Eric Brill and Philip Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Each noun with a possible semantic class of act or process in Wordnet [Miller, 1990] and that noun s arguments) can likewise be deemed a base level clause. The Processing Model Our system uses transformnation based error driven learning to automatically learn rules fi om training ex amples [Brill and Resnik, 1994]. One first runs the system on a training set, which starts with no grammatical relations marked. This training run moves in iterations with each iteration producing the next rule that yields the best bet gain in the training set (number of matching relationships found minus the number of ....

....text versus narratives) and different standards of grammatical relationships. An additional item of experimentation is to use our sys tem to adapt other systems, including earlier versions of our system, to these differing styles and standards. Like other Brill transformation rule systems [Brill and Resnik, 1994], our system can take in the output of another system and try to improve on it. This suggests a relatively low expense method to adapt a hard to alter system that performs well on a slightly different style or standm d. Our training approach accepts as a starting point an initial labeling of the ....

E. Brill aud P. Rsnik. A rule- based approach to prepositional phrase attachment disambiguation. In 15th International Conf on Computational Linguistics (COLING). 1994.

....synonymy sets. In (Ratnaparkhi et al., 1994) word classes are derived automatically with a clustering procedure. Franz, 1995) uses a loglinear model to estimate preferred attachments according to the linguistic features of co occurring words (e.g. bigrams, the accompanying noun determiner, etc. (Brill and Resnik, 1994) use transformationbased error driven learning (Brill, 1992) to derive dis ambiguation rules based on simple context information (e.g. right and left adjacent words or POSs) All these approaches need extensive collections of positive examples (i.e. hand corrected attachment instances) in order ....

Brill E., Resnik P., A rule-based ap- proach to prepositional phrase attachment disambiguation, in Proc. of COLING 94, 1198-1204

....We examine the different ways in which with can be used, and look at the methods that can be used to help determine the correct attachment point. In particular, we focus on the possibility that correct predictions can be made given the verb and its objects of the preposition. Experiments such as [BR94] and [WAB 91] seem to suggest that this may be sufficient information to produce good results, but psycholinguistic research implies that other information such as real world knowledge and contextual clues play a part in human sentence processing. We predict that if this type of information ....

....phrase attachment ambiguities by means of statistical approaches involving semantic word classes, e.g. WAB 91] BPV91] Bri93] HR93] These probabilistic models produce the interpretation with the greatest likelihood of occurrence. 26 3. 1 Rule Based Approaches Brill and Resnik [BR94] describe a rule based approach to disambiguation of prepositional phrase attachment, which uses information automatically extracted from a large corpus. Their approach relies on a rule based approach which attempts to predict the correct attachment of the preposition, with the aid of contextual ....

[Article contains additional citation context not shown here]

E. Brill and P. Resnik. A rule-based approach to prepositional phrase attachment disambiguation. In Proceedings of the Fifteenth International Conference on Computational Linguistics (COLING-1994), Kyoto, Japan, 1994.

.... In past research, methods of resolving ambiguities have been based, for example, on the assumption that case slots are mutually independent (Hindle and Rooth, 1991; Sekine et al. 1992; Resnik, 1993a; Grishman and Sterling, 1994; Alshawi and Carter, 1994) or at most two case slots are dependent (Brill and Resnik, 1994; Ratnaparkhi, Reynar, and Roukos, 1994; Collins and Brooks, 1995) 2.5 Structural Disambiguation 2.5.1 The lexical approach There have been many probabilistic methods proposed in the literature to address the structural disambiguation problem. Some methods tackle the basic problem of resolving ....

.... the conditional distribution by using the Maximum Entropy Estimation technique (see (Jaynes, 1978; Darroch and Ratcli#, 1972; Berger, Pietra, and Pietra, 1996) Another method of the quadruple approach is to employ transformation based error driven learning (Brill, 1995) as proposed in (Brill and Resnik, 1994). This method learns and uses IF THEN type rules, where the IF parts represent conditions like (p is with ) and (v is see ) and the THEN parts represent transformations from (attach to v) to (attach to n 1 ) and vice versa. The first rule is always a default decision, and all the other rules ....

