Technical Correspondence A Note on a Study of Cases 
A NOTE ON A STUDY OF CASES 
This note describes and illustrates a study of deep 
cases using a large sample of sentences. The purpose 
of the note is to draw attention to the value of the 
source material used for those interested in case- 
based representations of sentence meaning, and to 
indicate the potential utility of the study results. 
The purpose of this note is to draw attention to the utility 
of a specific source of data relevant to the use of case- 
based meaning representations in language processing, by 
illustrating the way we have exploited this source. 
Like many others, we have used a language analyser 
that builds meaning representations expressing semantic 
case roles; specifically, Boguraev's (1979) analyser builds 
dependency trees with word senses defined by semantic 
category primitive formulae, and with case labels, i.e., 
semantic relation primitives, on the constituents of verb 
(and some other) structures. 
Using the analyser for more varied and demanding 
purposes than Boguraev's original tests (see, e.g., Bogu- 
raev and Sparck Jones 1983) left us dissatisfied with the 
original set of case relations. We therefore carried out a 
detailed analysis of a large sample of English sentences 
to evaluate our proposals for a better-founded and more 
comprehensive set of case relations. This study exploited 
F.T. Wood's "English prepositional idioms" (Wood 
1967), which provides a careful account, supported by 
extensive examples, of the uses of English prepositions 
and preposition-like terms. For instance, 
WITHIN 
(1) Inside 
Within the house all was quiet. 
The Kingdom of God is within you. 
(2) Amongst the members of a group. 
Opinion within the profession is divided. 
(3) Inside specified bounds or limits. 
They were ordered to remain within the precincts of 
the college. 
The scholarship is open to anyone residing within fifty 
miles of the university. 
He always strove to live within his income. 
Our study was intended to establish both the justifica- 
tion for each case relation individually, by reference to a 
range of sentences, and the plausibility of the set of 
relations as a whole, by reference to the complete set of 
sentences. Looking at Wood's description of a preposi- 
tion's sense, and its accompanying illustration(s), we 
tried to assign a case label to the link between the 
sentence elements made by the preposition which we felt 
captured the essential nature of that link, at the level of 
generality represented by a set of 20-30 cases. Thus 
"location" would be the label associated with a number 
of specific space-relation prepositions, e.g., above, at, by. 
The study was primarily concerned with prepositionally- 
Computational Linguistics, Volume 13, Numbers 1-2, January-June 1987 65 
Technical Correspondence A Note on a Study of Cases 
based cases, but we considered other case instantiations, 
though not systematically. While additional cases might 
therefore be motivated by non-prepositional sources we 
have not examined, we believe that a complete set would 
have to include something like our list, to cover prep- 
ositions adequately. 
The study was a purely investigative one; we were 
interested in the legitimacy of the cases as characterisa- 
tions of sentence relations, and did not address the ques- 
tion of how the specific assignments, for the individual 
sentences, could be achieved automatically. However, as 
we had already demonstrated that a quite refined set of 
cases could be applied by Boguraev's existing analyser, 
and hive since, as practical need has arisen, implemented 
further cases, we feel some confidence in the feasibility 
of automatic assignment of the cases in the present set. 
(This of course accepts that some individual sentences 
may present considerable difficulties.) We are neverthe- 
less not offering our results here with any great claims 
about the especial novelty or merits of our case set, but 
rather as a practical contribution to the utilisation of case 
labels, justified by the large and varied sentence sample 
studied. Though domain-specific strategies may reduce 
the need to handle lexical ambiguity, especially in nouns 
and verbs but even in prepositions, it may nevertheless 
be necessary even in the domain-specific case to provide 
for distinct semantic functions in prepositions. 
Wood's complete set of sentences is substantial; we 
took one sentence for each straightforward sense defined 
by Wood's numbering, omitting items labelled "phrases" 
and with occasional modifications, e.g., to shorten 
sentences. This gave us 421 sentences altogether, for 86 
prepositions. As a result of our analysis we emerged with 
28 cases. These are listed below with notes on the types 
of source from which they may be derived, and (our) 
examples showing both prepositional and non-preposi- 
tional uses. This list is followed by illustrations from the 
Wood sample. 
In the list each case, or tag, name is prefaced by its 
• abbreviation. The meaning of the case is not explicitly 
defined, but is assumed to be ostended by the subsequent 
examples (and more particularly by the fuller sample 
derived from Wood). Possible sources for the cases 
include the lexicon and syntactic structures like embed- 
ded clauses identified by the parsing program (and in our 
analyser elements of structure identified by semantic 
pattern matching). The normal linkage marked by a case 
is between verb and noun group, but two nominal groups 
may also be case-linked, and also more than two items 
(though this is not common). In the illustrations we have 
adopted the convention of marking the head words of the 
two linked constituents by -. Thus the tag "after" labels 
the relation between left and breakfast in John left 
following breakfast, written as "John -left following 
-breakfast". In the cases where complete constituents 
like embedded clauses fill case roles, they are bracketed 
with \[ \], and the whole item is marked with -. The cases 
are alphabetically ordered here on their abbreviations. 
