D. Magerman. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. dissertation, Stanford University, February 1994. Section 3.4.2.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....successes, and the debate once thought closed by Chomsky s I saw a fragile whale is now as open as it ever has been. Most part of speech taggers now rely on statistics (Leech, Garside, Atwell 1983; Robert, Armstrong 1995) and it seems possible that parsers may also go this way (Church 1998; Magerman 1994, 1995; Briscoe, Carroll 1993) though more conventional methods are also increasing in quality and robustness (Strzalkowski, Scheyen 1993) It is possible that there is a natural ceiling to the advance of performance models (Wilks 1994) but the point of relevance for this report is that the ....

D. Magerman, Natural Language Parsing as Statistical Pattern Recognition, PhD thesis, Department of Computer Science, Stanford, 1994.

....and Schabes were run. Crossing brackets, zero crossing brackets, and the paired differences are presented in Table 1. All sentences output by the parser were made binary branching (see the section covering analysis of Bod s data) since otherwise the crossing brackets measures are meaningless (Magerman, 1994). A diff file between the original ATIS data and the cleaned up version, in a form usable by the ed program, is available by anonymous FTP from ftp: ftp.das.harvard.edu pub goodman atis ed tLtb.par ed and ti tb.pos ed. Note that the number of changes made was small. The diff files sum to 457 ....

David Magerman. 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. thesis, Stanford University University, February.

.... (e.g. a tagger or a chunker) is used as input by other memory based modules (e.g. syntactic relation assignment) Similar cascading ideas have been explored in other approaches to text analysis: e.g. finite state partial parsing [Abney, 1996, Grefenstette, 1996] statistical decision tree parsing IMagerman, 1994] maximum entropy parsing [Ratnaparkhi, 1997] and memory based learning [Cardie, 1994, Daelemans et al. 1996] Algorithms and Implementation For our experiments we have used TiMBL , an MBL software package developed in our group [Daelemans et al. 1999b] We used the fol lowing variants of ....

D. M. Magerman. Natural language parsing as statistical pattern recognition. Dissertation, Stanford University, 1994.

.... P (t) #P (t c) 1 #)P (t w) The model parameter # can be estimated from an annotated training text or via the forward backward algorithm [45] Clustering, decision trees, and decision lists are other general classification methods that have been applied to problems in computational linguistics [59, 89], including part of speech tagging [11] 6 A disadvantage of classification models is that they typically involve supervised training i.e. an annotated training corpus. On the other hand, as we have seen, HMM models often require as much manually prepared material as classification models do, ....

David Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford, 1994.

....al. 95] Similarly, there are many reports of work with large corpora of texts in general [Ejerhed Dagan 96] and the Penn Tree Bank [Marcus et al. 93] in particular. For example: extracting grammars [Gaizauskas 95] building taggers and parsers and measuring their performance [Bod 96, Magerman 94, Magerman 95, Brill 92b, Brill 95, Miller et al. 98] There remain significant barriers to reuse, to do with the format di#erences and distribution mechanisms (see section 5.3.1 and [Bird Liberman 98, Bird Liberman 99a, Peters et al. 98, Cunningham et al. 98a] but in general the picture is ....

D. Magerman. Natural Language Parsing as Statistical Pattern Recognition. Unpublished PhD thesis, Department of Computer Science, Stanford University, CA, 1994.

....a gradual ban on virtually all uses of asbestos , can only ask about the following four words: imposed ban on uses The notion of a head word here corresponds loosely to the notion of a lexical head. We use a small set of rules, called a Tree Head Table, to obtain the head word of a constituent [12]. We allow two types of binary valued questions: 1. Questions about the presence of any n gram ( of the four head words, e.g. a bigram maybe V = is , P = of . Features comprised solely of questions on words are denoted as word features. 2. Questions that involve the class ....

Magerman, D., 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph. D. dissertation, Stanford University, California.

....neighbor nodes. Magerman has also devised an e#cient algorithm for finding the parse tree (interpretation) with the highest likelihood value. The advantages of this method are its e#ective use of contextual information and its non use of a hand made grammar. See also (Magerman and Marcus, 1991; Magerman, 1994; Black et al. 1993; Ratnaparkhi, 1997; Haruno, Shirai, and Ooyama, 1998) Su and Chang (1988) propose the use of a probabilistic score function for disambiguation in generalized LR parsing (see also (Su et al. 1989; Chang, Luo, and Su, 1992; Chiang, Lin, and Su, 1995; Wright, 1990; Kita, 1992; ....