[Article contains additional citation context not shown here]

Brill, Eric and Philip Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation. Proceedings of the 15th International Conference on Computational Linguistics, pages 1198--1204.

.... uses the robust dependency based parser Link Grammar (LG) 16] which is able to handle a wide range of syntactic structures [17] Syntactically unresolvable ambiguities, such as prepositional phrase attachment or gerund and infinitive constructions, are treated with a corpus based approach [2]. Sentenceinternal pronouns are dealt with using the anaphora resolution algorithm [11] LG uses linkages to describe the syntactic structure of a sentence (see figure 2) Links connect pairs of words in such a way that the requirements of each word described in the sentences are satisfied, that ....

.... which we also base our MLFs) The syntax analysis in Falcon is based on a statistical parser [4] while we use a dependency parser that computes all syntactically possible structures which we then filter according to a combination of hand crafted rules and Brill and Resnik disambiguation procedure [2]. A similarity between ExtrAns and Falcon is that both build a semantic form starting from a dependency based representation of the questions. As for the type of inferencing used, while ExtrAns uses standard deduction (proving questions over documents) Falcon uses an abductive backchaining ....

Eric Brill and Philip Resnik, `A rule-based approach to prepositional phrase attachment disambiguation', in Proc. COLING '94, volume 2, pp. 998--1004, Kyoto, Japan, (1994).

.... of ML for NL include works on computational learnability of linguistic knowledge ( 4, 5] or example driven learning applied to text classification and document processing ( 14] More recent approaches span from very specific language processing tasks (e.g. induction of PP disambiguation rules [8], ML driven subcategorization frames acquisition [2] or case based discourse modeling [9] to integration of symbolic induction and quantitative techniques for the construction of extensive components of lexical knowledge bases (e.g. 17, 12, 3] 2 Adapting IE systems to domains and users In IE ....

....of the two sources of information is obtained by a linear combination via the coefficents 1 and 2 . As a result of this phase lexical semantic patterns of the form sem patt( break, subj,obj] capitalist : entity something ] law law of nature : psychological feature ] [8,17,45]) or semp patt( earn, subj,obj,in] person individual someone somebody mortal human soul : entity something ] information : psychological feature ] time period period period of time amount of time : abstraction ] 7,34] are derived. The last argument of the ....

E. Brill and P. Resnik, `A rule based approach to prepositional phrase attachment disambiguation', in Proc. of COLING-94, Kyoto, Japan, (1994).

....However, we observe that these conjunctive features are the common features used anyhow by most learning algorithms applied to various disambiguation and predictions tasks in NLP. For example, various methods have been used to study the PPA problem discussed in Ex. 3. 1 [ Collins and Brooks, 1995; Brill and Resnik, 1994; Ratnaparkhi et al. 1994; Krymolowski and Roth, 1998 ] All of them make use of essentially the same conjunctive features presented here. The same holds for many other problems. We conclude that MBL methods are linear over feature spaces that are commonly used for learning in NLP. Space ....

....make no abstractions. In terms of the vector representations z constructed in Sec. 3 that means that every feature that is observed active in the training set will have a non zero coe cient, and may contribute to future predictions. Algorithms like Transformations Based Learning [ Brill, 1995; Brill and Resnik, 1994 ] on the other hand, start with the same set of features as does MBL, but prune the feature space very aggressively, with the goal of outputting a very small list of active features. Bayesian algorithms, Golding, 1995; Gale et al. 1993 ] on the other hand, keep all the features alive, only ....

E. Brill and P. Resnik. A rulebased approach to prepositional phrase attachment disambiguation. In Proc. of COLING, 1994.

