Gerald Gazdar and Chris Mellish, Natural Language Processing in PROLOG, Addison Wesley, Workingham u.a., 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....by military and intelligence communities. These systems, what we call first generation, translate text almost word by word; the result was a failure. But considering the lack of theories, methods, and resources with semantics and ambiguities in natural language text, the result is not surprising [4]. Today with the advance of theories, resources, etc. MT is not a dream; even there are MT systems available in the market. Many components of NLP systems, like syntactic analyzers, text generators, taggers, and semantic disambiguators, need knowledge about words in the language. This ....

....access. fsdb(verb, existential, none, none, none, var, cat: maj:verb, syn: Figure 5.1: The entry for the existential verb var in the feature structure database. Feature structures are represented as a list of feature name:feature value pairs (see Gazdar and Mellish [4]) For example, the following feature structure with abstract representation would be represented in Prolog as in Figure 5.2: The morphological processor that our lexicon employs is implemented by Oflazer (see Oflazer [11] for the two level description of Turkish morphology) using a finite state ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog, chapter 7, pages 217--278. Addison-Wesley Publishing Company, 1989.

.... works by matching one empathy hierarchy against another, which makes it possible to deal with discourses with no explicit topic and those with cataphora [Halliday and Hasan, 1990] The theory is formalized through the definite clause grammar(DCG) formalism [Peteira and Warren, 1980] [Gazdar and Mellish, 1989; Longacre, 1979J. Finally, we show that graphology i.e. quotation mark, spacing, has an important effect on the interpretation of zero anaphora in Japanese discourse. 1 Introduction Over the past years, schemes like Focusing and Centering have dominated computational approaches to ....

Gerald Gazdar and Chris Mellish. Natural Language Processing in Prolog. Addison-Wesley Publishing Co., New York, 1989.

....We also use a rule based lemmatiser that produces an a#x and root for each noun and verb in the input text. The lemmatiser program is implemented as a set of regular expressions specified in flex and translated into C code. We are using an implementation of the Bottom up chart parsing described in [3], enriched with semantic rules that construct a naive semantic of each sentence in first order logical form. The parser is complete in the sense that every analysis licensed by the grammar is produced, though there is a mechanism to control this. On completion a best parse algorithm is run to ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison-Wesley, Reading, MA, 1989.

....are usually taken from speci c real world problems and representations are often constructed ad hoc. Morik [15] di erentiates between two di erent aspects of time, the linear precedence of events and immediate dominance of temporal categories. These terms originate from natural language theory [9]. Immediate dominance refers to the construction of higher level categories of the time dependent elements, exemplary learning tasks based on the concept of immediate dominance are the discovery of frequent episodes [14] or rst order logic learning based on Allen s interval relations [1] The ....

Gerald Gazdar and Chris Mellish. Natural Language Processing in PROLOG. Addison Wesley, Workingham u.a., 1989.

....serve as input to the GATE design and evaluation processes, and are presented in some detail for that reason. The section does not attempt a comprehensive review of LE technologies; wider coverage may be found in textbooks such as [Allen 95, Hutchins Somers 92, Reiter Dale 00, O Shaugnessy 87, Gazdar Mellish 89] Note also that because technologies intersect, certain subjects appear incidentally in the following that could occupy sections in their own right (so, for example, grammar is discussed in the section on NLG; this is not meant to suggest that NLG is the sole or main user of grammars in LE) ....

....describes the language syntax, the ways in which words may be combined into sentences. Lexical units may be further decomposed into morphemes, indivisible grammatical or semantic units that may stand alone as words or be grouped to form other words, e.g. print s, print ed, print ing, re print [Gazdar Mellish 89] Similarly, the unit of word construction in speech is the phoneme, defined such that if one phoneme is substituted for another in a word, the meaning of the word may be changed [Ainsworth 88] Only certain phoneme sequences are permissible in any given language. It should be noted here that ....

[Article contains additional citation context not shown here]

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison-Wesley, Reading, MA, 1989.

