File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/05/w05-1005_metho.xml
Size: 17,032 bytes
Last Modified: 2025-10-06 14:09:59
<?xml version="1.0" standalone="yes"?> <Paper uid="W05-1005"> <Title>Automatically Distinguishing Literal and Figurative Usages of Highly Polysemous Verbs</Title> <Section position="3" start_page="38" end_page="40" type="metho"> <SectionTitle> 2 Compositionality of Light Verbs </SectionTitle> <Paragraph position="0"/> <Section position="1" start_page="38" end_page="39" type="sub_section"> <SectionTitle> 2.1 Linguistic Properties: Syntactic Flexibility </SectionTitle> <Paragraph position="0"> We focus on a broadly-documented subclass of light verb constructions, in which the complement is an activity noun that is often the main source of semantic predication (Wierzbicka, 1982). Such complements are assumed to be indefinite, non-referential predicative nominals (PNs) that are often morphologically related to a verb (see the complements in examples (1a-c) above). We refer to this class of light verb constructions as &quot;LV+PN&quot; constructions, or simply LVCs.</Paragraph> <Paragraph position="1"> There is much linguistic evidence that semantic properties of a lexical item determine, to a large extent, its syntactic behaviour (e.g., Rappaport Hovav and Levin, 1998). In particular, the degree of compositionality (decomposability) of a multiword expression has been known to affect its participation in syntactic transformations, i.e., its syntactic flexibility (e.g., Nunberg et al., 1994). English &quot;LV+PN&quot; constructions enforce certain restrictions on the syntactic freedom of their noun components (Kearns, 2002). In some, the noun may be introduced by a definite article, pluralized, passivized, relativized, or even wh-questioned: give a book give a present give money give rightgive advice give opportunity give orders give permission give a speech give a smile give a laugh give a yell give a groan give a sweep give a push give a dust give a wipe give a pull give a kick more figurative give a bookgive a present give money give a wipe give a sweep give a dust give a push give a kick give a pull give orders give a speech give advice give permission give right give opportunity give a yell give a laugh give a groan give a smile 2. (a) Azin gave a speech to a few students. (b) Azin gave the speech just now. (c) Azin gave a couple of speeches last night. (d) A speech was given by Azin just now. (e) Which speech did Azin give? Others have little or no syntactic freedom: 3. (a) Azin gave a groan just now. (b) * Azin gave the groan just now. (c) ? Azin gave a couple of groans last night. (d) * A groan was given by Azin just now. (e) * Which groan did Azin give? Recall that give in give a groan is presumed to be a more abstract usage than give in give a speech. In general, the degree to which the light verb retains aspects of its literal meaning--and contributes them compositionally to the LVC--is reflected in the degree of syntactic freedom exhibited by the LVC. We exploit this insight to devise a statistical measure of compositionality, which uses evidence of syntactic (in)flexibility of a potential LVC to situate it on a scale of literal to figurative usage of the light verb: i.e., the more inflexible the expression, the more figurative (less compositional) the meaning.</Paragraph> </Section> <Section position="2" start_page="39" end_page="40" type="sub_section"> <SectionTitle> 2.2 A Statistical Measure of Compositionality </SectionTitle> <Paragraph position="0"> Our proposed measure quantifies the degree of syntactic flexibility of a light verb usage by looking at its frequency of occurrence in any of a set of relevant syntactic patterns, such as those in examples (2) and (3). The measure, COMP That is, the greater the association between LV and N, and the greater the difference between their association with positive syntactic patterns and negative syntactic patterns, the more figurative the meaning of the light verb, and the higher the score.</Paragraph> <Paragraph position="1"> The strength of the association between the light verb and the complement noun is measured using pointwise mutual information (PMI) whose standard formula is given here:1</Paragraph> <Paragraph position="3"> where n is an estimate of the total number of verb and object noun pairs in the corpus.</Paragraph> <Paragraph position="4"> PSpos represents the set of syntactic patterns preferred by less-compositional (more figurative) LVCs (e.g., as in (3a)), and PSneg represents less preferred patterns (e.g., those in (3b-e)). Typically, these patterns most affect the expression of the complement noun. Thus, to measure the strength of association between an expression and a set of patterns, we use the PMI of the light verb, and the complement noun appearing in all of the patterns in the set, as in:</Paragraph> <Paragraph position="6"> in which counts of occurrences of N in syntactic contexts represented by PSpos are summed over all patterns in the set. ASSOC(LV;Na1 PSneg) is defined analogously using PSneg in place of PSpos.</Paragraph> <Paragraph position="7"> DIFF measures the difference between the association strengths of the positive and negative pattern sets, referred to as ASSOC pos and ASSOCneg, respectively. Our calculation of ASSOC uses maximum likelihood estimates of the true probabilities. To account for resulting errors, we compare the two confidence intervals, a1 ASSOC pos</Paragraph> <Paragraph position="9"> the minimum distance between the two as a conservative estimate of the true difference: Taking the difference between confidence intervals lessens the effect of differences that are not statistically significant. (The confidence level, 1 a, is set to 95% in all experiments.)