.... will offer another automatic method of predicting abstract prosodic labels based on transformational rule based learning (TRBL) TRBL is currently used in the natural language processing community in a number of applications, such 3 as: part of speech tagging [17] prepositional phrase attachment [16], word segmentation [38] discourse processing [48] spelling correction [30] and partial parsing or chunking [43] 1 . There are four advantages to TRBL. First, it allows for automatic training. Second, the resulting model is easy to examine and can provide intuitions about the underlying ....

....the contemporary stochastic methods of part of speech classification, often with less data (64k words vs. 1M) TRBL has also been shown to have equivalent or greater success than other statistical or rule based algorithms in a number of diverse NLP tasks such as prepositional phrase attachment [16], partial parsing or chunking [43] dialogue act identification [48] word segmentation [38] and spell checking [30] TRBL has three basic requirements to define the system [17] an initial state annotator, a set of allowable questions and transformations, and a function for ranking potential ....

Brill, E. and Resnick, P.(1994). "A Rule-Based Approach to Prepositional Phrase Attachment Disambiguation." Proceedings of the Sixteenth International Conference on Computational Linguistics.

....in Cats eat. there is a subject relation that has eat as the target and Cats as the source. ing. 2 Di#erences Between the Two Systems Ferro et al. 1999) and Buchholz et al. 1999) both describe learning systems to find GRs. The former (TR) uses transformation based error driven learning (Brill and Resnik, 1994) and the latter (MB) uses memory based learning (Daelemans et al. 1999) In addition, there are other di#erences. The TR system includes several types of information not used in the MB system (some because memory based systems have a harder time handling set valued attributes) possible ....

E. Brill and P. Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation. In 15th International Conf.

.... size of computing machinery to accommodate larger models, the availability of resources such as the Penn Treebank (Marcus et al. 1993) and the success of machine learning techniques for lowerlevel NLP problems, such as part of speech tagging (Church, 1988; Brill, 1995) and PPattachment (Brill and Resnik, 1994; Collins and Brooks, 1995) However, perhaps even more signi cant has been the lexicalization of the grammar formalisms being probabilistically modeled: crucially, all the recent, successful statistical parsers have in some way made use of bilexical dependencies. This includes both the parsers ....

E. Brill and P. Resnik. 1994. A rulebased approach to prepositional phrase attachment disambiguation. In Fifteenth International Conference on Computational Linguistics (COLING-1994).

....F scores increased to about 70 . An alternative to performing translations is to use the untranslated initial annotations as an additional type of input to the rule system. This alternative, which we have yet to try, has the advantage of tting into the transformation based error driven paradigm (Brill and Resnik, 1994) more cleanly than having a translation stage. However, this additional type of input will also further slowdown an already slow rule learning module. 2.2 Overall Results For our experiment, we use the same 1151 word (748 GR) test set used in Ferro et al. 1999) but for a training set, we use ....

E. Brill and P. Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation. In 15th International Conf. on Computational Linguistics (COLING).

....step the rule that best repairs the current errors. Concrete rules are acquired by instantiation of a predefined set of template rules. This algorithm has been applied to a number of natural language problems, including part of speech tagging [21, 22, 23, 6, 186] PP attachment disambiguation [20], parsing [18] spelling correction [133] and word sense disambiguation [67] One major drawback of TBL is its computational cost since all instantiations of templates are tested at each iteration to find the best rule. Recently, Samuel [191] presented an efficient approximation called Lazy TBL ....

.... [181] Shallow Parsing [45, 1, 16, 212] 211] 7, 227, 33, 58] Parsing [12, 132, 92] 178] 210, 37, 36, 38] PP attachment disambiguation [180] 246] TBL NB NNs LSM EC Clause Boudaries [95] Shallow Parsing [18] 128, 129] 130] 223] Parsing [115, 43] 93] PP attachment disambiguation [20] [52] 125, 218] 107] 2] Table 2: References corresponding to syntactic analysis and structural ambiguity NLP problems DLs DTs NB TBL EM WSD [240, 150] 26, 150] 86, 150, 112] 67] 203, 166] Text categorization and filtering [117, 69, 230] 117, 190, 119, 142, 196] 162, 163] ....