Magerman, David M. 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. Thesis, Stanford Univ.

....always marked explicitly, and we use a heuristic procedure which uses the label of a node and its parent to make this distinction. Though the Treebank does not speci cally indicate syntactic heads, a deterministic procedure for identifying them is given in [10] this was originally developed by [20]) and we use a slightly modi ed version of this heuristic. Additionally, there are di erent types of null elements encoding traces, multiple attachments and attachment ambiguities. The presence of these null elements allows us to infer the correct categories even in relative clauses, ....

David M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD Thesis, Stanford University, 1994.

....al. 95] Similarly, there are many reports of work with large corpora of texts in general [Ejerhed Dagan 96] and the Penn Tree Bank [Marcus et al. 93] in particular. For example: ffl extracting grammars [Gaizauskas 95] ffl building taggers and parsers and measuring their performance [Bod 96, Magerman 94, Magerman 95, Brill 92b, Brill 95, Miller et al. 98] There remain significant barriers to reuse, to do with the format differences and distribution mechanisms (see section 5.3.1 and [Bird Liberman 98, Bird Liberman 99a, Peters et al. 98, Cunningham et al. 98a] but in general the picture ....

D. Magerman. Natural Language Parsing as Statistical Pattern Recognition. Unpublished PhD thesis, Department of Computer Science, Stanford University, CA, 1994.

.... It can focus attention upon subspaces of possible model parameterisations whose maximum correlates with the best performance at a given task [Pereira and Schabes, 1992] Numerous studies have demonstrated the utility of parsed corpora as an learning constraint (for example, Black et al., 1993b,Magerman, 1994,Collins, 1996] However, in practice, there are problems with available parsed corpora: they are limited in quantity (thus do not cover every construct in any given natural language) In addition, they often only partially specify derivations (for example, Noun Phrases in the parsed Wall Street ....

D. M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford University, February 1994.

.... Eeg Olofsson 1991, Brill 1993, Atwell et al. 1984, 2000a) Sentence structure analysis or parsing (mapping word and or word class sequences onto parses) e.g. Sampson et al. 1989, Atwell 1987, 1988, 1993, Black et al. 1993, Bod 1993, Briscoe 1994, Jelinek et al. 1992, Joshi and Srinivas 1994, Magerman 1994, O Donoghue 1993, Schabes, Roth and Osborne 1993, Sekine and Grishman 1995) semantic analysis or word sense tagging (mapping word sequences onto semantic tags or meaning analyses) e.g. Demetriou 1993, Demetriou and Atwell 1994, 2001, Bod et al. 1996, Kuhn and de Mori 1994, Weischedel et al. ....

Magerman D 1994 Natural Language Parsing as statistical pattern recognition. PhD thesis, Stanford University.

....Rule probabilities are initialized to one of two values, and the entire space of possible rules is considered for parsing. Apparently this is still too much of a burden for the InsideOutside algorithm, as they report mediocre results on both parsing accuracy and perplexity measures. 18 Magerman [25] has developed a system that addresses what he calls the treebank recognition problem. This is similar to parsing, only oriented towards reproducing the partial parse style bracketings seen in treebanks. In several respects, Magerman s work is the polar opposite to our own. Where we concentrate on ....

....been to show that information from a traditionally symbolic field (linguistics) can be usefully applied to a traditionally statistical problem (language modeling) e.g. we have shown that linguistically motivated constraints can be used as a partial cure for overfitting. Others such as Magerman [25] have taken the converse step of bringing statistical methods (decision trees) to parsing. Both Magerman s and our results show improvements over the current state of the art. Together they constitute a fairly strong argument that symbolic and statistical approaches have complementary strengths, ....

David Magerman, "Natural Language Parsing as Statistical Pattern Recognition, " Stanford University, PhD thesis, 1994.

