Putting Frames in Perspective
Nancy Chang, Srini Narayanan, and Miriam R.L. Petruck
International Computer Science Institute
1947 Center St., Suite 600, Berkeley, CA 94704
a0 nchang,snarayan,miriamp
a1 @icsi.berkeley.edu
Abstract
This paper attempts to bridge the gap between
FrameNet frames and inference. We describe a
computational formalism that captures structural re-
lationships among participants in a dynamic sce-
nario. This representation is used to describe the
internal structure of FrameNet frames in terms of
parameters for event simulations. We apply our for-
malism to the commerce domain and show how it
provides a flexible means of accounting for linguis-
tic perspective and other inferential effects.
1 Introduction
FrameNet (Fillmore et al., 2001) is an online lex-
ical resource1 designed according to the principles
of frame semantics (Fillmore, 1985; Petruck, 1996).
It thus takes as foundational the assumptions that
(1) lexical items draw on rich conceptual structures,
or frames, for their meaning and function; and (2)
conceptually related lexical items may foreground
different aspects of the same background frame.
Verbs involved with commercial events serve as
canonical examples:
(1) a. Chuck bought a car from Jerry for $1000.
b. Jerry sold a car to Chuck for $1000.
c. Chuck paid Jerry $1000 for a car.
d. Jerry charged Chuck $1000 for a car.
e. Chuck spent $1000 on a car.
The sentences in (1) might describe the same inter-
action – in which one individual (Chuck) transfers
money ($1000) to another (Jerry) in exchange for
some goods (a car) – but differ in the perspective
they impose on the scene.
The shared inferential structure of verbs like buy
and sell is captured in FrameNet by the COMMERCE
frame, which is associated with a set of situational
1http://www.icsi.berkeley.edu/framenet/
roles, or frame elements (FEs), corresponding to
event participants and props. These FEs are used to
annotate sentences like those in (1), yielding:
(2) a. [Chuck]Buyer bought [a car]Goods
[from Jerry]Seller [for $1000]Payment.
b. [Jerry]Seller sold [a car]Goods
[to Chuck]Buyer [for $1000]Payment.
FE tags act as a shorthand that allows diverse verbs
to tap into a common subset of encyclopedic knowl-
edge. Moreover, regularities in the set of FEs real-
ized with specific lexical items can be taken as cor-
related with their favored perspective.
A significant gap remains, however, between the
unstructured and intuitively chosen tag sets used in
FrameNet and a formal characterization of the inter-
related actions and relations holding among them.
An explicit representation of such frame-semantic
information is needed to fully realize FrameNet’s
potential use in text understanding and inference
(Fillmore and Baker, 2001). In this paper we at-
tempt to bridge the gap by defining a formalism
that unpacks the shorthand of frames into structured
event representations. These dynamic representa-
tions allow annotated FrameNet data to parameter-
ize event simulations (Narayanan, 1999b) that pro-
duce fine-grained, context-sensitive inferences. We
illustrate our formalism for the COMMERCE frame
and show how it can account for some of the wide-
ranging consequences of perspective-taking.
2 The FrameNet COMMERCE frame
The FrameNet project has thus far produced two
databases: a collection of approximately 80 frames
with frame descriptions, chosen to cover a broad
range of semantic domains; and a hand-annotated
dataset of about 50,000 sentences from the British
National Corpus (Baker et al., 1998). The databases
document both syntactic and semantic behavior of a
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                
Figure 1: Results of a query on the FrameNet COMMERCE frame, showing annotated data for the verb buy.
wide variety of lexical items (or lemmas) and thus
have the potential to allow corpus-based techniques
to be applied to semantically oriented tasks.2
The current release of the FrameNet databases3
defines a COMMERCE frame with frame elements
including the familiar Buyer, Seller, Payment and
Goods, along with several other FEs needed to
cover the data. The frame includes 10 verbs relevant
to commercial transactions, for a total of 575 anno-
tated sentences. Figure 1 shows a sampling of data
annotated with respect to the COMMERCE frame.
Considerable research has been devoted to ex-
plicating the connections among frames, perspec-
tive, and argument structure; see Gawron (ms.) and
Hudson (2002). But there has been relatively less
work that addresses inferential issues related to per-
spective. The COMMERCE frame, for example, is
implicitly associated with a complex, dynamic net-
work of interrelated events, actions and participants.
Our proposal is that perspectival effects may be best
understood in terms of subtle inferential effects on
interpretation licensed by this network.
Our task, then, is to make this inferential structure
2See (Gildea and Jurafsky, 2000) for some promising ini-
tial work in applying statistical techniques to the FrameNet
database to automatically label frame elements.