(ace) ACCOMPANIMENT 
dictionary: e.g., "with"; can link nominals 
John -went to the zoo (along) with -Mary. 
John -went everywhere with his -violin. 
(act) ACTIVITY 
dictionary: e.g., "at" 
John -beat Mary at -chess. 
Aloysius-beat Sebastian-running. 
(adest) ABSTRACT-DESTINATION 
dictionary: e.g., "to" 
I -reached my --conclusion. 
When heated, water -turns into -steam. 
(aft) AFTER 
dictionary: e.g., "after", program: ing-phrases, adverbi- 
als; can link nominals 
John -left after -Bill. 
John, -\[having bought the book\], -took it home. 
(ag) AGENT 
program: contents of "subj" register 
John was -hit by -Bill. 
-Malaria -killed the girl. 
(aloc) ABSTRACT-LOCATION 
dictionary: e.g., "in" 
Parry was -fixed in his -mind that the Mafia was out to 
get him sometime. 
• The four friends -shared the work between -them. 
(asour) ABSTRACT-SOURCE 
dictionary: e.g., "from" 
I -got them from -Bill. 
She -makes dresses of -silk. 
(attr) ATTRIBUTE 
dictionary: e.g., "with"; can link nominals 
The -girl in -blue was happy. 
The -girl with the pink -hat was sad. 
(bef) BEFORE 
dictionary: e.g., "before", program: adverbials; can link 
nominals 
John -left before -noon. 
The stock market -was very active ahead of the 
-Budget. 
(comp) COMPARISON 
dictionary: e.g., "as", program ?; can link nominals 
John -passed as -Bill. 
He was given -cash instead of -kind. 
Time -flies like an -arrow. 
(dest) DESTINATION 
dictionary: e.g., "to" 
John -went to -Paris. 
Bill -walked up to -John. 
66 Computational Linguistics, Volume 13, Numbers 1-2, January-June 1987 
Technical Correspondence A Note on a Study of Cases 
(dire) DIRECTION 
dictionary: e.g., "down" 
John -walked after-Bill. 
John -ran down the -hill. 
(force) FORCE 
dictionary: e.g., "of" 
The girl --died of -malaria. 
The girl -.died from an -accident. 
(goal) GOAL 
dictionary: e.g., "for", program: to-complements 
He -walked for -pleasure. 
John -went to town in order -\[to buy a shirt\]. 
(inst) INSTRUMENT 
dictionary: e.g., "by", program: complements 
The boy -caught the fish with a -minnow. 
John -went to Paris by -plane. 
(loc) LOCATION 
dictionary: e.g., "at" 
The -girl at -Smith's was not helpful about buying a 
book. 
Sebastian -felt pain in his -foot. 
(man) MANNER 
dictionary: e.g., "with", program: embedded clauses, 
adverbs 
They -ran with -speed. 
He -did it by -\[working very hard\]. 
(mobj) MENTAL-OBJECT 
dictionary: e.g., "about", program: embedded clauses, 
to-complements, that-complements 
John -planned -\[to go to London\]. 
John -threw light on the -problem. 
They -talked about -politics. 
(obj) OBJECT 
(almost a default), program: typically contents of "obj" 
register 
John -bought -books. 
John -threw -light on the problem. 
(poss) POSSESSED-BY 
program: noun phrases; only links nominals 
The -daughter of the -Mayor was blonde. 
The -pocket of my uncle's -coat was empty. 
(quant) QUANTITY 
program: noun phrases; can link nominals 
There were four -students including a -dwarf. 
It was too -long by twelve -inches. 
(reas) REASON 
dictionary: e.g., "of", "because of", program: embed- 
ded clauses 
John is -afraid of -\[being apprehended by the police\]. 
John is -happy about -school. 
(rec) RECIPIENT 
dictionary: e.g., "to", program: contents of "obj" regis- 
ter 
Susan -gave the flowers to -Jill. 
John -shot -Mary. 
(sour) SOURCE 
dictionary: e.g., "from" 
She -took some money out of her -bag. 
I -got them from -Paris. 
(state) STATE 
dictionary: e.g., "of", program: predicate adjectives 
His conduct -is -admirable. 
He wears -green -trousers. 
(subj) SUBJECT 
(almost a default), program: various places 
-Books -are nice. 
John is -fascinated by -books, especially ones 
published by private presses. 
(tloc) TIME-LOCATION 
dictionary: e.g., "at" 
They-ate breakfast at -noon. 
I-wrote my thesis during -1979. 
(tspan) TIME-SPAN 
dictionary: e.g., "throughout", program: embedded 
clauses, ing-phrases 
I -was a bus conductor for two ---days. 
John -sang -\[running in the park\]. 
Our treatment of the sample of sentences taken from 
Wood is illustrated by the two selections below. These 
show first the complete analyses for two prepositions, 
and then the complete set of sentences for two cases. 