....natural language understanding is not only viable, but a reasonable alternative to the more common linguistic approach. Already statistically trained part of speech taggers have matched or surpassed hand tailored ones in terms of performance[49, 50] The same is true for statistical parsing[71, 48, 46] and machine translation[7] This thesis intends to convince the reader that eventually this will be true for natural language understanding as well. 1.4 Organization of Thesis This thesis is organized as follows: ffl Chapter 2 introduces statistical natural language understanding at a very ....

....briefly discusses the components of the source channel model that are not investigated in this thesis. 2.2 Related Work In the last ten years, statistical methods have been used for more and more problems in computational linguistics. These have ranged from simple taggers[18] to complex parsers[71, 48, 46]. In addition, there has also been published work in natural language understanding. Two research efforts, at AT T and BBN, have used the source channel model, as done in this thesis. There are also a few NLU projects using different modeling paradigms. In the following subsections, some of these ....

[Article contains additional citation context not shown here]

D.M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford University, February 1994.

....of log linear models from incomplete data. This algorithm is applicable to log linear probability distributions in general, and has been shown here to be useful to estimate the parameters of probabilistic context sensitive NLP models. In contrast to related approaches such as that of [1] or [15], our statistical inference algorithm provides the means for automatic and reusable training of probabilistic constraint based grammars from unparsed corpora. Furthermore, heuristic search algorithms for nding the most probable analysis in the CLP model can be based upon this probability model. ....

David M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Department of Computer Science, Stanford University, 1994.

....certain formal grammar formalism such as the (generalized) context free grammar formalism, the tree adjoining grammar formalism and so on. The grammar rules are written and coded by linguists over many years. Alternatively, they can be learned from an annotated corpus (Carroll Charniak, 1992; Magerman, 1994). In any case, the underlying assumption of using a particular grammar formalism is that most, if not all, of the syntactic 1 Address: Computational Science Programme, Faculty of Science, National University of Singapore, Lower Kent Ridge Road, Singapore 0511. Tel: 65 373 2016, Fax: 65 775 ....

Magerman, D. (1994). Natural Language Parsing As Statistical Pattern Recognition. Ph.D. Thesis, Stanford University.

....Natural Language Toolkit to Analyse a Software Manual Corpus Miles Osborne 1 University of York 8. 1 Introduction Within the last decade there has been considerable research devoted to the problem of parsing unrestricted natural language (e.g. Alshawi, 1992; Black, Garside Leech, 1993; Magerman, 1994). By unrestricted, we mean language that is in everyday use. Examples of unrestricted language can be found in such places as requirement documents, newspaper reports, or software manuals. If unrestricted language can be successfully parsed, then we will be a lot closer to achieving long terms ....

Magerman, D. M. (1994). Natural Language Parsing as Statistical Pattern Recognition. Ph.D. thesis, Stanford University, February 1994.

....procedure to train G 0 on B. The probabilities actually used were distributions on rule applications given the mother rule and its daughter number as context K. We would have wished to extend the context K to larger portions of a tree in the sense of a history based grammar model (see Magerman 1994), but would have been running in sparse data problems then. Since the original tree bank was produced by a HPSG style uni cation grammar (where no structure sharing can be used in a tree bank) we could not use an inside outside algorithm to estimate our distributions. Instead, we extended a PCFG ....

Magerman, D. M. 1994. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford University, February.

.... = P (tjc) 1 Gamma )P (tjw) The model parameter can be estimated from an annotated training text or via the forward backward algorithm [45] Clustering, decision trees, and decision lists are other general classification methods that have been applied to problems in computational linguistics [59, 89], including part of speech tagging [11] A disadvantage of classification models is that they typically involve supervised training i.e. an annotated training corpus. On the other hand, as we have seen, HMM models often require as much manually prepared material as classification models do, if ....

David Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford, 1994.

....we use for estimating the probabilities includes both word identities and POS tags. To make effective use of this information, we need to allow the decision tree algorithm to generalize between words and POS tags that behave similarly. To learn which words behave similarly, Black et al. 1989) and Magerman (1994) used the clustering algorithm of Brown et al. 1992) to build a hierarchical classification tree. Figure 1 gives the classification tree that we built for the POS tags. The algorithm starts with each token in a separate class and iteratively finds two classes to merge that results in the smallest ....