3We refer to data from FrameNet I; an interim release of
FrameNet II is expected soon.
explicit. We take the original COMMERCE frame
as our starting point and define the interrelation-
ships present among its FEs. The additional struc-
ture we impose on the COMMERCE frame allows
us to distinguish a perspective-neutral description
of a commercial transaction from the perspectivized
situations described by particular verbs. The re-
sulting event representation can be integrated with
a simulation-based inference engine to account for
differences in the interpretation of sentences like
those in the annotated FrameNet data.
3 Structured event representations
In this section, we present a formal specification
used for mapping the flat set of FEs in COM-
MERCE onto explicitly structured event representa-
tions based on the Embodied Construction Gram-
mar (ECG) formalism. ECG is a constraint-
based formalism similar in many respects to other
unification-based linguistic formalisms, such as
HPSG (Pollard and Sag, 1994).4 It differs from
other lingustically motivated proposals in that it is
4ECG includes formalisms for both schemas (conceptual
representations) and constructions (conventionalized pairings
of form and meaning), described in (Bergen and Chang, 2002).
We refer here only to the schema formalism in a simplified
form. See (Chang et al., 2002) for a more complete version
that has been extended to accommodate additional cognitive
linguistic primitives.
designed to support a model of language under-
standing in which utterances evoke a complex net-
work of conceptual schemas that are then mentally
simulated in context to produce a rich set of infer-
ences. It is thus ideally suited for our current goal
of translating frames to conceptual representations.
Figure 2 presents the ECG schema definition
language. The indented block labeled roles lists
and constrains the schema’s local roles, which are
equivalent to features (or in this case, frame FEs).
Roles are declared with a local name (local-role) and
may be accompanied by type restrictions (indicated
with ‘:’). Identification (or binding) constraints (in-
dicated with ‘a0a2a1 ’) may appear in either the roles
or the constraints block; these cause roles and con-
straints to be shared between its arguments, similar
to unification or coindexation.5 The subcase rela-
tion defines a schema inheritance lattice, with the
local schema inheriting all roles and constraints.
schema name
a3 subcase of schema
a4 evokes
a5 schema as local-name
a6
rolesa7a8
a9
a8a10
local-role
local-role : restriction
local-role a11a13a12 role
local-role a11a13a12 role : restriction
a14
a8a15
a8
a16
a17a18 constraints
a19
role a11a13a12 role
phase :: condition a20
Figure 2: Schema definition formalism. Keywords
are shown in bold; a left square bracket ([) marks
optional blocks; and curly braces (a21a23a22 ) enclose a set
of optional statements. See text for details.
The formalism also has several novel features that
we will exploit in representing commercial transac-
tions. The most important of these are: (1) the abil-
ity to flexibly evoke and relate multiple schemas,
due mainly to the evokes relation; and (2) the abil-
ity to assert dynamic conditions that apply to spe-
cific event stages, through the use of simulation
constraints. We will describe each of these briefly,
deferring details to the example schemas below.
Schemas listed in the evokes block are instanti-
ated locally (as local-name), but the relationship be-
5Constraints may refer to locally declared roles, inherited
roles, and evoked schemas, as well as any roles available
through these structures. Standard slot-chain notation is used
to refer to role y of a structure x as x.y.
tween the defined schema and the evoked schema
is underspecified. This underspecification allows
one schema to be defined in terms of another
schema without implying either full inheritance of
the evoked schema’s roles or containment in either
direction. In some cases, the evoked schema cor-
responds to a subpart of the evoking schema; alter-
natively, the evoked schema may serve as a back-
ground schema against which the evoking schema
is defined. We will see examples of each below.
Simulation constraints use the ‘::’ notation to as-
sert some condition on a particular phase of simula-
tion – either a relation that must hold or an event or
action that must take place during that phase. Simu-
lation phases correspond to event stages; these con-
straints serve as the bridging connection to previous
work on modeling event structure and linguistic as-
pect using active representations (Narayanan, 1997;
Chang et al., 1998).
We now show how the ECG formalism can be
used to define more complex schemas that provide
the underlying structure we need to tackle the COM-
MERCE frame; the key schemas for the current dis-
cussion are shown in Figure 3.6
The Event schema is of primary importance: it ap-
pears directly or indirectly in the rest of the schema
definitions, and it serves as the crucial link to simu-