E. Brill and P. Resnik. A Rule-Based Approach to Prepositional Phrase Attachment Disambiguation. In Proceedings of the 15th International Conference on Computational Linguistics, COLING, Kyoto, Japan, August 1994. http://www.cs.jhu.edu/brill/acadpubs.html.

....can be used as we will see later by the retrieval procedure. 1.4. Disambiguator ExtrAns pruner uses heuristic rules to filter out all those dependency structures that are obviously wrong in an ambiguous sentence. In a subsequent stage, ExtrAns disambiguator uses a corpus based approach (Brill E. Resnik P. 1994) to eliminate (some of) those ambiguities that require domain knowledge. Brill Resnik s original disambiguator was designed to solve prepositional attachment ambiguity of sentences with a transitive verb and a prepositional phrase (PP) The algorithm decides whether the PP should attach to the ....

BRILL, Eric ; RESNIK, Philip (1994) : "A rule-based approach to prepositional phrase attachment disambiguation", in Proc. COLING '94, pp. 998--1004, Kyoto, Japan.

.... the size of the training data as learning progresses, such as is the case in decision tree induction (Quinlan 1986) Brill s work provides a proof of viability of transformation based techniques in the form of a number of processors, including a (widely distributed) part of speech tagger (Brill 1994, 1995) a procedure for prepositional phrase attachment (Brill Resnick 1994) and a bracketing parser (Brill 1993b) All of these provided performance comparable to or better than previous attempts. Transformation based learning has also been successfully applied to text chunking (Ramshaw ....

.... as is the case in decision tree induction (Quinlan 1986) Brill s work provides a proof of viability of transformation based techniques in the form of a number of processors, including a (widely distributed) part of speech tagger (Brill 1994, 1995) a procedure for prepositional phrase attachment (Brill Resnick 1994), and a bracketing parser (Brill 1993b) All of these provided performance comparable to or better than previous attempts. Transformation based learning has also been successfully applied to text chunking (Ramshaw Marcus 1995) morphological disambiguation (Oflazer Tur 1996 ) and phrase ....

Brill, e. & Resnick, p . (1994 ) "A rule-based approach to prepositional phrase attachment disambiguation". In Prcdgs. of the 15th Intl. Conference on Computational Linguistics (coling-94).

....by the pruner would need modi cation. 3 Disambiguation based on corpus information There are certain types of ambiguity that cannot be solved with syntactic information only. This is generally the case with attachment ambiguity. ExtrAns disambiguator module uses Brill Resnik s algorithm [2] to solve attachment ambiguity. Brill Resnik s algorithm follows a corpusbased approach. In particular, the algorithm is trained with a corpus and creates a set of rules that can be used to resolve the attachment ambiguity in sentences with a transitive verb and a prepositional phrase (PP) The ....

....the disambiguator (the rst piece of information) is compared with the decision made in the linkage (the second piece of information, shown in boxes above) Those linkages that show a highest ratio of correct attachment decisions pass the lter, and the rest are rejected. The original algorithm [2] can also be used to cover other types of attachment ambiguity, provided that the training corpus o ers the data necessary to cover them. The following types of ambiguity have been incorporated: V N particle VP. I ran the program by typing run program by type V N particle ....

Eric Brill and Philip Resnik. A rule-based approach to prepositional phrase attachment disambiguation. In Proceedings of the 15th International Conference on Computational Linguistics (COLING '94), volume 2, pages 998-1004, Kyoto, Japan, 1994.

.... Link Grammar does not carry out any morphological analysis of the words, ExtrAns uses a lemmatiser that generates the lemma of the inflected words [Humphreys et al. 96] Ambiguous attachments of prepositional phrases are disambiguated by a corpus based approach trained with data from the manpages [Brill Resnik 94] Anaphoric references are resolved using an adaptation of a purely syntactic approach [Lappin Leass 94] To finally construct the MLFs, ExtrAns transforms the syntax structures generated by Link Grammar ( linkages ) into a directed dependency structure. MLFs are converted into Horn clause ....

