Enter command>
Help
Set
List
Bar
Alpha Expand Past SOrt EDit Write
LISTEN
TOLD
GLOOMY
WOES
DISTRESS
PANG
EDWARD
FAINT
PAIN
CEASE
WAYS
COMMAND
ALREADY
SULLEN
STORY
CHATTERTON
DB:
\pl\coleridg\coleridg
Fig. 4. Expand list for 'sister'.
22Literary and Linguistic Computing, Vol. 5, No. 1, 1990
Librarian. Larger files with more redundancy would
probably need a lower proportion of additional disk
space (the developers quote an overhead of 55% in
discussing the earlier program, SIRE: Koll, et al., 1984).
WordCruncher took marginally more space to index the
same file. Users of Personal Librarian should note,
however, that during the indexing process a temporary
file is created which appears to be at least twice the size
of the source file, thus a large margin of empty disk space
will be needed when indexing a new file.
In several respects Personal Librarian offers a more
productive environment for studying texts than its
closest rival, WordCruncher. The search by document,
and the Expand facility, both provide suggestive ways of
tracing further documents at the level of themes and
ideas,
which will often be of value in the literature
classroom. Since the basis on which Personal Librarian
undertakes an Expand search is unclear, detailed study
of collocations will still require a scholarly tool such as
the Oxford Concordance Program; but for the support
of seminar and project work on literary texts Personal
Librarian offers a valuable new tool which can be
strongly recommended (Friedman et al., 1990).
Personal Librarian requires an AT or above, with hard disk
drive. Versions are also available for other operating systems
and mainframes, and a Macintosh version is expected shortly.
In the UK Personal Librarian is available from Systematic
Upgrade, 58-60 Edward Road, New Barnet, Herts., EN4 8AZ.
Telephone 01-449 9699. Price for a single user: £695. In the
USA: Personal Library Software, 15215 Shady Grove Road,
Suite 204, Rockville, Maryland 20850. Telephone 301-926-
1402.
Price: $895.
References
Friedman, E. A., McClellan, J. E., and Shapiro A. (1990).
'Automated Text Retrieval in Humanities Courses'. In D. S.
Miall (ed.), Humanities and the Computer: New Directions,
(ed.) 103-12. Oxford: Oxford University Press. Contains an
account of a course based on Personal Librarian.
Koll, M. B., Noreault, T., and McGill, M. J. (1984). 'Enhanced
Retrieval Techniques on a Microcomputer', National Online
Meeting: Proceedings—1984, 165-70. Medford, NJ:
Learned Information Inc.
Noreault, T., Koll, M., and McGill, M. J. (1977). 'Automatic
Ranked Output from Boolean Searches in SIRE', Journal of
the American Society for Information Science, 28, 333-9.
Literary and Linguistic Computing, Vol. 5, No. 1, 199023