lation. The definition given here is not intended to
capture the full complexity of the most generalized
event, which may have complex internal structure
(start and finish subevents, ongoing period, etc.).
At a coarser granularity, however, it may also be
viewed as a discrete temporal chunk that takes place
between two time slices. The schema as shown re-
flects this coarser view, which is sufficient for cur-
rent purposes: its roles include before, after, and
transition, all referring to simulation phases. An-
other role, the nucleus, is constrained only to hold
or take place during the transition phase. Together
these roles anchor the event to the passage of time.
The other schemas are more complex. The Trans-
fer schema corresponds to an event in which an agent
causes a theme to be transferred from the source
to the recipient. It is defined as evoking two other
schemas: an Action schema (with an actor role) and a
Receive schema (in which a receiver comes into pos-
session of the received entity). (These are not shown,
nor is the causal relation between them.) Note that
both act and rec are conceptually distinct from the
6Some schema definitions have been omitted or simplified
to conserve space; relevant details are mentioned in the text.
schema Event
roles
before : Phase
transition : Phase
after : Phase
nucleus
constraints
transition a0a1a0 nucleus
schema Transfer
subcase of Event
evokes
Action as act
Receive as rec
roles
agent a11a13a12 act.actor
source : Entity
theme a11a13a12 rec.received
recipient a11a13a12 rec.receiver
constraints
transition a0a1a0 act
transition a0a1a0 rec
after a0a1a0 has(recipient,theme)
schema Exchange
subcase of Event
roles
participant1 : Human
participant2 : Human
entity1 : Entity
entity2 : Entity
transfer1 : Transfer
transfer2 : Transfer
agent : Entity
constraints
transition a0a2a0 transfer1
transition a0a2a0 transfer2
transfer1.source a11a13a12 participant1
transfer1.theme a11a13a12 entity1
transfer1.recipient a11a13a12 participant2
transfer2.source a11a13a12 participant2
transfer2.theme a11a13a12 entity2
transfer2.recipient a11a13a12 participant1
Figure 3: The Event, Transfer and Exchange schemas.
nucleus role inherited from Event, although all are
constrained to take place during the event’s transi-
tion phase. The agent role is constrained to be the
same entity as the actor of act. Importantly, the
Transfer event schema makes no commitment as to
whether its agent – the entity seen as causing the
overall event – is the source, recipient or even theme.
It is in this respect that the Transfer schema can be
considered neutral in perspective.
The Exchange schema is structurally similar to the
Transfer schema and provides most of the relevant
constraints needed for commercial transactions. It
includes two transfer events that occur during the
transition phase and are parameterized straightfor-
wardly in the constraints block by two human par-
ticipants and two entities. An additional agent role
is not bound to any particular entity; this schema
is thus also perspective-neutral, since either partici-
pant (or both) might be viewed as active.
4 Commercial transaction schemas
We are now in a position to return to the commerce
domain and put our inventory of domain-general
schemas to use. We first define the Commercial-
Transaction (CT) schema as a subcase of the Ex-
change schema with appropriate role identifications
and an additional type restriction on entity1. The role
names in this schema differ slightly from those in
FrameNet’s COMMERCE, reflecting its perspective-
neutral status. But given the obvious mapping to
the FrameNet FEs, the CT schema fulfills part of
our original objective: based on its inherited and
evoked schemas and constraints, it concisely and
precisely states the conceptual underpinnings of the
basic commercial transaction.
schema Commercial-Transaction
subcase of Exchange
roles
customer a11a13a12 participant1
vendor a11a13a12 participant2
money a11a13a12 entity1 : Money
goods a11a13a12 entity2
goods-transfer a11 a12 transfer1
money-transfer a11 a12 transfer2
Figure 4: The Commercial-Transaction schema.
The CT schema provides the underlying in-
frastructure against which various perspectivized
schemas can be defined. As shown in Figure 5, we
treat Buy, Sell and Pay as schemas that evoke the CT
schema and identify their roles with specific partic-
ipants and event stages of the evoked CT schema.
Note the use of the keyword self (which we treat as
a special kind of role) to refer to the schema being
defined: Buy and Sell schemas each identify self with
the ct.nucleus role (that is, the nucleus of its evoked
commercial transaction), and is thus constrained to
take place during the evoked CT’s transition phase.
In contrast, since Pay identifies itself with ct.money-
transfer.nucleus, it refers specifically to a subpart of
the overall commercial transaction, such that its ex-
ecution does not necessarily entail the execution of
the goods-transfer in the event (i.e., you don’t always
get what you pay for).
The three schemas also differ in their partici-