</Paragraph> </Section> </Section> <Section position="4" start_page="40" end_page="41" type="metho"> <SectionTitle> 3 Acceptability Across Semantic Classes </SectionTitle> <Paragraph position="0"/> <Section position="1" start_page="40" end_page="40" type="sub_section"> <SectionTitle> 3.1 Linguistic Properties: Class Behaviour </SectionTitle> <Paragraph position="0"> In this aspect of our work, we narrow our focus onto a subclass of &quot;LV+PN&quot; constructions that have a PN complement in a stem form identical to a verb, preceded (typically) by an indefinite determiner (as in (1a-b) above). Kearns (2002), Wierzbicka (1982), and others have noted that the way in which LVs combine with such PNs to form acceptable LVCs is semantically patterned--that is, PNs with similar semantics appear to have the same trends of cooccurrence with an LV.</Paragraph> <Paragraph position="1"> Our hypothesis is that semantically similar LVCs--i.e., those formed from an LV plus any of a set of semantically similar PNs--distinguish a figurative subsense of the LV. In the long run, if this is true, it could be exploited by using class information to extend our knowledge of acceptable LVCs and their likely meaning (cf. such an approach to verb particle constructions by Villavicencio, 2003).</Paragraph> <Paragraph position="2"> As steps to achieving this long-term goal, we must first devise an acceptability measure which determines, for a given LV, which PNs it successfully combines with. We can even use this measure to provide evidence on whether the hypothesized class-based behaviour holds, by seeing if the measure exhibits differing behaviour across semantic classes of potential complements.</Paragraph> </Section> <Section position="2" start_page="40" end_page="41" type="sub_section"> <SectionTitle> 3.2 A Statistical Measure of Acceptability </SectionTitle> <Paragraph position="0"> We develop a probability formula that captures the likelihood of a given LV and PN forming an acceptable LVC. The probability depends on both the LV and the PN, and on these elements being used in an observation that higher frequency words are more likely to be used as LVC complements (Wierzbicka, 1982). We estimate this factor by f a0 PNa2a7a6 n, where n is the number of words in the corpus.</Paragraph> <Paragraph position="1"> The probability that a given LV and PN form an acceptable LVC further depends on how likely it is that the PN combines with any light verbs to form an LVC. The frequency with which a PN forms LVCs is estimated as the number of times we observe it in the prototypical &quot;LV a/an PN&quot; pattern across LVs. (Note that such counts are an overestimate, since we cannot determine which usages are indeed LVCs vs. literal uses of the LV.) Since these counts consider the PN only in the context of an indefinite determiner, we normalize over counts of &quot;a/an PN&quot; (noted as aPN) to form the conditional probability estimate of the second factor:</Paragraph> <Paragraph position="3"> where v is the number of light verbs considered.</Paragraph> <Paragraph position="4"> The third factor, Pra0 LV a5 PNa1 LVCa2 , reflects that different LVs have varying degrees of acceptability when used with a given PN in an LVC. We similarly estimate this factor with counts of the given LV and PN in the typical LVC pattern: f a0 LVa1 aPNa2a7a6 f a0 aPNa2 . Combining the estimates of the three factors yields:</Paragraph> <Paragraph position="6"/> </Section> </Section> <Section position="5" start_page="41" end_page="42" type="metho"> <SectionTitle> 4 Materials and Methods </SectionTitle> <Paragraph position="0"/> <Section position="1" start_page="41" end_page="41" type="sub_section"> <SectionTitle> 4.1 Light Verbs </SectionTitle> <Paragraph position="0"> Common light verbs in English include give, take, make, get, have, and do, among others. We focus here on two of them, i.e., give and take, that are frequently and productively used in light verb constructions, and are highly polysemous. The Word-Net polysemy count (number of different senses) of give and take are 44 and 42, respectively.</Paragraph> </Section> <Section position="2" start_page="41" end_page="41" type="sub_section"> <SectionTitle> 4.2 Experimental Expressions </SectionTitle> <Paragraph position="0"> Experimental expressions--i.e., potential LVCs using give and take--are drawn from two sources.</Paragraph> <Paragraph position="1"> The development and test data used in experiments of compositionality (bncD and bncT, respectively) are randomly extracted from the BNC (BNC Reference Guide, 2000), yielding expressions covering a wide range of figurative usages of give and take, with complements from different semantic categories. In contrast, in experiments that involve acceptability, we need figurative usages of &quot;the same type&quot;, i.e., with semantically similar complement nouns, to further examine our hypothesis on the class-based behaviour of light verb combinations.</Paragraph> <Paragraph position="2"> Since in these LVCs the complement is a predicative noun in stem form identical to a verb, we form development and test expressions by combining give or take with verbs from selected semantic classes of Levin (1993), taken from Stevenson et al. (2004).