File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/89/p89-1022_metho.xml
Size: 26,121 bytes
Last Modified: 2025-10-06 14:12:23
<?xml version="1.0" standalone="yes"?> <Paper uid="P89-1022"> <Title>AUTOMATIC ACQUISITION OF THE LEXICAL SEMANTICS OF VERBS FROM SENTENCE FRAMES*</Title> <Section position="4" start_page="177" end_page="177" type="metho"> <SectionTitle> STRUCTURED OVERCOM- MITMENT AND A LEARNING ALGORITHM </SectionTitle> <Paragraph position="0"> In this paper, we will present a computational model of verb acquisition which uses what we will call the principle of structured o~ercomrnitment to eliminate the need for such negative evidence. In essence, our learner learns by initially jumping to the strongest conclusions it can, simply assuming that everything within its descriptive system that it hasn't seen will never occur, and then later weakening its hypotheses when faced with contradictory evidence. Thus, the learner escapes from the need to be told that certain possibilities cannot occur (i.e. are&quot;ungrammatical') by the simple expedient of assuming that all properties it has observed are either always obligatory or always forbidden. If and when the learner discovers that it was wrong about such a strong assumption, it reclassifies the property from either obligatory or forbidden to merely optional.</Paragraph> <Paragraph position="1"> Note that this learning principal requires that no intermediate analysis is ever abandoned; analyses are only further refined by the weakening of universals (X ALWAYS has property P) to existenrials (X SOMETIMES has property P). It is in this sense that the overcommitment is&quot;structured.&quot; For such a learning strategy to work, it must be the case that the set of features which underlies the learning process are surface observable; the learner must be able to determine of a particular instance of (in this case) a verb structure whether some property is true or false of it. This would seem to imply, as far as we can tell, a commitment to the notion of em learning as selection widely presupposed in the linguistic study of generative grammar (as surveyed, for example, in Berwick(1985).</Paragraph> <Paragraph position="2"> Thus, we propose that the problem of learning the category of a verb does not require that a natural language understanding system synthesize em de novo a new structure to represent its semantic class, but rather that it determine to which of a predefined, presumably innate set of verb categories a given verb belongs. In what follows below, we argue that a relevant classification of verb categories can be represented by simple conjunctions of a finite number of predefined quasi-independent features with no need for disjunction or complex boolean combinations of features.</Paragraph> <Paragraph position="3"> Given such a feature set, the Principal of Structured Overcommitment defines a partial ordering (or, if one prefers, a tangled hierarchy) of verbs as follows: At the highest level of the hierarchy is a set of verb classes where all the primary four features, where defined, are either obligatory or forbidden. Under each of these &quot;primary&quot; categories there are those categories which differ from it only in that some category which is obligatory or forbidden in the higher class is optional in the lower class. Note that both obligatory and forbidden categories at one level lead to the same optional category at the next level down.</Paragraph> <Paragraph position="4"> The learning system, upon encountering a verb for the first time, will necessarily classify that verb into one of the ten top-level categories. This is because the learner assumes, for example, that if a verb is used with an object upon first encounter, that it always has an object; if it has no object, that it never has an object, etc. The learner will leave each verb classification unchanged upon encountering new verb instances until a usage occurs that falsifies at least one of the current feature values. When encountering such a usage i.e. a verb frame in which a property that is marked obligatory is missing, or a property that is marked forbidden is present (there are no other possibilities) - then the learner reclassifies the verb by moving down the hierarchy at least one level replacing the OBLIGATORY or FORBIDDEN value of that feature with OPTIONAL.</Paragraph> <Paragraph position="5"> Note that, for each verb, the learner's classifica.</Paragraph> <Paragraph position="6"> tion moves monotonically lower on this hierarchy, until it eventually remains unchanged because the learner has arrived at the correct value. (Thus this learner embodies a kind of em learning in the limit.</Paragraph> </Section> <Section position="5" start_page="177" end_page="180" type="metho"> <SectionTitle> 3 THE FEATURE SET AND THE VERB HIERARCHY </SectionTitle> <Paragraph position="0"> As discussed above, our learner describes each verb by means of a vector of features. Some of these features describe syntactic properties of the verb (e.g.&quot;Takes an Object&quot;), others describe aspects of the theta-structure (the predicate/argument structure) of the verb (e.g.