pant role bindings: all are defined as subcases of
schema Buy
subcase of Transitive-Action
evokes Commercial-Transaction as ct
roles
self a11 a12 ct.nucleus
buyer a11a13a12 actor a11a13a12 ct.agent a11a13a12 ct.customer
goods a11a13a12 undergoer a11a13a12 ct.goods
schema Sell
subcase of Transitive-Action
evokes Commercial-Transaction as ct
roles
self a11a13a12 ct.nucleus
seller a11a13a12 actor a11a13a12 ct.agent a11a13a12 ct.vendor
goods a11a13a12 undergoer a11a13a12 ct.goods
schema Pay
subcase of Transitive-Action
evokes Commercial-Transaction as ct
roles
self a11a13a12 ct.money-transfer.nucleus
payer a11a13a12 actor a11a13a12 ct.customer
a11 a12 ct.money-transfer.agent
payment a11a13a12 ct.money
payee a11a13a12 ct.vendor
Figure 5: The Buy, Sell and Pay schemas.
Transitive-Action (not shown), which corresponds to
a prototypical situation in which an actor entity af-
fects or manipulates an undergoer entity. The Buy
and Sell schemas both identify the undergoer with
ct.goods, and the actor with ct.agent. But the two
schemas impose different views on the same situ-
ation by virtue of a single additional constraint on
this latter role (which corresponds to the active par-
ticipant in the overall CT), binding it to either the
ct.customer (Buy) or the ct.vendor (Sell). The bind-
ings in the Pay schema assert that its actor is the
ct.customer, as well as the agent of the money-transfer.
Other schemas associated with the CT schema
lend themselves to similar analyses, though they
draw on additional schemas not defined here. For
example, the Spend schema evokes a schema for re-
source consumption (as in (Hudson, 2002)); Charge
involves the vendor’s communication of the price
to the customer as a prerequisite to the overall ex-
change of goods and money. In general, the CT
schema explicitly specifies the internal event struc-
ture of a commercial transaction but remains non-
committal about which of its participants is seen as
active. This flexibility in representation allows other
schemas to effect the bindings that make appropri-
ate commitments on an individual basis.
5 Simulation semantics
The structured event formalism we have described
allows us to translate FrameNet descriptions into
a representation suitable for simulative inference.
Central to the representation is an event model
called executing schemas (or x-schemas), moti-
vated by research in both sensorimotor control and
cognitive semantics (Narayanan, 1997). X-schemas
are active structures that cleanly capture sequen-
tiality, concurrency and event-based asynchronous
control. They thus provide a cognitively moti-
vated basis for modeling diverse linguistic phenom-
ena, including aspectual inference (Chang et al.,
1998), metaphoric inference (Narayanan, 1999a)
and event-based reasoning in narrative understand-
ing (Narayanan, 1999b). In this paper, we focus on
the problem of frame-based inference and the atten-
dent problem of modeling perspectival effects.
The event model is based on the Petri net, which
in its basic form is a weighted, bipartite graph
consisting of places (shown as circles) and transi-
tions (shown as rectangles) connected by directed
input and output arcs (Murata, 1989; Narayanan,
1997). Places may contain tokens (i.e., they may
be marked), and they typically represent states, re-
sources or conditions that apply. Transitions typi-
cally represent actions or events. X-schemas extend
the basic Petri net to include typed arcs, hierarchi-
cal control, durative transitions, parameterization,
typed (individual) tokens and stochasticity.
The most relevant property of the x-schema for
this paper is its well-specified execution semantics:
a transition is enabled when all its input places are
marked, such that it can fire by moving tokens from
input to output places. The active execution seman-
tics serves as the engine of context-sensitive infer-
ence in the simulation-based model of language un-
derstanding mentioned earlier.
The ECG formalism is designed to allow con-
straints on x-schema simulation to be expressed.
In particular, the Event schema in Figure 3 has
roles that refer to event phases; these correspond
to x-schema places and transitions. Other schema
roles specify x-schema parameters, which allow x-
schemas to give rise to different execution traces
through the network with different parameters.
The Commercial-Transaction schema has been im-
plemented in the KarmaSIM x-schema simulation
environment (Narayanan, 1997); Figure 6 shows
part of the network. The phase roles from the
schemas in Section 3 have been mapped onto the
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                
Figure 6: KarmaSIM simulation of the Commercial-Transaction schema. The highlighted execution is associ-
ated with the Pay schema, corresponding to the money-transfer event.
fine-grained temporal structure of each event, cor-
responding to the various control nodes in the net-
work (ready, ongoing, finish, done, etc.); the transi-
tion phase referenced in the schemas is expanded as