....serve as input to the GATE design and evaluation processes, and are presented in some detail for that reason. The section does not attempt a comprehensive review of LE technologies; wider coverage may be found in textbooks such as [Allen 95, Hutchins Somers 92, Reiter Dale 00, O Shaugnessy 87, Gazdar Mellish 89] Note also that because technologies intersect, certain subjects appear incidentally in the following that could occupy sections in their own right (so, for example, grammar is discussed in the section on NLG; this is not meant to suggest that NLG is the sole or main user of grammars in LE) ....

....describes the language syntax, the ways in which words may be combined into sentences. Lexical units may be further decomposed into morphemes, indivisible grammatical or semantic units that may stand alone as words or be grouped to form other words, e.g. print s, print ed, print ing, re print [Gazdar Mellish 89] Similarly, the unit of word construction in speech is the phoneme, defined such that if one phoneme is substituted for another in a word, the meaning of the word may be changed [Ainsworth 88] Only certain phoneme sequences are permissible in any given language. It should be noted here that ....

[Article contains additional citation context not shown here]

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison-Wesley, Reading, MA, 1989.

....abstract lexeme with other non syntactic features which can be described in DATR. 4 The DUTR project DUTR is an approach to join DATR with PATR into a single formalism. The connection between DATR and feature structures was realised on the bases of HUB DATR and a slightly extended Prolog PATR [14]. Both formalisms are Default and Unification Tree Representation realized in pure Prolog and run as one Prolog process. Prolog variables are used to interchange data between HUB DATR and PATR. Within a PATR feature structure free variables are bound by a DATR query predicate that derives the ....

....the derivation of Value from NodePathPair. The trace distinguishes between local and global inheritance. It exactly shows which equation is used to replace a descriptor. Unresolveable descriptors are marked with an error at the point they occur. 4. 3 PATR Our implementation of PATR follows mainly [14]. In order to use PATR, the file unify.pl has to be loaded. 4.3.1 The interface to HUB DATR The interface to HUB DATR is realized by the help of the HUB DATR query predicate datr 2. The builtin operator = uses this predicate to define a relation over feature values (see 3.4) as follows: ....

Gazdar, G. and Mellish, C. S. (1989): Natural Language Processing in PROLOG. Addison--Wesley, Wokingham.

....corpus into a multilingual aligned corpus using lexicons, grammars, etc. 4 Further support for the PR LR terminology may be gleaned from the argument in favour of declarative data structures for grammars, knowledge bases, etc. This argument was current in the late 1980s and early 1990s [Gazdar Mellish 89] partly as a response to what has been seen as the overly procedural nature of previous techniques such as augmented transition networks. Declarative structures represent a separation between data about language and the algorithms that use the data to perform language processing tasks; a similar ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison-Wesley, Reading, MA, 1989. References 92

....an ending to form a inflected word. Standard scientific considerations such as simplicity and generality apply to grammars in much the same way as they do to any other theories about natural phenomena. Other things being equal, a grammar with seven rules is to be preferred to one with 93 rules [Gazdar and Mellish, 1989]. The current dictionary has a considerable size: 43,000 lemma entries 5 , including 24,400 nouns, 7,600 verbs, and 11,000 adjectives. The model could be used for derivative morphology and compounds as well, but this has not been done yet, since further linguistic analysis must be done to ....

Gazdar, F. and Mellish, C.S. (1989). Natural Language Processing in Prolog. Addison Wesley.

....this problem is not strictly an ambiguity . There is no single general accepted definition of ambiguity, but it seems reasonable to describe a sentence or phrase as ambiguous if there is more than one clearly defined meaning for it. Source ambiguities occur at the semantic processing level (cf. [Gazdar and Mellish, 1989, Section 10.1] and as shown by the examples given above, they have effects on various aspects of that level. Since individual words with source ambiguities have effects on referent identification, and it is not possible to eliminate all such ambiguities at a lexical level, there is a need for a ....

Gazdar, G. and Mellish, C. (1989). Natural Language Processing in PROLOG. Addison-Wesley, Wokingham, UK.