....RBR RB VBG VBN RP DP PRP MD TO CC PREP JJ JJS JJR CD DT PRP WDT NN NNS NNP Figure 1: Classification Tree for POS Tags 3.2. 3 Questions about Word Identities For handling word identities, one could follow the approach used for handling the POS tags (e.g. Black et al. 1992; Magerman, 1994)) and view the POS tags and word identities as two separate sources of information. Instead, we view the word identities as a further refinement of the POS tags. We start the clustering algorithm with a separate class for each word and each POS tag that it takes on and only allow it to merge ....

D. Magerman. 1994. Natural language parsing as statistical pattern recognition. Doctoral dissertation, Dept. of Computer Science, Stanford.

....complete. Completeness here means that the learner can always make a decision. This is achieved by the learner making the ergodic assumption, which, roughly stated, means that any event is possible. For example, in the context of a learner using a treebank and a tree matching algorithm (such as [38]) this means always returning a non zero score for any local tree being matched. ffl There is little logistical effort required, other than the creation of (say) a treebank. This means that knowledge engineering is kept to a minimum. Linguists are not required to formulate grammatical models ....

....their features, thus forming more specific rules. Note also that these rule refinement operators may at times prevent idiosyncratic rules, necessary for generating unusual constructs found in a corpus, from being learnt. 3. 4 Data driven learning In other data driven systems (for example [20, 34, 38]) the data driven component uses a treebank consisting of shallow parse trees, generated by highly specific rules. Hence, these systems in turn tend to acquire rules that contain many categories in their right hand side and are also highly specific. As we have already explained, in our system new ....

David M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford University, February 1994.

....data, demonstrate that even a simple model of punctuation increases the plausibility of learnt grammars over grammars learnt without the use of punctuation. Introduction Natural Language Processing (NLP) systems have for many years suffered from what Magerman terms the toy problem syndrome (Magerman 1994). Systems are built, often with a small lexicon or modest grammar, that are usually demonstrational rather than being operational. That is, there has been little work in developing systems capable of dealing with unrestricted, naturally occurring language. Since the mid80s, this has begun to ....

Magerman, D. M. 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. Dissertation, Stanford University.

....unrestricted, naturally occuring language is the development of an appropriate lexicon, and not the coverage of the grammar. 1 Introduction Within the last decade there has been considerable research devoted to the problem of parsing unrestricted natural language (see for example [Als92, BGL93, Mag94] By unrestricted, we mean language, that is in everyday use. Examples of unrestricted language can be found in such places as requirement documents, newspaper reports, or software manuals. If unrestricted language can be successfully parsed, then we will be a lot closer to achieving long terms ....

David M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford University, February 1994.

....such as speech recognition. In addition, it would be interesting to see whether these results extend to elds other than language modeling where smoothing is used, such as prepositional phrase attachment (Collins and Brooks, 1995) part of speech tagging (Church, 1988) and stochastic parsing (Magerman, 1994). Acknowledgements The authors would like to thank Stuart Shieber and the anonymous reviewers for their comments on previous versions of this paper. We would also like to thank William Gale and Geo rey Sampson for supplying us with code for Good Turing frequency estimation without tears. This ....

Magerman, David M. 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. thesis, Stanford University, February.

....that are also in the key total number of answer brackets Each of these measures could arguably be used to assess the quality of a parse in the learning process. We have chosen to focus on the traditional crossed key measure that was used by Brill, as well as on the recall measure argued for in (Magerman 1994 ) The precision measure is not as directly relevant as the others, since the binary branching structures required by this approach always produce a misleadingly high precision error. Turning to implementation, an efficiently computed measure of recall is easy to obtain incrementally during the ....

....configuration of the bracketing parser. Indeed, recall levels of 74 and crossed key levels of 80 seem to be as good as this configuration gets. These were very exciting performance levels at the time Brill first published his work. However, recent results by Magerman and his colleagues (Magerman 1994) and by (Collins 1996 ) cite recall levels in the mid 80 range. It is thus fair to ask whether this approach might not have reached a glass ceiling. In our opinion, the problem is not with the overall approach, as the simple notion of transformation sequences has proved so workable in many ....

Magerman, d. (1994) Natural Language Parsing as Statistical Pattern Recognition. Ph.D. Dissertation, Stanford University, Department of Computer Science.