the start, ongoing and finish nodes. As shown, execu-
tion of the overall CT schema comprises the execu-
tion of two subsidiary events, the goods-transfer and
the money-transfer. These need not be synchronized,
but both must complete for the overall commercial
transaction to complete (enforced by the arcs from
ongoing(money-transfer) and ongoing(goods-transfer) to
finish(transfers)). All the frame-based inferences of
the CT frame (e.g., the seller (buyer) has the goods
(money) until the goods-transfer (money-transfer) is
completed, and the seller (buyer) has the money
(goods) when the money-transfer (goods-transfer) is
completed) come from simulating the CT frame.
In the simulation framework, perspectival effects
come in at least three flavors. First, the frame el-
ement binding patterns may differ among perspec-
tives, as illustrated by Figure 5, in which the lexi-
cal item buy identifies the actor of the transitive-action
with both the customer of the CT and the agent of the
money-transfer. This issue of binding has been the fo-
cus of previous work (see Section 2); our approach
is similar to construction-based proposals that ex-
plicitly represent the binding constraints for differ-
ent frame element binding patterns.
Second, some perspectives specify the specific
subevents (or collection of subevents) to simulate
while others require simulating the entire event
frame. An example of this is shown in Figure 6,
where the highlighted money-transfer portion of the
network corresponds to a simulation of the Pay
schema. The token in ongoing(ct) shows that there is
an ongoing transaction, but the finish(transfers) tran-
sition is not enabled. Technically, the done(ct) place
is not reachable (absent other information), since
the simulation of Pay does not provide direct evi-
dence for the occurrence of a goods-transfer.7 In con-
trast, both Buy and Sell involve simulating the en-
tire transaction, include both transfers as well as the
done(ct) node. (Thus, the entire network in Figure 6
can be considered an expansion of the CT schema’s
transition phase.)
A third, more subtle aspect of perspective is re-
lated to the problem of linguistic focus. The per-
spectival difference between Buy and Sell, for in-
stance, is only partially captured by their different
FE bindings to the CT frame. Another difference
stems from the foregrounding of specific relations:
buy foregrounds the interaction between the Buyer
and the Goods (including the eventual possession of
the Goods), while sell foregrounds the interaction be-
tween the Seller and the Goods. Work in progress
suggests that many foregrounding cases can be han-
dled by simulating different parts of the event at
varying degrees of detail. For example, the simu-
lation for Buy could execute x-schemas in which the
Buyer interacts with the Goods – such as the goods-
transfer and its resulting possession (abbreviated as
has(Chuck, car) in Figure 6) – at the default granu-
7Contextual or background knowledge could provide evi-
dence for the other transfer or allow it to be inferred by default.
larity, while other x-schemas are collapsed into less
detailed simulations. (See (Narayanan, 1997) for a
detailed model of simulation at multiple levels of
granularity.) While the model is able to handle some
of the issues pertaining to foregrounding and focus,
a full account remains a topic of ongoing research.
6 Discussion and conclusions
FrameNet shows considerable promise for enabling
qualitative breakthroughs on NLP applications re-
quiring increased semantic and pragmatic sophis-
tication, including information extraction, word-
sense disambiguation, and question answering.
FrameNet frames are intended to capture crucial
generalizations not available in other lexical re-
sources. WordNet (Fellebaum, 1998), for example,
includes only simple taxonomic relations (buy and
sell are listed as hyponyms of get and give, respec-
tively, and as antonyms of each other). The Prop-
Bank project (Kingsbury and Palmer, 2002) is, like
FrameNet, geared toward the creation of a seman-
tically annotated corpus (by adding general logical
predicates to the Penn Treebank), though without
any common background frame structures across
lexical items.
While frames and FE tags are meaningful to hu-
man interpreters, they are not yet suitable for use
in natural language understanding. In this paper
we have shown how FrameNet tags can be pre-
cisely defined in terms of structured event repre-
sentations, which can support parameterized simu-
lations that license active inferences. The formal-
ism appears expressive enough for the COMMERCE
frame, and uses methods of simulation semantics to
handle frame-based inferences and associated per-
spectival effects.
We are currently automating the process of map-
ping frame definitions to simulation parameteriza-
tions and extending the representation to cover the
entire FrameNet II database.
Acknowledgments
Thanks to Chuck Fillmore, Jerry Feldman, the FrameNet
and NTL groups, and the ScaNaLU workshop partici-
pants for early feedback.