E. Brill, P. Resnik, A rule-based approach to prepositional phrase attachment disambiguation, Proceedings of the 15th International Conference on Computational Linguistics (COLING'94), Vol. 2, pp. 998-1204, Kyoto, Japan, 1994.

....set of rules or restrictions for matching human expectations in interpretation. The intractability of heuristic methods, and the recent explosion of availability of on line text, have inspired a number of researchers to derive interpretation preferences from corpus analysis (e.g. Bod, 1995; Brill and Resnik, 1994; Hindle and Rooth, 1993; Weischedel, Meteer, Schwartz, Ramshaw, and Palmucci, 1993) The extraction of relevant statistics from large corpora has been claimed to enable us to determine human preferences in a more general and accurate manner than previously achieved. However, a number of studies ....

Brill, E. and P. Resnik (1994). A rule-based approach to prepositional phrase attachment disambiguation. In Proceedings of the 15th International Conference on Computational Linguistics (COLING-94).

....of tokens in a sentence. However, we observe that these conjunctive features are the common features used anyhow by most learning algorithms applied to various disambiguation and predictions tasks in NLP. For example, various methods have been used to study the PPA problem discussed in Ex. 3. 1 [4, 3, 12, 11]. All of them make use of essentially the same conjunctive features presented here. The same holds for many other problems. We conclude that MBL methods are linear over feature spaces that are commonly used for learning in NLP. Space Considerations The bound on the total number of features that ....

....[6] MBL algorithms make no abstractions. In terms of the vector representations z constructed in Sec. 3 that means that every feature that is observed active in the training set will have a non zero coe cient, and may contribute to future predictions. Algorithms like Transformations Based Learning [2, 3], on the other hand, start with the same set of features as does MBL, but prune the feature space very aggressively, with the goal of outputting a very small list of active features. Bayesian algorithms, 8, 7] on the other hand, keep all the features alive, only that they reweigh them as a ....

E. Brill and P. Resnik. A rule-based approach to prepositional phrase attachment disambiguation. In Proc. of COLING, 1994.

....set of our methods, as compared to the figures in the literature. Maximum Entropy is the model applied to this data set in Ratnaparkhi, Reynar, and Roukos (1994) Transformations is the figure cited in Collins and Brooks (1995) for the performance of Brill s transformation based learning method (Brill and Resnik (1994), Brill (1993) on this data set. Back off model is the model presented in Collins and Brooks (1995) The data on human performance are taken from Ratnaparkhi, Reynar, and Roukos (1994) performed best when k was set to 13, MVDM obtained its best score when k was set to 5. In the LexSpace based ....

....a paired t test. The LexSpace metric outperforms the Overlap metric, but shows no significant advantage over the MVDM metric. 6 Discussion Comparison with related work. A comparison with methods from the literature could be performed in the case of the PP attachment task, because other methods (Brill and Resnik (1994), Ratnaparkhi, Reynar, and Roukos (1994) and Collins and Brooks (1995) Method1 Method2 p value Overlap Overlap IG ratio 0 Overlap MVDM.k3 0.0012 Overlap IG ratio MVDM.k3 0.0115 Overlap LexSpace.k5 0.0007 Overlap IG ratio LexSpace.k5 0.6969 MVDM.k3 LexSpace.k5 0.4366 Table 6: The statistical ....

[Article contains additional citation context not shown here]

Brill, E. and P. Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation. In Proceedings of the 15th International Conference on Computational Linguistics (COLING'94).

....(Witten et al. 1994) A standard query to retrieve articles from the domain will be used, for example, in the management successions domain the query might be: chief executive officer head president chairman post succeed name ii. The corpus will be part of speech tagged using the Brill tagger (Brill 1994) and a simple frequency count of the nouns and verbs produced. After filtering common nouns and verbs (for example, forms of to be ) obtained from a standard stop list, the most frequent nouns verbs will provide an initial set of relevant single word nouns verbs which will be considered in ....