The average number of sentences per case in the sample 
is 15, with low variation. However, "location" has 80 
sentences, reflecting the very large number of different 
space-relation prepositions there are. It is possible that 
more specific space-location cases would be desirable, 
though the right level of discrimination and particular set 
would not be easy to establish. We are, in any case, not 
suggestiffg that the use of a case tag in the representation 
of a sentence delivered by the analyser makes it unneces- 
sary to indicate the specific lexical sense of the preposi- 
tion. 
AT 
/loc/ 
/tloc/ 
/loc/ 
/dest/ 
/act/ 
/act/ 
/man/ 
/mah/ 
/force/ 
/man/ 
Mr Brown -is at the --office. 
The concert -starts at half past -seven. 
At the second -roundabout, -turn left. 
We have -arrived at our ---destination. 
The two tribes -were constantly at -war. 
She loved to watch the -children at -play. 
He -left at a moment's -notice. 
He -vaulted over the gate at a single -leap. 
I have ---called to see you at the -request of a 
friend. 
Cars are -parked at their owners' -risk. 
Computational Linguistics, Volume 13, Numbers 1-2, January-June 1987 67 
Technical Correspondence A Note on a Study of Cases 
/man/ 
/force/ 
/dest/ 
/act/ 
BY 
/ag/ 
/inst/ 
/obj/ 
/man/ 
/man/ 
/loc/ 
/loc/ 
/bef/ 
/tspan/ 
/quant/ 
/reas/ 
/inst/ 
/loc/ 
/inst/ 
/attr/ 
/quant/ /.9/ 
/quant/ 
/quant/ 
/inst/ 
/force/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
/inst/ 
The car -whizzed along at sixty -miles an hour. 
She felt sure she would -faint at the -sight of 
blood. 
As we passed the gate the dog -flew out at -us. 
That boy -is clever at -mathematics. 
The new library was --opened by the -Mayor. 
I shall -go by the 10:30 -train. 
He -seized me by the -arm. 
He -succeeded by hard -work. 
I -know him by -sight. 
She -sat by the -pillar. 
We -went by the -shop without realising it. 
They should have -been here by -now. 
They decided to -travel by -night. 
We -lost the match by one -goal. 
She -was very generous by -nature. 
They always -addressed each other by their 
Christian -names. 
I -went from London to Manchester by 
-Sheffield. 
He -swore by all the -Gods he would conquer. 
He brought a -person by the -name of Smoth. 
The carpet is three -yards by -four. 
The box is too heavy for me to lift by myself. 
-Little by -little his savings accumulated. 
Milk is -sold by the -pint. 
By my -watch, it' -s time for lunch. 
The meeting will be -held in the school, by 
-permission of the headmaster. 
By my -watch it' -s time for lunch. 
He -looked at it through his -binoculars. 
He -swore by all the -Gods he would conquer. 
I managed to -beat the dog off with a -stick. 
I shall -go by the 10:30 -train. 
Many locomotives nowadays -run on -oil. 
Marian Evans -wrote under the -name George 
Eliot. 
The parcel was -wrapped in -paper. 
The poem has been -set to -music. 
The prosecution was -brought under the 
-Vagrancy Act. 
The repairs -cost me over five -pounds. 
The roof of the building is -supported on 
-pillars. 
They always -addressed each other by their 
Christian -names. 
/inst/ You can -use that saucer for an -ashtray. 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
/man/ 
Cars are -parked at their owners' -risk. 
He -is, without --doubt, a very able person. 
He -left at a moment's -notice. 
He -said nothing in -reply to my questions. 
He -succeeded by hard -work. 
He -vaulted over the gate at a single -leap. 
I ---.did it without -thinking. 
I -know him by -sight. 
I only -said it in -fun. 
It -is all right in -theory. 
She -did the work to the -best of her ability. 
She -lay back with her ---eyes closed. 
The car -whizzed along at sixty -miles an hour. 
The child -screamed in -terror. 
The last few years of his life were -spent in 
-poverty. 
The water -shot up in a -fountain. 
The work has been ---carded out according to 
your -instructions. 
They -played out of -tune. 
They let him -have it on -loan. 
To our -surprise, the train -was early. 
Our complete sample is available in listings, with a fuller 
description. We acknowledge the immense value of 
Wood's work as a gource, and are grateful to the publish- 
er Macmillan for permission to reproduce and use 
Wood's material. 
Karen Sparck Jones and Branimir Boguraev 
Computer Laboratory 
University of Cambridge 
Corn Exchange Street 
Cambridge CB2 3QG, England 
arpa: sparckjones%cl.cam.ac.uk@cs.ucl.ac.uk 

REFERENCES 
Boguraev, B.K. 1979 Automatic Resolution of Linguistic Ambiguities. 
Technical Report 11, Computer Laboratory, University of 
Cambridge. 
Boguraev, B.K. and Sparck Jones, K. 1983 How to Drive a Database 
Front End Using General Semantic Information. Conference on 
Applied Natural Language Processing, Santa Monica: 81-88. 
Wilks, Y. 1977 Good and Bad Arguments about Semantic Primitives. 
Communication and Cognition 10:181-221, 
Wood, F.T. 1979 English Prepositional Idioms. Macmillan, London, 
England. 