&quot;Takes an Agent&quot;,~Ikkes a Theme&quot;), while others describe some key properties of the mapping between theta-structure and syntactic structure (e.g.&quot;Theme Appears As Surface Object&quot;). Most of these features are three-valued; they describe properties that are either always true (e.g. that&quot;devour&quot; always Takes An Object), always false (e.g. that &quot;fall&quot; never Takes An Object) or properties that are optionally true (e.g. that&quot;eat&quot; optionally Takes An Object). Always true values will be indicated as&quot;q-&quot; below, always false values as&quot;-&quot; and optional values as~0 &quot;.</Paragraph> <Paragraph position="1"> All verbs are specified for the first three features mentioned above: &quot;Takes an Object&quot; (OBJ),&quot;Takes an Agent&quot; (AGT), and&quot;Takes a Theme&quot; (THEME). All verbs that allow OBJ and THEME are specified for&quot;Theme Appears As Object&quot; (TAO), otherwise TAO is undefined. At the highest level of the hierarchy is a set of verb classes where all these primary features, where defined, are either obligatory or forbidden. Thus there are at most 10 primary verb types; of the eight for the first three features, only two (-I--q-, and -H-+) split for TAO.</Paragraph> <Paragraph position="2"> The full set of features we assume include the primary set of features (OBJ, AGT, THEME, and TAO), as described above,,and a secondary set of features which play a secondary role in the learning algorithm, as will be discussed below. These secondary features are either thematic properties, or correlations between thematic and syntactic roles. The thematic properties are: LOC - takes a locative; INST - takes an instrument; and DAT takes a dative. The first thematic-syntactic mapping feature &quot;Instrument as Subject&quot; is fake if no instrument can. appear in subject position (or, true if the subject is always an instrument, al-&quot; though this is never the case.) The second such feature &quot;Theme as Chomeuf (TAC) is the only non-trinary-valued feature in our learner; it specifies what preposition marks the theme when it is not realized as subject or object. This feature, if not -, either takes a lexical item (a preposition, actually, as its value, or else the null string. We treat verbs with double objects (e.g. &quot;John gave Mary the ball.&quot;) as having a Dative as object, and the theme as either marked by a null preposition or, somewhat alternatively, as a bare NP chomeur.</Paragraph> <Paragraph position="3"> (The facts we deal with here don't decide between these two analyses.) Note that this analysis does not make explict what can appear as object; it is a claim of the analysis that if the verb is OBJ:/ or OBJ:0 and is TAO:- or TAO:0, then whatever other thematic roles may occur can be realized as the object. This may well be too strong, but we are still seeking a counterexample.</Paragraph> <Paragraph position="4"> Figure 1 shows our classification of some verb classes of English, given this feature set. (This classification owes much to Levin(1985), as well as to Grimshaw(1983) and Jackendoff(1983).) This is only the beginning of such a classification, clearly; for example, we have concentrated our efforts solely on verbs that take simple NPs as complements. Our intention is merely to provide a rich enough set of verb classes to show that our classification scheme has merit, and that the learning algorithm works. We believe that this set of features is rich enough to describe not only the verb classes covered here but other similar classes. It is also our hope that an analysis of verbs with richer complement structures will extend the set of features without changing the analysis of the classes currently handled.</Paragraph> <Paragraph position="5"> It is interesting to note that although the partial ordering of verb classes is defined in terms of features defined over syntactic and theta structures, that there appears to be at least a very strong semantic reflex to the network. Due to lack of space, we label verb cla-~ses in Figure 1 only with exemplars; here we give a list of either typical verbs in the class, and/or a brief description of the class, in semantic terms: * Spray, load, inscribe, sow: Verbs of physical contact that show the completive/noncompletire 1 alternation. If completive, like &quot;fill&quot;. * Clear, empty: Similar to spray/load, but if completive, like &quot;empty&quot;.</Paragraph> <Paragraph position="6"> * Wipe: Like clear, but no completive pattern.</Paragraph> <Paragraph position="7"> * Throw: The following four verb classes all involve an object and a trajectory. '~rhrow&quot; verbs don't require a terminus of the trajectory. null * Present: Like &quot;throw&quot;, as far as we can tell. * Give: Requires a terminus.</Paragraph> <Paragraph position="9"/> <Paragraph position="11"> * Poke, jab, stick, touch: Some object follows a trajectory, resulting in surface contact.</Paragraph> <Paragraph position="12"> * Hug: Surface contact, no trajectory.</Paragraph> <Paragraph position="13"> * Fill: Inherently C/ompletive verbs.