....found in large thesaura or on line dictionaries. The lexical information that trigger induction varies from simple word tokens to syntactically annotated or semantically typed collocations (e.g. powerful vs. strong tea (Smadja,1989) syntactic disambiguation rules (e.g. Hindle and Rooths,1993) (Brill and Resnik,1994)) or sense disambiguation rules (Yarowsky,1992) McRoy,1992) Such information is lexical as it encodes constraints (of different types) at the word level, to be thus inherited by morphologic variants of a given lemma. This strongly lexicalized knowledge, as it is extracted from corpus data, ....

Brill E., Resnik P.,1994, A rule-based approach to prepositional phrase attachment disambiguation, in Proc. of COLING 94, 1198-1204 Brill, E. 1994. Some advances in transformation based part of speech tagging. Proceedings of the Twelfth National Conference on Artificial Intelligence (AAAI-94). Cambridge, Massachusetts: AAAI Press/MIT Press.

....or to the VP joined , giving two alternative structures. In this case the VP attachment is correct) NP attach: joined ( the board) as a nonexecutive director) VP attach: joined (the board) as a nonexecutive director) Work by Ratnaparkhi, Reynar and Roukos [RRR94] and Brill and Resnik [BR94] has considered corpus based approaches to this problem, using a set of examples to train a model which is then used to make attachment decisions on test data. Both papers describe methods which look at the four head words involved in the attachment the VP head, the first NP head, the ....

....the attachment decision for 880 test examples, and the method performed at 80 accuracy on these cases. Note that it is difficult to compare this result to results on Wall Street Journal, as the two corpora may be quite different. The Wall Street Journal Treebank [MSM93] enabled both [RRR94] and [BR94] to extract a large amount of supervised training material for the problem. Both of these methods consider the second noun, n2, as well as v, n1 and p, with the hope that this additional information will improve results. BR94] use 12,000 training and 500 test examples. A greedy search is used to ....

[Article contains additional citation context not shown here]

E. Brill and P. Resnik. A Rule-Based Approach to Prepositional Phrase Attachment Disambiguation. In Proceedings of the fifteenth international conference on computational linguistics (COLING-1994), 1994.

....In its most general setting, the TBL hypothesis is not a classifier (Brill 1995) The reason is that the truth value of the condition of the ith rule may change while evaluating one of the preceding rules. However, in many applications and, in particular, in Spell (Mangu Brill 1997) and PPA (Brill Resnik 1994) which we discuss later, this is not the case. There, the conditions do not depend on the labels, and therefore the output hypothesis of the TBL method can be viewed as a classifier. The following analysis applies only for this case. Using the terminology introduced above, let (x i 1 ; c i 1 ) x ....

....3097 previously unseen examples from this corpus; all the system use the same feature set. Test Baseline NB TBL BO SNOW cases 3097 59.0 83.0 81.9 84.1 83.9 the steering wheel or the verb phrase, as in Buy the car with his money. Earlier works on this problem (Ratnaparkhi, Reynar, Roukos 1994; Brill Resnik 1994; Collins Brooks 1995) consider as input the four head words involved in the attachment the VP head, the first NP head, the preposition and the second NP head (in this case, buy, car, with and steering wheel, respectively) These four tuples, along with the attachment decision constitute the ....

Brill, E., and Resnik, P. 1994. A rule-based approach to prepositional phrase attachment disambiguation. In Proc.

....second noun and therefore could not distinguish between sentence 1 and 2. Their method is also difficult to extend to more elaborate combinations of information sources. More recently, a number of statistical methods better suited to larger numbers of features have been proposed for PP attachment. Brill and Resnik (1994) applied Error Driven Transformation Based Learning, Ratnaparkhi, Reynar and Roukos (1994) applied a Maximum Entropy model, Franz (1996) used a Loglinear model, and Collins and Brooks (1995) obtained good results using a Back Off model. In this paper, we examine whether Memory Based Learning ....