....we use for estimating the probabilities includes both word identities and POS tags. To make effective use of this information, we need to allow the decision tree algorithm to generalize between words and POS tags that behave similarly. To learn which words behave similarly, Black et al. 1989) and Magerman (1994) used the clustering algorithm of Brown et al. 1992) to build a hierarchical classification tree. Figure 1 gives the classification tree that we built for the POS tags from the Trains corpus. The algorithm starts with each token in a separate class and iteratively finds two classes to merge that ....

....RP DP PRP MD TO CC PREP JJ JJS JJR CD DT PRP WDT NN NNS NNP Figure 1: Classification Tree for POS Tags tions (Heeman, 1997) 3. 4 Questions about Word Identities For handling word identities, one could follow the approach used for handling the POS tags (e.g. Black et al. 1992; Magerman, 1994)) and view the POS tags and word identities as two separate sources of information. Instead, we view the word identities as a further refinement of the POS tags. We start the clustering algorithm with a separate class for each word and each POS tag that it takes on and only allow it to merge ....

D. Magerman. 1994. Natural language parsing as statistical pattern recognition. Doctoral dissertation, Stanford University.

....having to design accurate model classes. It can focus attention upon subspaces of possible model parameterisations whose maximum correlates with the best performance at a given task [26] Numerous studies have demonstrated the utility of parsed corpora as an estimation constraint (for example, [4, 22, 12]) However, in practice, there are problems with available parsed corpora: they are limited in quantity (thus do not cover every construct in any given natural language) In addition, they often only partially specify derivations (for example, Noun Phrases in the parsed Wall Street Journal (WSJ) ....

David M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford University, February 1994.

.... a tagger or a chunker) is used as input by other memory based modules (e.g. syntactic relation assignment) Similar cascading ideas have been explored in other approaches to text analysis: e.g. finite state partial parsing [ Abney, 1996; Grefenstette, 1996 ] statistical decision tree parsing [ Magerman, 1994 ] maximum entropy parsing [ Ratnaparkhi, 1997 ] and memory based learning [ Cardie, 1994; Daelemans et al. 1996 ] Algorithms and Implementation For our experiments we have used TiMBL 1 , an MBL software package developed in our group [ Daelemans et al. 1999b ] We used the following ....

D. M. Magerman. Natural language parsing as statistical pattern recognition. Dissertation, Stanford University, 1994.

....seen is allowed to influence the expansion of a non terminal node. Decision trees were used in this work to estimate the rule probabilities. The authors tested a HBG and found it to be more accurate than a PCFG it was compared to. Another decision tree based model was used by Magerman in [33] with good results. While some of these systems seem quite adept at finding good parses of input sentences, they can not be directly used as language models as they don t calculate the prior probabilities we are interested in. Unfortunately, we know of no examples where language models based on ....

D. M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Department of Computer Science, Stanford University, Stanford, CA, Feb 1994.

....and target clauses: The main verb of the source clause occurs in the passive voice and the main verb of the target clause in the active voice. 33) since regardless of which bit is initially assigned, it will be flipped if more information is gained by doing so. flipping it] from text of Magerman (1994, page 29) 11 Some of the VP ellipsis data (cf. 32a) have been called into question. Schachter (1977) provides a number of felicitous examples of VP ellipsis with situationally evoked referents, such as (i) and (ii) i) John tries to kiss Mary. She says: John, you mustn t. ii) John ....

Magerman, David. 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph.D.

....The correct rate (44 correct parsing, 38 correct parsing and thematic role assignment) that is obtained in this research is lower than other published approaches in computational linguistics. For instance, Fujisaki et al. 5] achieved 85 correct analysis among 84 test sentences, and Magerman [11] obtained 78 correct among 1473 test sentences. Many improvements are necessary before any claim that the current research program has proven something about language acquisition can be made. However, the learning conditions of the model that is presented here are different from those assumed in ....

Magerman, D. M. (1994). Natural language parsing as statistical pattern recognition. Ph.D. thesis, Stanford University.

....learning taggers is that they can be trained using only a dictionary listing the allowable parts of speech for each word and not a manually tagged corpus. The recent work also includes a hybrid approach that combines automatically learned context constraints using statistical decision trees [67] and manually written linguistic constraints for the POS tagging [68] a constraint based morphological tagger that uses a two level morphological analysis with a large lexicon and a morphological description [94] The advantages of incorporating a part of speech tagger in an information ....