</Paragraph> <Paragraph position="14"> * Search: Verbs that show a completive/noncompletive alternation that doesn't involve physical contact.</Paragraph> <Paragraph position="15"> * Die, flower: Change of state. Inherently nonagentive. null * Break: Change of state, undergoing causitive alternation.</Paragraph> <Paragraph position="16"> * Destroy: Verbs of destruction.</Paragraph> <Paragraph position="17"> * Pierce: Verbs of destruction involving a trajectory. null * Devour, dynamite: Verbs of destruction with incorporated instruments * Put: Simple change of location.</Paragraph> <Paragraph position="18"> * Eat: Verbs of ingesting allowing instruments * Breathe: Verbs of ingesting that incorporate instrument * Fall, swim: Verbs of movement with incorporated theme and incorporated manner. * Push: Exerting force; maybe something moves, maybe not.</Paragraph> <Paragraph position="19"> * Stand: Like &quot;break s, but at a location. * Rain: Verbs which have no agent, and incor null porate their patient.</Paragraph> <Paragraph position="20"> The set of verb classes that we have investigated interacts with our learning algorithm to define the partial order of verb classes illustrated schematically in Figure 2.</Paragraph> <Paragraph position="21"> For simplicity, this diagram is organized by the values of the four principle features of our system. Each subsystem shown in brackets shares the same principle features; the individual verbs within each subsystem differ in secondary features as shown. If one of the primary features is made optional, the learning algorithm will map all verbs in each subsystem into the same subordinate subsystem as shown; of course, secondary feature values are maintained as well. In some cases, a sub-hierarchy within a subsystem shows the learning of a secondary feature.</Paragraph> <Paragraph position="22"> We should note that several of the primary verb classes in Figure 2 are unlabelled because they correspond to no English verbs: The class &quot;----&quot; would be the class of rain if it didn't allow forms like ~hail stones rained from the sky&quot;, while the class '~+--I--t-&quot; would be the class of verbs like &quot;destrof' if they only took instruments as subjects. Such classes may be artifacts of our analysis, or they may be somewhat unlikely classes that are filled in languages other than English.</Paragraph> <Paragraph position="23"> Note that sub-patterns in the primary feature subvector seem to signal semantic properties in a straightforward way. So, for example, it appears that verbs have the pattern {OBJ:+, THEME:+, TAO:-} only if they are inherently completive; consider &quot;search&quot; and &quot;fill&quot;. Similarly, the rare verbs that have the pattern {OBJ:-, THEME:-}, i.e those that are truly intransitive, appear to incorporate their theme into their meaning; a typical case here is =swim&quot;. Verbs that are {OBJ:-, AGT:-} (e.g. =die&quot;) are inherently stative; they allow no agency. Those verbs that are {AGT:+} incorporate the instrument of the operation into their meaning. We will have to say about this below. null</Paragraph> </Section> <Section position="6" start_page="180" end_page="182" type="metho"> <SectionTitle> 4 THE LEARNING ALGORITHM AT WORK </SectionTitle> <Paragraph position="0"> Let us now see how the learning algorithm works for a few verbs.</Paragraph> <Paragraph position="1"> Our model presupposes that the learner receives as input a parse of the sentence from which to derive the subject and object grammatical relations, and a representation of what NPs serve as agent, patient, instrument and location. This may be seen as begging the question of verb acquisition, because, it may be asked, how could an intelligent learner know what entities function as agent, patient, etc. without understanding the meaning of the verb? Our model in fact presupposes that a learner can distinguish between such general categories as animate, inanimate, instrument, and locative from direct observation of the environment, without explicit support from verb meaning; i.e. that it will be clear from observation em who is acting on em what em where. This assumption is not unreasonable; there is strong experimental evidence that children do in fact perceive even something as subtle as the difference between animate and inanimate motion well before the two word stage (see Golinkoff et al, 1984). Thisnotion that agent, patient and the like can be derived from direct observation (perhaps focussed by what NPs appear in the sentence) is a weak form of what is sometimes called the em semantic bootstrapping hypothesis (Pinker(1984)). The theory that we present here is actually a combination of this weak form of semantic bootstrapping with what is called em syntactic bootstrapping, the notion that syntactic frames alone offer enough information to classify verbs (see Naigles, Gleitman, and Gleitman (in press) and Fisher, Gleitman and Gleitman(1988).) null With this preliminary out of the way, let's turn to a simple example. Suppose the learner encounters the verb &quot;break&quot;, never seen before, in the context (6) The window broke.