....process is much more laborious than the automatic computation of IG weights on the training set. The other methods for which results have been reported on this dataset include decision trees, Maximum Entropy (Ratnaparkhi, Reynar, and Roukos, 1994) and Error Driven TransformationBased Learning (Brill and Resnik, 1994), 3 which 3 The results of Brill s method on the present benchmark were reconstructed by Collins and Brooks (1995) were clearly outperformed by both IB1 and IB1 IG, even though e.g. Brill Resnik used more elaborate feature sets (words and WordNet classes) Adding more elaborate features is ....

[Article contains additional citation context not shown here]

Brill, E. and P. Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation.

....(PP) attaches to the direct object NP as in Buy the car with the steering wheel (nattachment) or to the verb phrase buy, as in Buy the car with his money (v attachment) PPA is a common cause of structural ambiguity in natural language. Earlier works on this problem (Ratnaparkhi et al. 1994; Brill and Resnik, 1994; Collins and Brooks, 1995; Zavrel et al. 1997) represented an example by the 4 tuple v, n1, p, n2 containing the VP head, the direct object NP head, the preposition, and the indirect object NP head respectively. The first example in the previous paragraph is thus represented by buy, car, ....

....of the class information may be irrelevant or even slightly misleading. The results provide an assesment of the relative relevance of each knowledge source. When a noun belongs to a class, one may replace the explicit noun feature by its classes. Using the classes in addition to the original noun (Brill and Resnik, 1994; Resnik, 1992; Resnik, 1995) seems, however, a better strategy. Consider, for example, the feature prep,indirect object=n2 . Suppose the noun n2 belongs to two classes c1 and c2. The class information will be incorporated by creating two additional features: prep,indirect object=c1 and ....

[Article contains additional citation context not shown here]

E. Brill and P. Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation. In Proc. of COLING.

....to specify the initial state, the space of transformations available to the learner, and the scoring function. When given a context b in test data, we begin with the default outcome for b, and apply, in respective order, the transformations learned during the training phase. Past literature[Brill, 1994] on transformation based learning claims that the rules learned by the procedure are easier to understand than the statistics of comparable probabilistic approaches. Transformation based learning is extremely flexible, and has been used (among other tasks) for part ofspeech tagging [Brill, 1994] ....

....1994] on transformation based learning claims that the rules learned by the procedure are easier to understand than the statistics of comparable probabilistic approaches. Transformation based learning is extremely flexible, and has been used (among other tasks) for part ofspeech tagging [Brill, 1994], prepositional phrase attachment [Brill and Resnik, 1994] and parsing [Brill, 1993b] Maximum entropy models are equally flexible in the kinds of evidence they allow and the types tasks they can perform. In the maximum entropy framework, we do not specify the space of possible transformations, ....

[Article contains additional citation context not shown here]

Brill, E. and Resnik, P. (1994). A Rule Based Approach to Prepositional Phrase Attachment Disambiguation. In Proceedings of the Fifteenth International Conference on Computational Linguistics (COLING).

....the preposition is not within the same trigram window as the noun and the verb due to modifiers of the noun and the verb. However, despite this limitation, the trigram approach performs as well as Ratnaparkhi, Reynar Roukos [1994] and Hindle and Rooth [1991] and is outperformed only by methods (Brill and Resnik [1994], Collins and Brooks[1995] that use four words: the verb, the object noun, the preposition and the complement of the preposition in making the attachment decision. Other Constructions In this section, we summarize the performance of the supertagger in identifying constructions such as ....

E. Brill and P. Resnik. A rule-based approach to prepositional phrase attachment disambiguation. In Proceedings of the International Conference on Computational Linguistics (COLING '94), Kyoto, Japan, 1994.