</Paragraph> </Section> <Section position="3" start_page="41" end_page="41" type="sub_section"> <SectionTitle> 4.3 Corpora </SectionTitle> <Paragraph position="0"> We gather estimates for our COMP measure from the BNC, processed using the Collins parser (Collins, 1999) and TGrep2 (Rohde, 2004). Because some LVCs can be rare in classical corpora, our ACPT estimates are drawn from the World Wide Web (the subsection indexed by AltaVista). In our comparison of the two measures, we use web data for both, using a simplified version of COMP. The high level of noise on the web will influence the performance of both measures, but COMP more severely, due to its reliance on comparisons of syntactic patterns.</Paragraph> <Paragraph position="1"> Web counts are based on an exact-phrase query to AltaVista, with the number of pages containing the search phrase recorded as its frequency.2 The size of the corpus is estimated at 3.7 billion, the number of hits returned in a search for the. These counts are underestimates of the true frequencies, as a phrase may appear more than once in a web page, but we assume all counts to be similarly affected.</Paragraph> </Section> <Section position="4" start_page="41" end_page="42" type="sub_section"> <SectionTitle> 4.4 Extraction </SectionTitle> <Paragraph position="0"> Most required frequencies are simple counts of a word or string of words, but the syntactic patterns used in the compositionality measure present some complexity. Recall that PSpos and PSneg are pattern sets representing the syntactic contexts of interest.</Paragraph> <Paragraph position="1"> Each pattern encodes several syntactic attributes: v, the voice of the extracted expression (active or passive); d, the type of the determiner introducing N (definite or indefinite); and n, the number of N (singular or plural). In our experiments, the set of patterns associated with less-compositional use, PSpos, consists of the single pattern with values active, indefinite, and singular, for these attributes. PSneg consists of all patterns with at least one of these attributes having the alternative value.</Paragraph> <Paragraph position="2"> While our counts on the BNC can use syntactic mark-up, it is not feasible to collect counts on the web for some of the pattern attributes, such as voice. We develop two different variations of the measure, one for BNC counts, and a simpler one for respect to human compositionality ratings.</Paragraph> <Paragraph position="3"> web counts. We thus subscript COMP with abbreviations standing for each attribute in the measure: COMPvdn for a measure involving all three attributes (used on BNC data), and COMPd for a measure involving determiner type only (used on web data).</Paragraph> </Section> </Section> <Section position="6" start_page="42" end_page="42" type="metho"> <SectionTitle> 5 Human Judgments </SectionTitle> <Paragraph position="0"/> <Section position="1" start_page="42" end_page="42" type="sub_section"> <SectionTitle> 5.1 Judgments of Compositionality </SectionTitle> <Paragraph position="0"> To determine how well our proposed measure of compositionality captures the degree of literal/figurative use of a light verb, we compare its scores to human judgments on compositionality.</Paragraph> <Paragraph position="1"> Three judges (native speakers of English with sufficient linguistic knowledge) answered yes/no questions related to the contribution of the literal meaning of the light verb within each experimental expression. The combination of answers to these questions is transformed to numerical ratings, ranging from 0 (fully non-compositional) to 4 (largely compositional). The three sets of ratings yield linearly weighted Kappa values of .34 and .70 for give and take, respectively. The ratings are averaged to form a consensus set to be used for evaluation.3 The lists of rated expressions were biased toward figurative usages of give and take. To achieve a spectrum of literal to figurative usages, we augment the lists with literal expressions having an average rating of 5 (fully compositional). Table 2 shows the distribution of the experimental expressions across three intervals of compositionality degree, 'low' (ratings</Paragraph> <Paragraph position="3"> ings a2 3). Table 3 presents sample expressions with different levels of compositionality ratings.</Paragraph> <Paragraph position="4"> 3We asked the judges to provide short paraphrases for each expression, and only use those expressions for which the majority of judges expressed the same sense.</Paragraph> </Section> <Section position="2" start_page="42" end_page="42" type="sub_section"> <SectionTitle> Sample Expressions </SectionTitle> <Paragraph position="0"> Human Ratings give take 'low' give a squeeze take a shower 'medium' give help take a course 'high' give a dose take an amount</Paragraph> </Section> <Section position="3" start_page="42" end_page="42" type="sub_section"> <SectionTitle> 5.2 Judgments of Acceptability </SectionTitle> <Paragraph position="0"> Our acceptability measure is compared to the human judgments gathered by Stevenson et al. (2004).</Paragraph> <Paragraph position="1"> Two expert native speakers of English rated the acceptability of each potential &quot;LV+PN&quot; construction generated by combining give and take with candidate complements from the development and test Levin classes. Ratings were from 1 (unacceptable) to 5 (completely natural; this was capped at 4 for test data), allowing for &quot;in-between&quot; ratings as well, such as 2.5. On test data, the two sets of ratings yielded linearly weighted Kappa values of .39 and .72 for give and take, respectively. (Interestingly, a similar agreement pattern is found in our human compositionality judgments above.) The consensus set of ratings was formed from an average of the two sets of ratings, once disagreements of more than one point were discussed.</Paragraph> </Section> </Section> class="xml-element"></Paper>