....nding the antecedent for a pronoun is there any way to avoid considering all potential antecedents ) But these results relate to worst case, not normal or typical case behaviour. 9 10 Reading Partee et al. 1990, Part E, Ch18 ) is an introduction to formal language theory for linguists. Gazdar and Mellish (1989, 132 ) is a brief discussion of some issues. Gazdar and Mellish (1989, Chs1 3) discuss FSAs. The argument about NLs not being Finite State can be found in Chomsky (1957) convincing arguments for the non context freeness of NLs were given in Shieber (1985) and there is an excellent discussion in ....

....all potential antecedents ) But these results relate to worst case, not normal or typical case behaviour. 9 10 Reading Partee et al. 1990, Part E, Ch18 ) is an introduction to formal language theory for linguists. Gazdar and Mellish (1989, 132 ) is a brief discussion of some issues. Gazdar and Mellish (1989, Chs1 3) discuss FSAs. The argument about NLs not being Finite State can be found in Chomsky (1957) convincing arguments for the non context freeness of NLs were given in Shieber (1985) and there is an excellent discussion in Pullum (19 ) Barton et al. 1987) and Ristad (1993) are discussions ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison Wesley, Wokingham, 1989.

....we will turn our attention to issues of generation. Some understanding of Prolog (such as may be obtained from following LG519, or from a basic introduction such as Matthews (1998) is essential. Material for the course is drawn from sources such as Covington (1994) especially Chapter 6, Gazdar and Mellish (1989), especially chapters 5 and 6, and Pereira and Shieber (1987) General background can be found in Allen (1987) and Smith (1991) 2 Aims and Objectives to give a theoretically grounded introduction to contemporary work in Computational Linguistics. to introduce standard methods for parsing ....

....Basic Notions of Parsing Strategies: Top down and Bottom up Parsing (Covington, 1994) 2. A crash course in Phrase Structure Grammar and Linguistics (Borsley, 1996) Covington, 1994) 3. A crash course in the formal background (Aho et al. 1986) Human Sentence Processing; Chart Parsing (Gazdar and Mellish, 1989). 4. Shift Reduce Parsing: tabular techniques (Aho et al. 1986) 5. Left Corner Parsing Matsumoto, BUP (Matsumoto et al. 1985) Matsumoto et al. 1983) 6. Stack Based Bottom Up Parsing Tomita (Tomita, 1986) Tomita, 1987) 7. Parsing as Deduction(Pereira and Warren, 1983) 8. ....

[Article contains additional citation context not shown here]

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison Wesley, Wokingham, 1989.

.... DET on top of its stack, and saw as the rst word of input, such a table can be used to tell the parser to reduce this to N, disregarding the alternative of reduction to V (cf. LR parsing) 10 6 Reading Useful Discussions of top down and bottom up parsing can be found in Covington (1994, Ch6) Gazdar and Mellish (1989, Ch5) and Allen (1987, Ch3) Technical discussion unrelated to NLP can be found in most books on compiler design, e.g. Aho et al. 1986, Ch4) There is an interesting and very intuitive discussion which looks at psychological plausibility in Johnson Laird (1983) Since Tomita s work on ....

....of practically oriented work on this approach. See, for example, Tomita (1991) Bunt and Tomita (1996) There has been some work on the psychological plausibility of shift reduce parsing: e.g. Pereira (1985) Shieber (1983a) Shieber (1983b) There is discussion of Augmented Transition Networks in Gazdar and Mellish (1989), original papers include Woods (1970, 1986) and Kaplan (1972) ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison Wesley, Wokingham, 1989.

....search respectively) using the current strategy this would not matter. 2.3 Use of an oracle The left corner parsing can be improved by using an oracle, which prevents the addition to the chart of edges which cannot lead to a solution. As the details of the oracle are not provided by [Gazdar Mellish 89] or [Winograd 83] I provide an outline of its operation. 2.3.1 What the oracle stores The oracle is used with the left corner parser. Because active edges are included in the chart, there is always an idea of what the parser is looking for (as with top down parsing) When a category has been ....