</Paragraph> <Paragraph position="2"> The learner sees that the referent of &quot;the window&quot; is inanimate, and thus is the theme. Given this and the syntactic fzarne of (6), the learner can see that em break (a) does not take an object, in this case, (b) does not take an agent, and (c) takes a patient. By Structured Overcommitment, the learner therefore assumes that em break em never takes an object, em never takes a subject, and em always takes a patient. Thus, it classifies em break as {OBJ:-, AGT:-, THEME:+, TAO:-} (ifTAO is undefined, it is assigned &quot;-'). It also assumes that em break is {DAT:-, LOC:-, INST:-, ... } for similar reasons. This is the class of DIE, one of the toplevel verb classes.</Paragraph> <Paragraph position="3"> Next, suppose it sees (7) John broke the window.</Paragraph> <Paragraph position="4"> and sees from observation that the referent of &quot;John&quot; is an agent, the referent of &quot;the window&quot; a patient, and from syntax that &quot;John&quot; is subject, and &quot;the window&quot; object. That em break takes an object conflicts with the current view that em break NEVER takes an object, and therefore this strong assumption isweakened to say that em break SOMETIMES takes an object. Similarly, the learner must fall back to the position that em break SOMETIMES can have the theme serve as object, and can SOMETIMES have an agent. This takes {OBJ:-, AGT:-, THEME:+, TAO:-} to {OBJ:0, AGT:0, THEME:+, TAO:0}, which is the class of both em break and em stand.</Paragraph> <Paragraph position="5"> However, since it has never seen a locative for ern break, it assumes that em break falls into exactly the category we have labelled as &quot;break&quot;.2 2And how would it distinguish between The vase stood on the table.</Paragraph> <Paragraph position="6"> mad There are, of course, many other possible orders in which the learner might encounter the verb em break. Suppose the learner first encounters the pattern (8) John broke the window.</Paragraph> <Paragraph position="7"> beR)re any other occurrences of this verb. Given only (8), it will assume that em break always takes an object, always takes an agent, always has a patient, and always has the patient serving as object. The learner will also assume that em break never takes a location, a dative, etc. This will give it the initial description of {OBJ:+, AGT:+, THEME:+, TAO:+, ..., LOC:-), which causes the learner to classify em break as falling into the toplevel verb class of DEVOUR, verbs of destruction with the instrument incorporated into the verb meaning.</Paragraph> <Paragraph position="8"> Next, suppose the learner sees (9) The hammer broke the window.</Paragraph> <Paragraph position="9"> where the learner observes that '~hammer&quot; is an inanimate object, and therefore must serve as instrument, not agent. This means that the earlier assumption that agent is necessary was an overcommitment (as was the unmentioned assumption that an instrument was forbidden). The learner therefore weakens the description of em break to {OBJ:+, AGT:0, THEME:-{-, TAO:+, ..., LOC:-, INST:0}, which moves em break into the verb class of DESTROY, destruction without incorporated instrument.</Paragraph> <Paragraph position="10"> Finally (as it turns out), suppose the learner sees (10) The window broke.</Paragraph> <Paragraph position="11"> Now it discovers that the object is not obligatory, and also that the theme can appear as subject, not object, which means that TAO is optional, not obligatory. This now takes em break to {OBJ:0, AGT:0, THEME:+, TAO:0, ... }, which is the verb class of break.</Paragraph> <Paragraph position="12"> We interposed (9) between (8) and (10) in this sequence just to exercise the learner. If (10) followed (8) directly, the learner would have taken em break to verb class BREAK all the more quickly.</Paragraph> <Paragraph position="13"> Although we will not explicitly go through the exercise here, it is important to our claims that any permutation of the potential sentence frames of em break will take the learner to BREAK, although some combinations require verb classes not shown The base broke on the table? This is a probl~n we discuss at the end of this paper. on our chart for the sake of simplicity (e.g. the class {OBJ:0, AGT:-, THEME:+, TAO:0} if it hasn't yet seen an agent as subject.).</Paragraph> <Paragraph position="14"> We were somewhat surprised to note that the trajectory of em break takes the learner through a sequence of states whose semantics are useful approximations of the meaning of this verb. In the first case above, the learner goes through the class of &quot;change of state without agency&quot;, into the class of BREAK, i.e. &quot;change of state involving no location&quot;. In the second case, the trajectory takes the learner through &quot;destroy with an incorporated instrument&quot;, and then DESTROY into BREAK.</Paragraph> <Paragraph position="15"> In both of these cases, it happens that the trajectory of em break through our hierarchy causes it to have a meaning consistent with its final meaning at each point of the way. While this will not always be true, it seems that it is quite often the case. We find this property of our verb classification very encouraging, particularly given its genesis in our simple learning principle.