D. Magerman. Natural language parsing as statistical pattern recognition. Ph.D Thesis.

.... others chose to make no direct challenge to linguistic orthodoxy while using methods which undermine some key tenets of theoretical linguistics (Carroll and Charniak [24] for example consider phrases good if they occur frequently, regardless of what linguists think; also, Magerman [87] declares in the preface of his thesis dissertation that a significant goal of his work was to replace linguistics with statistical analysis of corpora) Investigating the amount of linguistic structure in language utterances is an interesting theoretical research topic in itself, though it also ....

David M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford University Computer Science Department, February 1994.

....for best analyses which is not addressed by Abney (1996) The expressive power of log linear models even allows us to couch other approaches to probabilisitic processing beyond context freeness in terms of this framework. Statistical decision trees as used in the probabilistic parsing model of Magerman (1994) can be cast in the log linear framework by encoding the questions building up a decision tree as binary valued, disjoint property functions. Property selection then can be seen as closely related to growing a decision tree and iterative maximization can be seen as maximum likelihood estimation ....

....a decision tree as binary valued, disjoint property functions. Property selection then can be seen as closely related to growing a decision tree and iterative maximization can be seen as maximum likelihood estimation for such defined decision trees. However, in contrast to the algorithms used by Magerman (1994), which require large samples of complete data, our approach allows induction of the probabilistic model from incomplete data. A similar statement can be made for the probabilistic tree substitution model of Bod (1995) This approach can be couched as a log linear model employing all subtrees of a ....

Magerman, D. M. (1994). Natural Language Parsing as Statistical Pattern Recognition. Ph. D. thesis, Department of Computer Science, Stanford University.

....source and target clauses: the main verb of the source clause occurs in the passive voice and the main verb of the target clause in the active voice. 271) since regardless of which bit is initially assigned, it will be flipped if more information is gained by doing so. flipping it] text of Magerman (1994, page 29) 272) Section 1 provides the examples to be derived by Gapping, and a formulation of Gapping capable of doing so. deriving the examples] text of Neijt (1981) 273) As an imperial statute the British North America Act could be amended only by the British Parliament, which did so on ....

Magerman, David. 1994. Natural Language Parsing as Statistical Pattern Recognition.

No context found.

D. Magerman. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. dissertation, Stanford University, February 1994. Section 3.4.2.

No context found.

D. Magerman. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. dissertation, Stanford University, February 1994. Section 3.4.2.

No context found.

Magerman, David M. 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. thesis, Stanford University, February.

No context found.

Magerman, D. 1994. Natural Language Parsing as Statistical Pattern Recognition PhD Thesis, Stanford University, CA.

No context found.

Magerman, D. 1994. Natural Language Parsing as Statistical Pattern Recognition PhD Thesis, Stanford University, CA.

No context found.

Magerman, D. M. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford University, 1994.

No context found.

David Magerman. 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. thesis, Stanford University, CA.

No context found.

Magerman, D. 1994. Natural Language Parsing as Statistical Pattern Recognition PhD Thesis, Stanford University, CA.

No context found.

Magerman, D. (1994): Natural Language Parsing as Statistical Pattern Recognition, PhD Thesis, Stanford University, CA.

No context found.

D. Magerman. 1994. Natural Language Parsing as Statistical Pattern Recognition Ph.D. Thesis, Stanford University, CA

No context found.

Magerman, D. (1994): Natural Language Parsing as Statistical Pattern Recognition, PhD Thesis, Stanford University, CA.

No context found.

David Magerman. 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. thesis, Stanford University.

No context found.

David M. Magerman. Natural language parsing as statistical pattern recognition. Thesis CS-TR-94-1502, Stanford University, Department of Computer Science, February 1994.

No context found.

David M. Magerman. Natural Language Parsing as Statistical Pattern Recognition. PhD thesis, Stanford University, February 1994.

No context found.

David M. Magerman, Natural Language Parsing as Statistical Pattern Recognition, PhD thesis, Stanford University, Palo Alto, California, February 1994.

No context found.

David M. Magerman. 1994. Natural Language Parsing as Statistical Pattern Recognition. Ph.D. thesis, Department of Computer Science, Stanford University.

First 50 documents  Next 50

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