References

Collin F. Baker, Charles J. Fillmore, and John B. Lowe.
1998. The Berkeley FrameNet Project. In Proc.
COLING-ACL, Montreal, Canada.

Benjamin K. Bergen and Nancy C. Chang. 2002. Em-
bodied Construction Grammar in simulation-based
language understanding. Technical Report TR-02-
004, International Computer Science Institute.

Nancy Chang, Daniel Gildea, and Srini Narayanan.
1998. A dynamic model of aspectual composition.
In Proc. 20th Cognitive Science Society Conference,
Madison, Wisconsin.

Nancy Chang, Jerome Feldman, Robert Porzel, and
Keith Sanders. 2002. Scaling cognitive linguistics:
Formalisms for language understanding. In Proc.
1st International Workshop on Scalable Natural Lan-
guage Understanding, Heidelberg, Germany.

Christianne Fellebaum, editor. 1998. WordNet: An Elec-
tronic Database. MIT Press.

Charles J. Fillmore and Collin F. Baker. 2001. Frame
semantics for text understanding. In Proc. WordNet
and Other Lexical Resources Workshop, Pittsburgh.
NAACL.

Charles J. Fillmore, Charles Wooters, and Collin Baker.
2001. Building a large lexical databank which pro-
vides deep semantics. In Proc. Pacific Asian Con-
ference on Language, Information and Computation,
Hong Kong.

Charles J. Fillmore. 1985. Frames and the semantics of
understanding. Quaderni di Semantica, IV(2).

Jean Mark Gawron. 2002. Frames, modifiers, and argu-
ment structure. Ms., San Diego State University.
Daniel Gildea and Daniel Jurafsky. 2000. Automatic la-
beling of semantic roles. In Proceedings of the 38th
Annual Meeting of the ACL, Hong Kong, October.

Richard Hudson. 2002. Buying and selling in Word
Grammar. In press.

Paul Kingsbury and Martha Palmer. 2002. From Tree-
bank to PropBank. In Proc. 3rd International Confer-
ence on Language Resources and Evaluation (LREC-
2002).

Tadao Murata. 1989. Petri nets: Properties, analysis,
and applications. In Proc. IEEE-89, volume 77, pages
541–576.

Srini Narayanan. 1997. Knowledge-based Action Repre-
sentations for Metaphor and Aspect (KARMA). Ph.D.
thesis, University of California at Berkeley.

Srini Narayanan. 1999a. Moving right along: A compu-
tational model of metaphoric reasoning about events.
In Proc. Sixteenth National Conference of Artificial
Intelligence (AAAI-99). AAAI Press, Menlo Park.

Srini Narayanan. 1999b. Reasoning about actions in nar-
rative understanding. In Proc. Sixteenth International
Joint Conference on Artificial Intelligence (IJCAI-99).
Morgan Kaufmann Press.

Miriam R.L. Petruck. 1996. Frame semantics. In J. Ver-
schueren, J. ¨Ostman, J. Blommaert, and C. Bulcaen,
editors, Handbook of Pragmatics. John Benjamins,
Philadelphia.

Carl Pollard and Ivan A. Sag. 1994. Head-Driven
Phrase Structure Grammar. University of Chicago
Press, Chicago.