</Paragraph> <Paragraph position="16"> We now consider a similar example for a different verb, the verb em load, in somewhat terser form. And again, we have chosen a somewhat indirect route to the final derived verb class to demonstrate complex trajectories through the space of verb classes. Assume the learner first encounters (II) John loads the hay onto the truck.</Paragraph> <Paragraph position="17"> From (11), the learner builds the representation {OBJ:+, AGT:+, THEME:+, TAO:+, ..., LOC:+, ..., DAT:-}, which lands the learner into the class of PUT, i.e. &quot;simple change of location&quot;. We aasume that the learner can derive that &quot;the truck&quot; is a locative both from the prepositional marking, and from direct observation.</Paragraph> <Paragraph position="18"> Next the learner encounters (12) John loads the hay.</Paragraph> <Paragraph position="19"> From this, the learner discovers that the location is not obligatory, but merely optional, shifting it to {OBJ:+, AGT:+, THEME:+, TAO:+, ..., LOC:O ..., DAT:-}, the verb class of HUG, with the general mean/ng of &quot;surface contact with no trajectory.&quot; The next sentence encountered is (13) John loads the truck with hay.</Paragraph> <Paragraph position="20"> This sentence tells the learner that the theme need only optionally serve as object, that it can be * shifted to a non-argument position marked with the preposition em with. This gives em load the description of {OBJ:+, AGT:+, THEME:+, TAO:0, TAC:with, ..., LOC:0 .... DAT:-}. This new description takes em load now into the verb class of POKE/TOUCH, surface contact by an object that has followed some trajectory. (We have explicitly indicated in our description here that {DAT:-} was part of the verb description, rather than leaving this fact implicit, because we knew, of course, that this feature would be needed to distinguish between the verb classes of GIVE and POKE/TOUCH. We should stress that this and many other features are encoded as &quot;-&quot; until encountered by the learner; we have simply suppressed explicitly representing such features in our account here unless needed.) Finally, the learner encounters the sentence (14) John loads the truck.</Paragraph> <Paragraph position="21"> which makes it only optional that the theme must occur, shifting the verb representation to {OBJ:+, AGT:+, THEME:0, TAO:0, TAC:with, ..., LOC:0 ..., DAT:-}. The principle four features of this description put the verb into the general area of WIPE, CLEAR and SPRAY/LOAD, but the optional locative, and the fact that the theme can be marked with em with select for the class of SPRAY/LOAD, verbs of physical contact that show the completive/noncompletive alternation: null Note that in this case again, the semantics of the verb classes along the learning trajectory are reasonable successive approximations to the meaning of the verb.</Paragraph> </Section> <Section position="7" start_page="182" end_page="183" type="metho"> <SectionTitle> 5 FURTHER RESEARCH AND SOME PROBLEMS </SectionTitle> <Paragraph position="0"> One difficulty with this approach which we have not yet confronted is that real data is somewhat noisy. For example, although it is often claimed that Motherese is extremely clean, one researcher has observed that the verb &quot;put&quot;, which requires both a location and an object to be fully grammatical, has been observed in Motherese (although extremely infrequently) without a location. We strongly suspect, of course, that the assumption that one instance suffices to change the learner's model is too strong. It would be relatively easy to extend the model we give here with a couple of bits to count the number of counterexamples seen for each obligatory or forbidden feature, with two or three examples needed within some limited time period to shift the feature to optional.</Paragraph> <Paragraph position="1"> Can the model we describe here be taken as a psychological model? At first glance, clearly not, because this model appears to be deeply conservative, and as Pinker(1987) demonstrates, children freely use verbs in patterns that they have not seen. In our terms, they use verbs as if they had moved them down the hierarchy without evidence. The facts as currently understood can be accounted for by our model given one simple assumption: While children summarize their exposure to verb usages as discussed above, they will use those verbs in highly productive alternations (as if they were in lower categories) for some period after exposure to the verb. The claim is that their em usage might be non-conservative, even if their representations of verb class are. By this model, the child would restrict the usage of a given verb to the represented usages only after some period of time. The mechanisms for deriving criteria for productive usage of verb patterns described by Pinker(1987) could also be added to our model without difficulty. In essence, one would then have a non-conservative learner with a conservative core.</Paragraph> </Section> class="xml-element"></Paper>