....or bottom up (or both) 1 as Ritchie notes, and as described in chapter 2, the strategy I have been using should really be called left corner. I will use bottom up here and for the remainder of this dissertation for consistency with his paper. It is also common in other texts, such as [Gazdar Mellish 89, page 197] to see this strategy called bottom up 21 Formally, the definition of the grammar should be rewritten, so that, rather than having a single set of rules we require: 1. a set of Nonterminal symbols, N 2. a set of Terminal symbols, Sigma 3. a distinguished symbol, S, and 4. a ....

[Article contains additional citation context not shown here]

G.M. Gazdar and C.M. Mellish. Natural Language Processing in Prolog. Addison-Wesley, 1989.

....available for UML 1.3 Statecharts, superceding the gentler introduction in the UML 1.3 Users Guide [6] 1. 1 Semantics of State Machines Classical finite state machines are amenable to formal reasoning in terms of their equivalence to orders of grammar and formal language in the Chomsky hierarchy [7]. For example, a recursive language is defined by a context free grammar and is recognisable by a pushdown automaton, a variant of a finite state automaton with a global stack. However, different Statechart formalisms are subject to a number of different semantic interpretations. These result from ....

....to a consistent formal interpretation. 3.1 Classical Machines and Flowcharts In a classical finite state automaton, the states are quiescent vertices in the graph and all computational activity happens on the transition arcs, as events are processed. Mealy machines may be styled as transducers [7] which read an input symbol as each arc is traversed and generate an output symbol at the same time (figure 4a) Initial Medial Final in1 DoThis DoThat (a) b) out1 in2 out2 Initial Medial Final in1 out1 in2 out2 in1 out1 in2 out2 (c) done] done] Figure 4: Comparison of ....

Gazdar G and Mellish C. Natural Language Processing in Prolog. Addison Wesley, Reading MA, 1991

....(as opposed to the noun ground) will not be able to distinguish these cases, and will always get one of them wrong (the default is to treat the verbal form ground as grind ed) Chapter 8 buChart Parser 8. 1 Overview The parser is a modi cation of the Gazdar and Mellish bottom up chart parser [12]. It uses a feature based uni cation grammar to perform bottom up phrase structure syntactic analysis of each sentence in its input and during parsing a semantic representation of each constituent is constructed compositionally from the semantic representations of the constituent s components. The ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison-Wesley, Wokingham, 1989.

....are different tokens of the same type. For a dependency rule like (3b) however, the two occurrences of X are one and the same token. 238 Tom B.Y. LAI,HUANG Changning We emulate these rules with re write rules delimited by unificationbased constraints. Adapting the PATR formalism (SHIEBER 1986; GAZDAR MELLISH 1989) for our use, syntactic structures like the following are produced: 4) tv, n, john] saw, n, mary] In these dependency structures, there are no intermediate phrasal nodes. And it is in this sense that our formalism is dependency based rather than phrase structure based. A ....

GAZDAR, G. and MELLISH, C. (1989): Natural Language Processing in Prolog. Wokingham, UK: Addision Wesley.

.... In some situations, like in the processing of Finnish, where the words don t have a specific order, they are a good solution to an otherwise difficult problem (see [LJN85] One can find more complete descriptions of the use of regular expressions in natural language processing in [MN97] and [GM89]. We present our techniques in the following order. Section 2 presents definitions and conventions that we use. Among which is the description of the parse trees themselves. 1 Strictly speaking, the regular expressions are not context free grammars. But we can give to a regular expression a ....

Gazdar, G., Mellish, C.: Natural Language Processing in PROLOG. Addison-Wesley. 1989

....scheduling [68] Nevertheless, AI and symbolic computation remain the application areas that have benefited the most from the logic programming paradigm. Prolog has been used with extreme success in developing applications in a wide variety of AI related areas, such as natural language processing [26], planning [65] expert systems [5] etc. Not only can AI and symbolic computation problems be efficiently coded in Prolog, our aim in this paper is to show that such problems can be effectively parallelized and their execution speeded up on parallel Prolog systems. 3 1.3 Parallelism in Logic ....

....rules of an expert systems where multiple rules can be fired simultaneously to achieve a goal. It also arises in applications that involve natural language sentence parsing: the various grammar rules can be applied in or parallel to arrive at a parse for a sentence (e.g. non deterministic parsing [26]) Or parallelism frequently arises in database applications where there are large numbers of clauses. Or parallelism also arises in generate and test kind of problems the various alternatives to be tested can be generated in or parallel. 3.1 Implementation of Or parallelism In principle, ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison-Wesley, 1989.

....is done by searching a series of prestored lists of organisations, locations, date forms, currency names, etc. most of which have been derived from the gazetteer lists provided as part of the MUC 6 training data. buChart Parser: a modification of the Gazdar and Mellish bottom up chart parser [17], the module builds a semantic representation compositionally, and a best parse algorithm is applied to each final chart, providing a partial parse if no complete sentence span can be constructed. The parser operates in two stages, first applying a Named Entity grammar to construct proper noun ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. AddisonWesley, Wokingham, 1989.

....subsections. This section concludes wih a list of LE 0 required by the learning methods. 3.1 Structuring time phenomena For the overall view, we may structure time phenomena by two aspects, linear precedence and immediate dominance. These terms have been defined in natural language theory [15]. Linear precedence refers to the ordering of elements in a sequence. It is the relation between elements occuring along the time axis, horizontally depicted in Figure 1. Most statistical approaches are restricted to this aspect of time. Immediate dominance refers to categories of the ....

Gerald Gazdar and Chris Mellish. Natural Language Processing in PROLOG. Addison Wesley, Workingham u.a., 1989.

....1997) that he cites and its later developments, work by constraining senses and are perfectly able to report results with more than one sense still attaching to a word, just as some POS taggers result in more than one tag per word in the output. Close scholars of AI will also remember that Mellish [28], Hirst [33] and Small [52] all proposed methods by which polysemy might be computationally reduced by degree and not in an all or nothing manner. Or, as one might put it, under specification, Buitelaar s key technical term, can seem no more than an implementation detail in any effective tagger ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison-Wesley, 1989.

....the the Amsterdam Hypermedia Model. Many concepts in this paper are not totally new. Wherever possible, we consistently adopt terminology and concepts from other studies on information ( 15] multimedia ( 2, 23, 5] computer graphics ( 6] user models ( 10, 16] and knowledge based systems ([11, 32, 30]) In conclusion, the model is intended to fit real systems, like the cited ones. This is the main aim of this work, and we constantly have been guided towards such a goal. However, this does not necessarily mean that the proposed architecture is to be followed in an implementation. It is only a ....

G. Gazdar and C. Mellish. Natural Language Processing in PROLOG. Addison-Wesley Publishing Company, 1989.

....it to a minimum. In our approach, we do not hand code any semantic knowledge for individual words as was done in LDOCE. We start with the following elements: 1. 1800 words, with their part of speech and their textual definitions; 2. Morphological rules for the tagging of words; 3. A chart parser [66] with multiple parse rules for generating the parse tree(s) as well as additional information to help the parser: a) some word specific heuristics, for example, parse rules unlikely to make sense for some words (ex. a trip can go well, using rule [np n n] to make a compound noun of trip can ....

G. Gazdar and C. Mellish. Natural Language Processing in PROLOG, chapter 6. Addison-Wesley Publishing Company, 1989.

....Linton Miller is a Masters student with the Department of Computer Science at Victoria University of Wellington. 1 Introduction Parsers for general context free grammars have found wide application in natural language processing, because grammars for natural languages are close to context free [4]. To be useful in natural language processing, however, a parsing algorithm must cope well with grammatical ambiguity, as natural language grammars are highly ambiguous. The number of valid syntactic parses of a natural language sentence can be larger than exponential in the length of the sentence ....

Gerald Gazdar and Chris Mellish. Natural Language Processing in Prolog. Addison-Wesley, 1989.

....What it means is we end up with twice as many rules for S : the ones with the fullstop in the end and the ones without. This can also be true for other nonterminals. 3. 2 Parsing using the resulting rules To test the plausibility of the grammar we inferred, a simple Prolog chart parser from [6] was created and adjusted for the probabilistic model. This parser parses short sentences including the one displayed in Figures 1 and produces the same parse as in the Penn Treebank. However, even with a reduced version of this grammar (see Section 4) it took about an hour to parse a sentence as ....

G. Gazdar and C. Mellish, Natural Language Processing in Prolog, Addison Wesley, 1989

....An inactive edge is entered and rules are found for which this edge represents the initial constituent. 2. Combining with active edges. An inactive edges is entered and active edges are looked for with which it may combine. 3. Extension of active edges (usually termed the fundamental rule: Gazdar Mellish 1989:193) An active edge is entered and inactive edges are looked for to complete or extend the span of the edge. A number of primitive operations are required to support these general operations. vi MATTHEW F. HURST ffl matching: matching must be carried out between the constituents of rules. ffl ....

....a similar form to the standard production described by a context free rule. Instead of a simple series of daughters, the right hand side consists of an FSA (Figure 1) The use of the language of regular expressions to describe finite state automata is well documented (Aho, Sethi Ullman 1986:83; Gazdar Mellish 1989:134) as are algorithms for constructing the machines from these descriptions. The processes required to form well formed substring tables must be modified to accommodate the more complex rule description system of the FSA. In fact this alteration is not at all complex and is really a transfer of ....

Gazdar, Gerald & Chris Mellish 1989 Natural Language Processing in PROLOG.

....to new domains often requires a partial or even total reengineering of the system. The problem of creating programs that emulate human performance in natural language understanding is an instance of the knowledge acquisition bottleneck. Many have attempted to overcome this problem (Allen, 1995; Gazdar Mellish, 1989; Cardie, 1993; Riloff, 1993; Zelle Mooney, 1993) but have only contributed small pieces of the puzzle. This work employs machine learning techniques to contribute more and larger pieces to the puzzle, and advances a technique that integrates multiple levels of the NLP problem. Natural language ....

Gazdar, G., & Mellish, C. (1989). Natural Language Processing in Prolog. Adison-Wesley Publishing Company, New York.

....to be internally coherent) The internal structure of fs:s is not intended to be visible in the grammar, The predicates = 2 and denotes 2 allow us to assign values to features and to hide their internal structure. This kind of Prolog reconstruction of patr ii style notation is described by Gazdar and Mellish 1989. The arguments to the two relations are compiled into the internal representation of the ftfsg parser. The following kinds of argument are allowed and they are understood as follows: ffl Variables: A variable stand for any kind of fs. Cooccurrence of variables indicate sharing of substructure. ....

....Each time a lexical rule matches a lexical entry a new lexical entry (as defined by the sharing in the lexical rule) is added to the lexicon. The new entry is not compared to the already present ones before being added. 3. 3 Parsing mechanism The ftfsg parser is an ordinary chart parser (cf. Gazdar and Mellish 1989: Ch 6) This mechanism allows us to apply the grammar to strings (needless to say) and to inspect how it works or fails to work. The locations of words and substrings are defined by pairs of vertices. A vertex is an integer: 0 is the beginning of the input string. 1 separates the first word from ....

Gazdar, G. and Mellish, C., 1989, Natural Language Processing in PROLOG, Addison-Wesley Publishing Company, Wokingham.

....searching for perspicuous, rule based representations of the knowledge required for language processing. Over time, a myriad of frameworks, from augmented transition networks to unification grammars, have been proposed for representing and computing with linguistic knowledge. Allen, 1995; Gazdar Mellish, 1989). Certainly, great progress has been made in understanding the fundamental issues involved in natural language understanding. Modern representations allow grammarians to model many aspects of language elegantly and facilitate the construction of grammars for domain specific language processing ....

Gazdar, G., & Mellish, C. (1989). Natural Language Processing in Prolog. AdisonWesley Publishing Company, New York.

....the preceding table will not correct the token WISCONSEN to WISCONSON . Automatic generation of code to correct mispellings is needed to address this brittleness and is discussed in Section 3.1.3. 3.1. 2 Supertokenization and Correction of Line Breaks Bottom up parsing is easily done in Prolog [3]. As an example, consider that San Francisco and San Lius Obispo are towns in California, while San Luis Potosi is a town (and a province) in Mexico. A simple approach to supertokenizing these input strings could consist of the rules: cftrans( SAN , FRANCISCO Tail] Tail, SAN FRANCISCO ) ....

Gazdar, G., and Mellish, C. Natural Language Processing in Prolog. Addison Wesley, 1989.

....(feature values) in the graph, see Figure 3. Similarly, a lexical item is a single graph. Once a parser has been populated with feature structures there is no need to explicitly build new feature structures as the generative unification operation simply combines feature structure information (see [GM89]) Consequently, after creating unique feature structures for rules and lexical entries, it is possible to encode all other possible feature structures that may be generated as annotations to this original set. The straight forward implementation described here relies on the monotonic qualities ....

Gerald Gazdar and Chris Mellish. Natural Language Processing in PROLOG. Addison Wesley, 1989.

....to the left or to the right) Here the latter strategy is chosen because it makes it possible to keep the grammar simple and because it is implementable without great effort given that a simple chart parser is used. With this approach some problems are obvious: 3 This technique is described in (Gazdar and Mellish, 1989). 24 ffl We cannot find heads in structures where the head can be in alternating positions. ffl The task of describing each head category is complex and a possible source of error. The first case is not a great problem in Swedish but in other languages there are constructions that might cause ....

Gazdar, G. and Mellish, C. (1989). Natural Language Processing in Prolog. Addison -Wesley.

....knowledge engineering, it is necessary for lexicalist approaches to factor away at the lexiconencoding interface as many as possible of the commonalities between lexical items. To this end, we adopt the principles of default inheritance (Gazdar 1987) as embodied in the datr language (Evans and Gazdar 1989b) Areas where abstractions may be made over the lexicon are morphosyntax (Gazdar 1990) and transitivity (Charniak and McDermott 1985, Pollard and Sag 1987, Hudson 1990) and combinations of these leading to lexical rules such as passive. To this we have added the area of lexico semantic ....

....at Node2. For example, evaluating the datr theory in (3) yields the theorems for node Ex2 shown in (4) 3) Ex1: head major = noun head case = nom. Ex2: syn = Ex1: 4) Ex2: syn head major = noun. Ex2: syn head case = nom. For a more detailed description of datr see Evans and Gazdar 1989a. 3.2 The linguistic framework Linguistic knowledge is structured in terms of a simple unification categorial grammar (Calder et al. 1988) in which featural constraints at the levels of morphology, syntax, and semantics may all occur in a lexical sign. The basic sign structure of lexical entries ....

[Article contains additional citation context not shown here]

Gazdar, G. and C. Mellish (1989). Natural Language Processing in Prolog.

....of dependencies can bridge arbitrarily large amounts of intervening structure. However, solutions to the problem of representing DDs in a succinct manner are provided by Transformational Grammars (TGs, Chomsky (1957) Definite Clause Grammars (DCGs, Pereira and Warren (1980) and others (see Gazdar and Mellish (1989)) The similarity between the grammar in Figure 2 of the accompanying article and the set of patterns in Figure 3 of that article might suggest that grammars in the form of patterns suffer the same shortcomings as CF PSGs. The suggestion here is that, given an appropriate system for finding good ....

....time. However, later research has shown that the same kinds of regularities in the syntax of English auxiliary verbs can be described quite well without recourse to transformational rules, using DCGs or other systems which do not use that type of rule (see, for example, Pereira and Warren, 1980; Gazdar and Mellish, 1989). An example showing how English auxiliary verbs may be described using the DCG formalism may be found in Wolff (1987, pp. 183 4) 5.2 ICMAUS and English Auxiliary Verbs Figure 15 shows an ICMAUS grammar for English auxiliary verbs which exploits several of the ideas described earlier in ....

Gazdar, G. and Mellish, C. (1989) Natural Language Processing in Prolog. Addison-Wesley, Wokingham.

....we can adapt a parser to process grammars of this sort. As we will see below, it only takes relatively minor modifications to an existing chart parser for HPSG. 5.4 Chart parser Chart parsing has its origins in the works of M. Kay and R. Kaplan (see a textbook introduction to this technique in Gazdar and Mellish (1989) and references therein) It has been popular in the natural language processing community since the early 1980s for its several advantages: ffl various parsing algorithms are applicable (it does not commit us to either topdown or bottom up approach) ffl the method is neutral with respect to ....

Gazdar, G. and C. S. Mellish (1989). Natural language processing in Prolog. Wokingham: Addison-Wesley Publishing.

....calculi. 4.2 TOBI s architecture Figure 8 presents the architecture of TOBI. Here is a list of TOBI s modules with a short description of their tasks: SYNT: morphoSYNTactic analyzer that parses the input utterance, producing its syntactic structure. SYNT uses a lexicon and a DCG grammar [19] as knowledge bases; SEM: SEMantic analyzer; it takes the syntactic structure produced by SYNT and produces as output the logical form, that is a representation of the utterance in a slot filler notation, in which events and semantic roles are TOWARDS RECURSIVE MODELS . Informatica 21 ....

G. Gazdar and C. S. Mellish. Natural Language Processing in Prolog. Addison Wesley, Wokingham, England, 1989.

.... of non terminal symbols (variables) and terminals (letters or tokens) Chomsky has proposed a four layer hierarchy of languages corresponding to four syntactically caracterizable sets of grammars: the type 0 languages (almost anything) the context sensitive languages (e.g. Swiss German, cf. GM89] or a n b n c n ) the context free languages (e.g. block languages, most programming languages) and the regular languages (language generated by regular expressions) There are three main categories of recognizers: automata and machines, transition network, and definite clause grammars. ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison Wesley, 1989.

....Up Parsing Doug Arnold doug essex.ac. uk In Bottom Up parsing, the idea is to build the parse tree bottom up (i.e. starting with leaves) Two basic approaches: Compile grammar into special form (BUP) e.g. Gazdar and Mellish (1989)(Ch5) Interpret normal grammar on the y , e.g. Pereira and Shieber (1987) pp178 . 1 A Left Corner Algorithm 1.1 The Basic Idea Our problem is to analyze part of a string as an S (in fact, we will only be happy if this is all the string, but we can ignore that for the moment) 1) S Sam ....

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison Wesley, Wokingham, 1989.

No context found.

Gerald Gazdar and Chris Mellish, Natural Language Processing in PROLOG, Addison Wesley, Workingham u.a., 1989.

No context found.

Gerald Gazdar and Chris Mellish. 1989. Natural Language Processing in Prolog. Addison-Wesley, Wokingham, England.

No context found.

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison-Wesley, 1989.

No context found.

Gerald Gazdar and Chris Mellish. Natural Language Processing in PROLOG. Addison-Wesley, 1989.

No context found.

GAZDAR, G., and MELLISH, C. 1989. Natural Language Processing in Prolog. AddisonWesley, Reading, MA.

No context found.

G. Gazdar and C. Mellish. Natural Language Processing in Prolog. Addison-Wesley Publishing Company, Workingham, England, 1989.

No context found.

Gazdar G., Mellish C.: Natural Language Processing in Prolog. Addison-Wesley (1989)

No context found.

Gerald Gazdar and Christopher Mellish. 1989. Natural Language Processing in Prolog. Addison-Wesley.

No context found.

) Gazdar, G. and Mellish, C., Natural Language Processing in Prolog. Wokingham: Addison-Wesley, 1989.

No context found.

Gazdar and Mellish. Natural Language Processing in Prolog, Addison-Wesley, 1989.

No context found.

Blackwell. Gazdar, G. and Mellish, C. (1989). Natural Language Processing in Prolog. Addison-Wesley Publishing Company.

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