....isolation. 2 Previous Work on PP Attachment Systems Currently, the best PP attachment disambiguation techniques combining high accuracy with low computational cost involve recording phrase head patterns from corpora. Three teams using such an approach are Hindle and Rooth [HR93] Brill and Resnik [BR94], and Collins and Brooks [CB95] Hindle and Rooth used a corpus of AP news stories to record PP attachment probabilities given the heads of the VP, NP, and PP involved. The corpus they used was not parsed in advance so the correct attachments were unknown. However, their idea was that ....

E. Brill and P. Resnik. A rule-based approach to prepositional phrase attachment disambiguation. In 15th International Conference on Computational Linguistics (COLING94) , Kyoto, Japan, 1994.

.... tagging (Brill, 1995; Ramshaw and Marcus, 1994) spelling correction (Mangu and Brill, 1997) word sense disambiguation (Gale et al. 1992) message understanding (Day et al. 1997) discourse tagging (Samuel et al. 1998) accent restoration (Yarowsky, 1994) prepositional phrase attachment (Brill and Resnik, 1994) and base noun phrase identi cation (Ramshaw and Marcus, In Press; Cardie and Pierce, 1998; Veenstra, 1998; Argamon et al. 1998) Many of these rule based systems learn a short list of simple rules (typically on the order of 50 300) which are easily understood by humans. Since these rule based ....

E. Brill and P. Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation. In Proceedings of the fteenth International Conference on Computational Linguistics (COLING-1994).

....constraints. Progress on broad coverage prepositional phrase attachment ambiguity has been particularly notable, now that the dominant approach has shifted from structural strategies to quantitative analysis of lexical relationships (Whittemore, Ferrara, Brunner, 1990; Hindle Rooth, 1993; Brill Resnik, 1994; Ratnaparkhi Roukos, 1994; Li Abe, 1995; Collins Brooks, 1995; Merlo, Crocker, Berthouzoz, 1997) Noun compounds have received comparatively less attention (Kobayasi, Takunaga, Tanaka, 1994; Lauer, 1994, 1995) as has the problem of coordination ambiguity (Agarwal Boggess, 1992; ....

Brill, E., & Resnik, P. (1994). A rule-based approach to prepositional phrase attachment disambiguation.

No context found.

Eric Brill and Philip Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation.

No context found.

Brill, E., & Resnik, P. (1994). A rule-based approach to prepositional phrase attachment disambiguation. COLING 1994 . Available at: http://www.cs.jhu.edu/~brill/pp-attachment.ps

No context found.

E. Brill and P. Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation.

No context found.

Eric Brill and Philip Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation.

No context found.

Brill E, Resnik P. A rule-based approach to prepositional phrase attachment disambiguation. In: COLING; 1994.

No context found.

E. Brill and P. Resnik, "A Rule-Based Approach to Prepositional Phrase Attachment Disambiguation," Proc. COLING1994, Kyoto, Japan, 5--9 Aug. 1994.

No context found.

Eric Brill and Philip Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation. In Proceedings of the #fteenth International Conference on Computational Linguistics.

No context found.

E. Brill and P. Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation. In Proceedings of COLING.

No context found.

Eric Brill and Philip Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation. In Proceedings of the fifteenth International Conference on Computational Linguistics.

No context found.

Brill, E. and Resnik,P. \A rule-based approach to prepositional phrase attachment ", Proceedings of the 15th Computational Linguistics conference (COLING) 1994, pp 1198-1204.

No context found.

Eric Brill and Philip Resnik. 1994. A Rule-Based Approach to Prepositional Phrase Attachment Disambiguation. In Proceedings of COLING-94.

No context found.

E. Brill and P. Resnik. A rule-based approach to prepositional phrase attachment disambiguation. Proc. of COLING'94, pp. 1198--1204. P. F. Brown, V. J. Della Pietra, P. V. deSouza, J. C. Lai, and R. L. Mercer. 1992. Class-based ngram models of natural language. Comp. Ling., 18(4):283--298.

No context found.

Eric Brill and Philip Resnik. 1994. A rule-based approach to prepositional phrase attachment disambiguation.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC