Quick shipping. Satisfaction guaranteed! Seller assumes all responsibility for this listing. Shipping and handling This item will ship to Germany , but the seller has not specified shipping options. Contact the seller - opens in a new window or tab and request a shipping method to your location. Shipping cost cannot be calculated. Please enter a valid ZIP Code. This item will be shipped through the Global Shipping Program and includes international tracking.
Please enter a number less than or equal to 1. Select a valid country. Please enter 5 or 9 numbers for the ZIP Code. Delivery times may vary, especially during peak periods. Handling time Will usually ship within 2 business days of receiving cleared payment - opens in a new window or tab. Taxes Taxes may be applicable at checkout. Learn more Return policy After receiving the item, contact seller within Refund will be given as Return shipping 30 days Money back Buyer pays for return shipping Refer to eBay Return policy for more details. You are covered by the eBay Money Back Guarantee if you receive an item that is not as described in the listing.
Payment details Payment methods Special financing available Select PayPal Credit at checkout to have the option to pay over time. Other offers may also be available. Interest will be charged to your account from the purchase date if the balance is not paid in full within 6 months. Minimum monthly payments are required. Subject to credit approval. No ratings or reviews yet.
Skip to main content. Email to friends Share on Facebook - opens in a new window or tab Share on Twitter - opens in a new window or tab Share on Pinterest - opens in a new window or tab. Add to Watchlist. People who viewed this item also viewed. Picture Information. Have one to sell?
Natural Language Processing In Lisp: An Introduction To Computational Linguistics by Gerald Gazdar
Sell now - Have one to sell? Get an immediate offer. Get the item you ordered or get your money back. Learn more - opens in new window or tab. Seller information rockymtntext Contact seller. Visit store. See other items More See all. Item Information Condition:. Examples are seen in the following sentences. Humankind in 3. The name-like character of the term is apparent from the fact that it cannot readily be premodified by an adjective. The subjects in 3. Here - ness is a predicate modifier that transforms the predicate polite , which applies to ordinary usually human individuals, into a predicate over quantities of the abstract stuff, politeness.
This allows for modification of the nominal predicate before reification, in phrases such as fluffy snow or excessive politeness. The subject of 3. Finally 3. Here we can posit a reification operator Ke that maps sentence intensions into kinds of situations. This type of sentential reification needs to be distinguished from that -clause reification, such as appears to be involved in 3. We mentioned the possibility of a modal-logic analysis of 3. The use of reification operators is a departure from a strict Montgovian approach, but is plausible if we seek to limit the expressiveness of our semantic representation by taking predicates to be true or false of individuals, rather than of objects of arbitrarily high types, and likewise take quantification to be over individuals in all cases, i.
Some computational linguists and AI researchers wish to go much further in avoiding expressive devices outside those of standard first-order logic. One strategy that can be used to deal with intensionality within FOL is to functionalize all predicates, save one or two. Here loves is regarded as a function that yields a reified property, while Holds or in some proposals, True , and perhaps equality, are the only predicates in the representation language. Then we can formalize 3. The main practical impetus behind such approaches is to be able to exploit existing FOL inference techniques and technology.
Another important issue has been canonicalization or normalization : What transformations should be applied to initial logical forms in order to minimize difficulties in making use of linguistically derived information? The uses that should be facilitated by the choice of canonical representation include the interpretation of further texts in the context of previously interpreted text and general knowledge , as well as inferential question answering and other inference tasks.
We can distinguish two types of canonicalization: logical normalization and conceptual canonicalization. An example of logical normalization in sentential logic and FOL is the conversion to clause form Skolemized, quantifier-free conjunctive normal form. The rationale is that reducing multiple logically equivalent formulas to a single form reduces the combinatorial complexity of inference. For example, in a geographic domain, we might replace the relations between countries is next to, is adjacent to, borders on, is a neighbor of, shares a border with, etc.
In the domain of physical, communicative, and mental events, we might go further and decompose predicates into configurations of primitive predicates. As in the case of logical normalization, conceptual canonicalization is intended to simplify inference, and to minimize the need for the axioms on which inference is based. A question raised by canonicalization, especially by the stronger versions involving reduction to primitives, is whether significant meaning is lost in this process.
For example, the concept of being neighboring countries, unlike mere adjacency, suggests the idea of side-by-side existence of the populations of the countries, in a way that resembles the side-by-side existence of neighbors in a local community. More starkly, reducing the notion of walking to transporting oneself by moving one's feet fails to distinguish walking from running, hopping, skating, and perhaps even bicycling. Therefore it may be preferable to regard conceptual canonicalization as inference of important entailments, rather than as replacement of superficial logical forms by equivalent ones in a more restricted vocabulary.
- GUCL: Computational Linguistics @ Georgetown?
- Rhetoric of Machine Aesthetics.
- CSC 247/447 (and BCS 235, LIN 247): Natural Language Processing;
- On Lisp: Advanced Techniques for Common Lisp. Paul Graham | SpringerLink.
- Natural Language Processing in LISP - An Introduction to Computational Linguistics (Hardcover)?
- Featured channels!
We will comment further on primitives in the context of the following subsection. While many AI researchers have been interested in semantic representation and inference as practical means for achieving linguistic and inferential competence in machines, others have approached these issues from the perspective of modeling human cognition.
Prior to the s, computational modeling of NLP and cognition more broadly were pursued almost exclusively within a representationalist paradigm, i. In the s, connectionist or neural models enjoyed a resurgence, and came to be seen by many as rivalling representationalist approaches. We briefly summarize these developments under two subheadings below. Some of the cognitively motivated researchers working within a representationalist paradigm have been particularly concerned with cognitive architecture , including the associative linkages between concepts, distinctions between types of memories and types of representations e.
Others have been more concerned with uncovering the actual internal conceptual vocabulary and inference rules that seem to underlie language and thought. Ross Quillian's semantic memory model, and models developed by Rumelhart, Norman and Lindsay Rumelhart et al. A common thread in cognitively motivated theorizing about semantic representation has been the use of graphical semantic memory models, intended to capture direct relations as well as more indirect associations between concepts, as illustrated in Figure This particular example is loosely based on Quillian Quillian suggested that one of the functions of semantic memory, conceived in this graphical way, was to enable word sense disambiguation through spreading activation.
In particular, the activation signals propagating from sense 1 the living-plant sense of plant would reach the concept for the stuff, water , in four steps along the pathways corresponding to the information that plants may get food from water , and the same concept would be reached in two steps from the term water , used as a verb, whose semantic representation would express the idea of supplying water to some target object.
Such conceptual representations have tended to differ from logical ones in several respects. One, as already discussed, has been the emphasis by Schank and various other researchers e. However, this involves a questionable assumption that subtle distinctions between, say, walking to the park, ambling to the park, or traipsing to the park are simply ignored in the interpretive process, and as noted earlier it neglects the possibility that seemingly insignificant semantic details are pruned from memory after a short time, while major entailments are retained for a longer time.
Another common strain in much of the theorizing about conceptual representation has been a certain diffidence concerning logical representations and denotational semantics. The relevant semantics of language is said to be the transduction from linguistic utterances to internal representations, and the relevant semantics of the internal representations is said to be the way they are deployed in understanding and thought. For both the external language and the internal mentalese representation, it is said to be irrelevant whether or not the semantic framework provides formal truth conditions for them.
The rejection of logical semantics has sometimes been summarized in the dictum that one cannot compute with possible worlds. However, it seems that any perceived conflict between conceptual semantics and logical semantics can be resolved by noting that these two brands of semantics are quite different enterprises with quite different purposes. Certainly it is entirely appropriate for conceptual semantics to focus on the mapping from language to symbolic structures in the head, realized ultimately in terms of neural assemblies or circuits of some sort , and on the functioning of these structures in understanding and thought.
But logical semantics, as well, has a legitimate role to play, both in considering how words and larger linguistic expressions relate to the world and how the symbols and expressions of the internal semantic representation relate to the world. This role is metatheoretic in that the goal is not to posit cognitive entities that can be computationally manipulated, but rather to provide a framework for theorizing about the relationship between the symbols people use, externally in language and internally in their thinking, and the world in which they live.
It is surely undeniable that utterances are at least sometimes intended to be understood as claims about things, properties, and relationships in the world, and as such are at least sometimes true or false. It would be hard to understand how language and thought could have evolved as useful means for coping with the world, if they were incapable of capturing truths about it. Moreover, logical semantics shows how certain syntactic manipulations lead from truths to truths regardless of the specific meanings of the symbols involved in these manipulations and these notions can be extended to uncertain inference, though this remains only very partially understood.
Thus, logical semantics provides a basis for assessing the soundness or otherwise of inference rules. While human reasoning as well as reasoning in practical AI systems often needs to resort to unsound methods abduction, default reasoning, Bayesian inference, analogy, etc. A strong indication that cognitively motivated conceptual representations of language are reconcilable with logically motivated ones is the fact that all proposed conceptual representations have either borrowed deliberately from logic in the first place in their use of predication, connectives, set-theoretic notions, and sometimes quantifiers or can be transformed to logical representations without much difficulty, despite being cognitively motivated.
As noted earlier, the s saw the re-emergence of connectionist computational models within mainstream cognitive science theory e. We have already briefly characterized connectionist models in our discussion of connectionist parsing. But the connectionist paradigm was viewed as applicable not only to specialized functions, but to a broad range of cognitive tasks including recognizing objects in an image, recognizing speech, understanding language, making inferences, and guiding physical behavior. The emphasis was on learning, realized by adjusting the weights of the unit-to-unit connections in a layered neural network, typically by a back-propagation process that distributes credit or blame for a successful or unsuccessful output to the units involved in producing the output Rumelhart and McClelland From one perspective, the renewal of interest in connectionism and neural modeling was a natural step in the endeavor to elaborate abstract notions of cognitive content and functioning to the point where they can make testable contact with brain theory and neuroscience.
But it can also be seen as a paradigm shift, to the extent that the focus on subsymbolic processing began to be linked to a growing skepticism concerning higher-level symbolic processing as models of mind, of the sort associated with earlier semantic network-based and rule-based architectures. For example, Ramsay et al. But others have continued to defend the essential role of symbolic processing.
For example, Anderson , contended that while theories of symbolic thought need to be grounded in neurally plausible processing, and while subsymbolic processes are well-suited for exploiting the statistical structure of the environment, nevertheless understanding the interaction of these subsymbolic processes required a theory of representation and behavior at the symbolic level. What would it mean for the semantic content of an utterance to be represented in a neural network, enabling, for example, inferential question-answering?
The input modifies the activity of the network and the strengths of various connections in a distributed way, such that the subsequent behavior of the network effectively implements inferential question-answering. However, this leaves entirely open how a network would learn this sort of behavior. The most successful neural net experiments have been aimed at mapping input patterns to class labels or to other very restricted sets of outputs, and they have required numerous labeled examples e. A less radical alternative to the eliminativist position, termed the subsymbolic hypothesis , was proposed by Smolensky , to the effect that mental processing cannot be fully and accurately described in terms of symbol manipulation, requiring instead a description at the level of subsymbolic features, where these features are represented in a distributed way in the network.
Such a view does not preclude the possibility that assemblies of units in a connectionist system do in fact encode symbols and more complex entities built out of symbols, such as predications and rules. It merely denies that the behavior engendered by these assemblies can be adequately modelled as symbol manipulation.
In fact, much of the neural net research over the past two or three decades has sought to understand how neural nets can encode symbolic information e. Distributed schemes associate a set of units and their activation states with particular symbols or values. For example, Feldman proposes that concepts are represented by the activity of a cluster of neurons; triples of such clusters representing a concept, a role, and a filler value are linked together by triangle nodes to represent simple attributes of objects.
Language understanding is treated as a kind of simulation that maps language onto a more concrete domain of physical action or experience, guided by background knowledge in the form of a temporal Bayesian network. Global schemes encode symbols in overlapping fashion over all units. Propositional symbols can then be interpreted in terms of such states, and truth functions in terms of simple max-min operations and sign inversions performed on network states.
See Blutner, ; however, Blutner ultimately focuses on a localist scheme in which units represent atomic propositions and connections represent biconditionals. Holographic neural network schemes e. A distinctive characteristic of such networks is their ability to classify or reconstruct patterns from partial or noisy inputs. The status of the subsymbolic hypothesis remains an issue for debate and further research. Certainly it is unclear how symbolic approaches could match certain characteristics of neural network approaches, such as their ability to cope with novel instances and their graceful degradation in the face of errors or omissions.
Researchers more concerned with practical advances than biologically plausible modeling have also explored the possibility of hybridizing the symbolic and subsymbolic approaches, in order to gain the advantages of both e. A quite formal example of this, drawing on ideas by Dov Gabbay, is d'Avila Garcez Finally, we should comment on the view expressed in some of the cognitive science literature that mental representations of language are primarily imagistic e.
Certainly there is ample evidence for the reality and significance of mental imagery Johnson-Laird ; Kosslyn But as was previously noted, symbolic and imagistic representations may well coexist and interact synergistically. Moreover, cognitive scientists who explore the human language faculty in detail, such as Steven Pinker , or any of the representationalist or connectionist researchers cited above, all seem to reach the conclusion that the content derived from language and the stuff of thought itself is in large part symbolic—except in the case of the eliminativists who deny representations altogether.
It is not hard to see, however, how raw intuition might lead to the meanings-as-images hypothesis. It appears that vivid consciousness is associated mainly with the visual cortex, especially area V1, which is also crucially involved in mental imagery e. Consequently it is entirely possible that vast amounts of non-imagistic encoding and processing of language go unnoticed, while any evoked imagistic artifacts become part of our conscious experience.
Further, the very act of introspecting on what sort of imagery, if any, is evoked by a given sentence may promote construction of imagery and awareness thereof. In its broadest sense, statistical semantics is concerned with semantic properties of words, phrases, sentences, and texts, engendered by their distributional characteristics in large text corpora. For example, terms such as cheerful, exuberant, and depressed may be considered semantically similar to the extent that they tend to occur flanked by the same or in turn similar nearby words.
For some purposes, such as information retrieval, identifying labels of documents may be used as occurrence contexts. Through careful distinctions among various occurrence contexts, it may also be possible to factor similarity into more specific relations such as synonymy, entailment, and antonymy. One basic difference between standard logical semantic relations and relations based on distributional similarity is that the latter are a matter of degree.
Further, the underlying abstractions are very different, in that statistical semantics does not relate strings to the world, but only to their contexts of occurrence a notion similar to, but narrower than, Wittgenstein's notion of meaning as use. However, statistical semantics does admit elegant formalizations. Various concepts of similarity and other semantic relations can be captured in terms of vector algebra, by viewing the occurrence frequencies of an expression as values of the components of a vector, with the components corresponding to the distinct contexts of occurrence.
But how does this bear on meaning representation of natural language sentences and texts? In essence, the representation of sentences in statistical semantics consists of the sentences themselves. The idea that sentences can be used directly, in conjunction with distributional knowledge, as objects enabling inference is a rather recent and surprising one, though it was foreshadowed by many years of work on question answering based on large text corpora. Recognizing textual entailment requires judgments as to whether one given linguistic string entails a second one, in a sense of entailment that accords with human intuitions about what a person would naturally infer with reliance on knowledge about word meanings, general knowledge such as that any person who works for a branch of a company also works for that company, and occasional well-known specific facts.
Some examples are intermediate; e. Initial results in the annual competitions were poor not far above the random guessing mark , but have steadily improved, particularly with the injection of some reasoning based on ontologies and on some general axioms about the meanings of words, word classes, relations, and phrasal patterns e. It is noteworthy that the conception of sentences as meaning representations echoes Montague's contention that language is logic. But research in textual entailment seems to be moving towards a similar conception, as exemplified in the work of Dagan et al. One way of construing degrees of entailment in this framework is in terms of the entailment probabilities relating each possible logical form of the premise sentence to each possible logical form of the hypothesis in question.
Having surveyed three rather different brands of semantics, we are left with the question of which of these brands serves best in computational linguistic practice. If the goal, for example, is to create a dialogue-based problem-solving system for circuit fault diagnosis, emergency response, medical contingencies, or vacation planning, then an approach based on logical or at least symbolic representations of the dialogue, underlying intentions, and relevant constraints and knowledge is at present the only viable option. Here it is of less importance whether the symbolic representations are based on some presumed logical semantics for language, or some theory of mental representation—as long as they are representations that can be reasoned with.
The most important limitations that disqualify subsymbolic and statistical representations of meaning for such purposes are their very limited inferential reach and response capabilities. They provide classifications or one-shot inferences rather than reasoning chains, and they do not generate plans, justifications, or extended linguistic responses. However, both neural net techniques and statistical techniques can help to improve semantic processing in dialogue systems, for example by disambiguating word senses, or recognizing which of several standard plans is being proposed or followed, on the basis of observed utterances or actions.
On the other hand, if the computational goal is to demonstrate human-like performance in a biologically plausible or biologically valid! However, to the extent that language is symbolic, and is a cognitive phenomenon, subsymbolic theories must ultimately explain how language can come about.
In the case of statistical semantics, practical applications such as question-answering based on large textual resources, retrieval of documents relevant to a query, or machine translation are at present greatly superior to logical systems that attempt to fully understand both the query or text they are confronted with and the knowledge they bring to bear on the task. But some of the trends pointed out above in trying to link subsymbolic and statistical representations with symbolic ones indicate that a gradual convergence of the various approaches to semantics is taking place.
For the next few paragraphs, we shall take semantic interpretation to refer to the process of deriving meaning representations from a word stream, taking for granted the operation of a prior or concurrent parsing phase. In other words, we are mapping syntactic trees to logical forms or whatever our meaning representation may be. In the heyday of the proceduralist paradigm, semantic interpretation was typically accomplished with sets of rules that matched patterns to parts of syntactic trees and added to or otherwise modified the semantic representations of input sentences.
The completed representations might either express facts to be remembered, or might themselves be executable commands, such as formal queries to a database or high-level instructions placing one block on another in a robot's simulated or real world. When it became clear in the early s, however, that syntactic trees could be mapped to semantic representations by using compositional semantic rules associated with phrase structure rules in one-to-one fashion, this approach became broadly favored over pure proceduralist ones. In our earlier discussion in section 3.
There we saw sample interpretive rules for a small number of phrase structure rules and vocabulary. The interpretive rules are repeated at the tree nodes from section 3. As can be seen, the Montagovian treatment of NPs as second-order predicates leads to some complications, and these are exacerbated when we try to take account of quantifier scope ambiguity.
We mentioned Montague's use of multiple parses, the Cooper-storage approach, and the unscoped-quantifier approach to this issue in section 3. It is easy to see that multiple unscoped quantifiers will give rise to multiple permutations of quantifier order when the quantifiers are brought to the sentence level. At this point we should pause to consider some interpretive methods that do not conform with the above very common but not universally employed syntax-driven approach. First, Schank and his collaborators emphasized the role of lexical knowledge, especially primitive actions used in verb decomposition, and knowledge about stereotyped patterns of behavior in the interpretive process, nearly to the exclusion of syntax.
These ideas had considerable appeal, and led to unprecedented successes in machine understanding of some paragraph-length stories. Another approach to interpretation that subordinates syntax to semantics is one that employs domain-specific semantic grammars Brown and Burton While these resemble context-free syntactic grammars perhaps procedurally implemented in ATN-like manner , their constituents are chosen to be meaningful in the chosen application domain. For example, an electronics tutoring system might employ categories such as measurement, hypothesis , or transistor instead of NP, and fault-specification or voltage-specification instead of VP.
The importance of these approaches lay in their recognition of the fact that knowledge powerfully shapes our ultimate interpretation of text and dialogue, enabling understanding even in the presence of noisy, flawed, and partial linguistic input. Statistical NLP has only recently begun to be concerned with deriving interpretations usable for inference and question answering and as pointed out in the previous subsection, some of the literature in this area assumes that the NL text itself can and should be used as the basis for inference.
We will mention examples of this type of work, and comment on its prospects, in section 8. We noted earlier that language is potentially ambiguous at all levels of syntactic structure, and the same is true of semantic content, even for syntactically unambiguous words, phrases and sentences. For example, words like bank , recover , and cool have multiple meanings even as members of the same lexical category; nominal compounds such as ice bucket, ice sculpture, olive oil, or baby oil leave unspecified the underlying relation between the nominals such as constituency or purpose.
Many techniques have been proposed for dealing with the various sorts of semantic ambiguities, ranging from psychologically motivated principles, to knowledge-based methods, heuristics, and statistical approaches. Psychologically motivated principles are exemplified by Quillian's spreading activation model described earlier and the use of selectional preferences in word sense disambiguation. Examples of knowledge-based disambiguation would be the disambiguation of ice sculpture to a constitutive relation based on the knowledge that sculptures may be carved or constructed from solid materials, or the disambiguation of a man with a hat to a wearing -relation based on the knowledge that a hat is normally worn on the head.
The possible meanings may first be narrowed down using heuristics concerning the limited types of relations typically indicated by nominal compounding or by with -modification. Heuristic principles used in scope disambiguation include island constraints quantifiers such as every and most cannot expand their scope beyond their local clause and differing wide-scoping tendencies for different quantifiers e.
Statistical approaches typically extract various features in the vicinity of an ambiguous word or phrase that are thought to influence the choice to be made, and then make that choice with a classifier that has been trained on an annotated text corpus. The features used might be particular nearby words or their parts of speech or semantic categories, syntactic dependency relations, morphological features, etc..
Such techniques have the advantage of learnability and robustness, but ultimately will require supplementation with knowledge-based techniques. For example, the correct scoping of quantifiers in contrasting sentence pairs such as. For example,. Thus in general appears to be the implicit default adverbial.
But when the quantifying adverb is present, the sentence admits both an atemporal reading according to which many purebred racehorses are characteristically skittish, as well as a temporal reading to the effect that purebred racehorses in general are subject to frequent episodes of skittishness.
If we replace purebred by at the starting gate , then only the episodic reading of skittish remains available, while often may quantify over racehorses, implying that many are habitually skittish at the starting gate, or it may quantify over starting-gate situations, implying that racehorses in general are often skittish in such situations; furthermore, making formal sense of the phrase at the starting gate evidently depends on knowledge about horse racing scenarios. The interpretive challenges presented by such sentences are or should be of great concern in computational linguistics, since much of people's general knowledge about the world is most naturally expressed in the form of generic and habitual sentences.
Systematic ways of interpreting and disambiguating such sentences would immediately provide a way of funneling large amounts of knowledge into formal knowledge bases from sources such as lexicons, encyclopedias, and crowd-sourced collections of generic claims such as those in Open Mind Common Sense e. Many theorists assume that the logical forms of such sentences should be tripartite structures with a quantifier that quantifies over objects or situations, a restrictor that limits the quantificational domain, and a nuclear scope main clause that makes an assertion about the elements of the domain e.
The challenge lies in specifying a mapping from surface structure to such a logical form. While many of the principles underlying the ambiguities illustrated above are reasonably well understood, general interpretive algorithms are still lacking. The dividing line between semantic interpretation computing and disambiguating logical forms and discourse understanding—making sense of text—is a rather arbitrary one. Language has evolved to convey information as efficiently as possible, and as a result avoids lengthy identifying descriptions and other lengthy phrasings where shorter ones will do.
The reverse sequencing, cataphora , is seen occasionally as well. Determining the co referents of anaphors can be approached in a variety of ways, as in the case of semantic disambiguation. Linguistic and psycholinguistic principles that have been proposed include gender and number agreement of coreferential terms, C-command principles e. An early heuristic algorithm that employed several features of this type to interpret anaphors was that of Hobbs But selectional preferences are important as well.
Another complication concerns reference to collections of entities, related entities such as parts , propositions, and events that can become referents of pronouns such as they, this, and that or of definite NPs such as this situation or the door of the house without having appeared explicitly as a noun phrase. Like other sorts of ambiguity, coreference ambiguity has been tackled with statistical techniques. These typically take into account factors like those mentioned, along with additional features such as antecedent animacy and prior frequency of occurrence, and use these as probabilistic evidence in making a choice of antecedent e.
Parameters of the model are learned from a corpus annotated with coreference relations and the requisite syntactic analyses. Coming back briefly to nominal compounds of form N N, note that unlike conventional compounds such as ice bucket or ice sculpture —ones approachable using an enriched lexicon, heuristic rules, or statistical techniques—some compounds can acquire a variety of meanings as a function of context. For example, rabbit guy could refer to entirely different things in a story about a fellow wearing a rabbit suit, or one about a rabbit breeder, or one about large intelligent leporids from outer space.
Such examples reveal certain parallels between compound nominal interpretation and anaphora resolution: At least in the more difficult cases, N N interpretation depends on previously seen material, and on having understood crucial aspects of that previous material in the current example, the concepts of wearing a rabbit suit, being a breeder of rabbits, or being a rabbit-like creature. In other words N N interpretation, like anaphora resolution, is ultimately knowledge-dependent, whether that knowledge comes from prior text, or from a preexisting store of background knowledge.
A strong version of this view is seen in the work of Fan et al. For example, in a chemistry context, HCL solution is assumed to require elaboration into something like: solution whose base is a chemical whose basic structural constituents are HCL molecules. Algorithms are provided and tested empirically that search for a relational path subject to certain general constraints from the modified N to the modifying N, selecting such a relational path as the meaning of the N N compound.
As the authors note, this is essentially a spreading-activation algorithm, and they suggest more general application of this method see section 5. One pervasive phenomenon of this type is of course ellipsis , as illustrated earlier in sentences 2. Interpreting ellipsis requires filling in of missing material; this can often be found at the surface level as a sequence of consecutive words as in the gapping and bare ellipsis examples 2.
Further complications arise when the imported material contains referring expressions, as in the following variant of 5. Here the missing material may refer either to Felix's boss or my boss called the strict and sloppy reading respectively , a distinction that can be captured by regarding the logical form of the antecedent VP as containing only one, or two, occurrences of the lambda-abstracted subject variable, i. The two readings can be thought of as resulting respectively from scoping his boss first, then filling in the elided material, and the reverse ordering of these operations Dalrymple et al.
Other challenging forms of ellipsis are event ellipsis, as in 5. In applications these and some other forms of ellipsis are handled, where possible, by a making strong use of domain-dependent expectations about the types of information and speech acts that are likely to occur in the discourse, such as requests for flight information in an air travel adviser; and b interpreting utterances as providing augmentations or modifications of domain-specific knowledge representations built up so far.
Corpus-based approaches to ellipsis have so far focused mainly on identifying instances of VP ellipsis in text, and finding the corresponding antecedent material, as problems separate from that of computing correct logical forms e. They would like the machine to read an unprepared text, to test it for correctness, to execute the instructions contained in the text, or even to comprehend it well enough to produce a reasonable response based on its meaning. Human beings want to keep for themselves only the final decisions. The necessity for intelligent automatic text processing arises mainly from the following two circumstances, both being connected with the quantity of the texts produced and used nowadays in the world:.
For example, a secretary in an office cannot take into consideration each time the hundreds of various rules necessary to write down a good business letter to another company, especially when he or she is not writing in his or her native language. It is just cheaper to teach the machine once to do this work, rather than repeatedly teach every new generation of computer users to do it by themselves.
For example, to find information in the Internet on, let us say, the expected demand for a specific product in the next month, a lot of secretaries would have to read texts for a hundred years without eating and sleeping, looking through all the documents where this information might appear. In such cases, using a computer is the only possible way to accomplish the task. Thus, the processing of natural language has become one of the main problems in information exchange. The rapid development of computers in the last two decades has made possible the implementation of many ideas to solve the problems that one could not even imagine being solved automatically, say, 45 years ago, when the first computers appeared.
Intelligent natural language processing is based on the science called computational linguistics. Computational linguistics is closely connected with applied linguistics and linguistics in general. Therefore, we shall first outline shortly linguistics as a science belonging to the humanities. Linguistics is a science about natural languages. It studies the general structure of various natural languages and discovers the universal laws of functioning of natural languages.
Many concepts from general linguistics prove to be necessary for any researcher who deals with natural languages. General linguistics is a fundamental science that was developed by many researchers during the last two centuries, and it is largely based on the methods and results of grammarians of older times, beginning from the classical antiquity. As far as general linguistics is concerned, its most important parts are the following:. Structure of linguistic science. Semantics deals with the meaning of individual words and entire texts, and pragmatics studies the motivations of people to produce specific sentences or texts in a specific situation.
The second name is explained by the fact that comparison is the main method in this branch of linguistics. Comparative linguistics is even older than general linguistics, taking its origin from the eighteenth century. Many useful notions of general linguistics were adopted directly from comparative linguistics. Therefore, diachrony describes changes of a language along the time axis.
Comparative study reveals many common words and constructions within each of the mentioned families—Romance, Germanic, and Slavonic—taken separately. At the same time, it has noticed a number of similar words among these families. This finding has led to the conclusion that the mentioned families form a broader community of languages, which was called Indo-European languages. Several thousand years ago, the ancestors of the people now speaking Romance, Germanic, and Slavonic languages in Europe probably formed a common tribe or related tribes.
At the same time, historic studies permits to explain why English has so many words in common with the Romance family, or why Romanian language has so many Slavonic words these are referred to as loan words. Comparative linguistics allows us to predict the elements of one language based on our knowledge of another related language. For example, it is easy to guess the unknown word in the following table of analogy :. S panis h. Based on more complicated phonologic laws, it is possible even to predict the pronunciation of the French word for the Spanish agua namely [o], eau in the written form , though at the first glance these two words are quite different actually, both were derived from the Latin word aqua.
As to computational linguistics, it can appeal to diachrony, but usually only for motivation of purely synchronic models. History sometimes gives good suggestions for description of the current state of language, helping the researcher to understand its structure. The following are examples of classification of languages not connected with their origin.
Part of speech is defined as a large group of words having some identical morphologic and syntactic properties. Romance and Germanic languages use articles, as well as Bulgarian within the Slavonic family. Meantime, many other languages do not have articles nearly all Slavonic family and Lithuanian, among others. The availability of articles influences some other features of languages. Latin had nominative direct case and five oblique cases: genitive, dative, accusative, ablative, and vocative.
Russian has also six cases, and some of them are rather similar in their functions to those of Latin. Inflected parts of speech, i. In English, there is only one oblique case, and it is applicable only to some personal pronouns: me, us, him, her, them. In Spanish, two oblique cases can be observed for personal pronouns, i. Grammatical cases give additional mean for exhibiting syntactic dependencies between words in a sentence. The subject is in a standard form i.
This is referred to as non-ergative construction. Meantime, a multiplicity of languages related to various other families, not being cognate to each other, are classified as ergative languages. In a sentence of an ergative language, the agent of the action is in a special oblique called ergative case, whereas the object is in a standard form. All ergative languages are considered typologically similar to each other, though they might not have any common word. The similarity of syntactical structures unites them in a common typological group. It suffices to compare the words and their combinations you use in your own formal documents and in conversations with your friends.
It can be said that dialectology describes variations of a language throughout the space axis while diachrony goes along the time axis. For example, in different Spanish-speaking countries, many words, word combinations, or even grammatical forms are used differently, not to mention significant differences in pronunciation.
Natural Language Processing in LISP - an introduction to computational linguistics
A study of Mexican Spanish, among other variants of Spanish language is a good example of a task in the area of dialectology. The results of lexicography are very important for many tasks in computational linguistics, since any text consists of words. Any automatic processing of a text starts with retrieving the information on each word from a computer dictionary compiled beforehand.
Among areas of its special interest, psycholinguists studies teaching language to children, links between the language ability in general and the art of speech, as well as other human psychological features connected with natural language and expressed through it. Mathematical linguistics. There are two different views on mathematical linguistics. In the narrower view, the term mathematical linguistics is used for the theory of formal grammars of a specific type referred to as generative grammars.
This is one of the first purely mathematical theories devoted to natural language. Alternatively, in the broader view, mathematical linguistics is the intersection between linguistics and mathematics, i. Since the theory of generative grammars is nowadays not unique among linguistic applications of mathematics, we will follow the second, broader view on mathematical linguistics.
One of the branches of mathematical linguistics is quantitative linguistic. It studies language by means of determining the frequencies of various words, word combinations, and constructions in texts. Currently, quantitative linguistics mainly means statistical linguistics. It provides the methods of making decisions in text processing on the base of previously gathered statistics. Another application of statistical methods is in the deciphering of texts in forgotten languages or unknown writing systems.
Until the middle of the twentieth century, applications of linguistics were limited to developing and improving grammars and dictionaries in a printed form oriented to their broader use by non-specialists, as well as to the rational methods of teaching natural languages, their orthography and stylistics. This was the only purely practical product of linguistics. The ancient Mayan writing system was deciphered with statistical methods. In the latter half of the twentieth century, a new branch of applied linguistics arose, namely the computational , or engineering , linguistics. Actually, this is the main topic of this book, and it is discussed in some detail in the next section.
The processing of natural language should be considered here in a very broad sense that will be discussed later. The algorithms, the corresponding programs, and the programming technologies can vary, while the basic linguistic principles and methods of their description are much more stable. A broader set of notions and models of general linguistics and mathematical linguistics are described below. For the purposes of this course, it is also useful to draw a line between the issues in text processing we con sider linguistic—and thus will discuss below—and the ones we will not.
In our opinion, for a computer system or its part to be considered linguistic, it should use some data or procedures that are:. Thus, not every program dealing with natural language texts is related to linguistics. Thus, they are language-dependent. However, they do not rely on large enough linguistic resources. Therefore, simple hyphenation programs only border upon the software that can be considered linguistic proper.
As to spell checkers that use a large word list and complicated morphologic tables, they are just linguistic programs. As it could be noticed, the term word was used in the previous sections very loosely. Its meaning seems obvious: any language operates with words and any text or utterance consists of them.
This notion seems so simple that, at the first glance, it does not require any strict definition or further explanation: one can think that a word is just a substring of the text as a letter string, from the first delimiter usually, a space to the next one usually, a space or a punctuation mark. Nevertheless, the situation is not so simple.
How many words does it contain? One can say 14 and will be right, since there are just 14 letter substrings from one delimiter to another in this sentence. For these observations, no linguistic knowledge is necessary. However, one can also notice that devuelvo and devuelves are forms of the same verb devolver , and libros and libro are forms of the same noun libro , so that the number of different words is only Indeed, these pairs of wordforms denote the same action or thing.
If one additionally notices that the article los is essentially equivalent to the article el whereas the difference in grammatical number is ignored, then there are only 10 different words in this sentence. At last, one can consider me the same as yo , but given in oblique grammatical case , even though there are no letters in common in these substrings. For such an approach, the total number of different words is nine. We can conclude from the example that the term word is too ambiguous to be used in a science with the objective to give a precise description of a natural language.
To introduce a more consistent terminology, let us call an individual substring used in a specific place of a text without taking into account its possible direct repetitions or similarities to other substrings a word occurrence. Now we can say that the sentence above consisted of 14 word occurrences. Some of the substrings usually similar in the appearance have the same core meaning. We intuitively consider them as different forms of some common entity. A set of such forms is called lexeme. Each entry of such a set—a letter string without regard to its position in the text—is called wordform.
Each word occurrence represents a wordform, while wordforms but not word occurrences can repeat in the text. Now we can say that the sentence in the example above contains 14 word occurrences, 13 different wordforms, or nine different lexemes. The considerations that gave other figures in the example above are linguistically inconsistent. A lexeme is identified by a name. Usually, one of its wordforms, i. Just these names are used as titles of the corresponding entries in dictionaries mentioned above. The dictionaries cover available information about lexemes of a given language, sometimes including morphologic information, i.
Various dictionaries compiled for the needs of lexicography, dialectology, and sociolinguistics have just lexemes as their entries rather than wordforms. Therefore, the term word , as well as its counterparts in other languages, such as Spanish palabra , is too ambiguous to be used in a linguistic book.
Instead, we should generally use the terms word occurrence for a specific string in a specific place in the text, wordform for a string regardless to its specific place in any text, and lexeme for a theoretical construction uniting several wordforms corresponding to a common meaning in the manner discussed above. However, sometimes we will retain the habitual word word when it is obvious which of these more specific terms is actually meant.
In the past few decades, many attempts to build language processing or language understanding systems have been undertaken by people without sufficient knowledge in theoretical linguistics. They hoped that they would succeed thanks to clever mathematical algorithms, fine programming in Assembler language, or just the speed of their computers.
To our knowledge, all such attempts have failed. Even now it is still worthwhile to explain the necessity to have fundamental knowledge for those who would develop such systems, and thus to clarify why we decided to start a course in computational linguistics from notions of general linguistics.
General linguistics is a fundamental science belonging to the humanities. An analogy with another fundamental science—physics—is appropriate here. A specialist with deep knowledge of physics would easily understand the structure of any new electrical device.
- Satellite Orbits;
- The 80s (Designers & Dragons, Volume 2)!
- ALU: Lisp Books.
- European Masters in Language and Speech.
- Metric Handbook - Planning and Design Data;
- Computational Linguistcis: Models, Resources, Applications;
- Development as Freedom.
Moreover, since the fundamentals of physics are changing very slowly, such a specialist would be able to understand those new or revolutionary engineering principles that did not even exist at the time when he or she studied in a university. Indeed, the underlying fundamentals of physics have remained the same. On the contrary, somebody with narrow specialization only in laser devices might be perplexed with any new principle that does not relate to his or her deep but narrow knowledge.
The same situation can be observed with linguistics. Even experienced programmers who have taken solid courses on software systems for natural language processing might become helpless in cases where a deeper penetration into linguistic notions and laws is needed. What is more, they will hardly be able to compile a formal grammar, grammar tables, or a computer dictionary of a natural language, whereas the program heavily relies on such tables and dictionaries as on its integral parts.
They will not even be able to understand the suggestions of professional linguists who regrettably in many cases prove to know little about programming and will not be able to explain to them the meaning of the data structures the program needs to use. On the contrary, a good background in linguistics will allow the new specialists in computational linguistics to work productively in an interdisciplinary team, or just to compile all the necessary tables and dictionaries on their own. It might even guide them to some quite new ideas and approaches.
These specialists will better understand the literature on the subject, which is very important in our rapidly developing world. In our books, the stress on Spanish language is made intentionally and purposefully. For historical reasons, the majority of the literature on natural languages processing is not only written in English, but also takes English as the target language for these studies.
In our opinion, this is counter-productive and thus it has become one of the causes of a lag in applied research on natural language processing in many countries, compared to the United States. The Spanish-speaking world is not an exception to this rule. The number of Spanish-speaking people in the world has exceeded now million, and Spanish is one of the official languages of the United Nations. As to the human-oriented way of teaching, Spanish is well described, and the Royal Academy of Madrid constantly supports orthographic  and grammatical research  and standardization.
There are also several good academic-type dictionaries of Spanish, one the best of which being . However, the lexicographic research reflected in these dictionaries is too human-oriented. Formal description and algorithmization of a language is the objective of research teams in computational linguistics.
Several teams in this field oriented to Spanish work now in Barcelona and Madrid, Spain. However, even this is rather little for a country of the European Community, where unilingual and multilingual efforts are well supported by the governmental and international agencies.
Now, the team headed by Prof. A very useful dictionary of modern Mexican Spanish, developed by the team headed by Prof. Some additional information on Spanish-oriented groups can be found in the Appendix on the page One of the most powerful corporations in the world, Microsoft, has announced the development of a natural language processing system for Spanish based on the idea of multistage processing. As usually with commercial developments, the details of the project are still rather scarce. We can only guess that a rather slow progress of the grammar checker of Word text processor for Windows is related somehow with these developments.
Thus, one who needs to compile all facts of Spanish relevant for its automatic processing faces with a small set of rather old monographs and manuals oriented to human learners, mainly written and first published in Spain and then sometimes reprinted elsewhere in Latin America. Meantime, a development of natural language processing tools is quite necessary for any country to be competitive in the twenty-first century. We hope that our books will contribute to such developments in Mexico.
The twenty-first century will be the century of the total information revolution. The development of the tools for the automatic processing of the natural language spoken in a country or a whole group of countries is extremely important for the country to be competitive both in science and technology. To develop such applications, specialists in computer science need to have adequate tools to investigate language with a view to its automatic processing. One of such tools is a deep knowledge of both computational linguistics and general linguistic science.
A course on linguistics usually follows one of the general models, or theories, of natural language, as well as the corresponding methods of interpretation of the linguistic phenomena. A comparison with physics is appropriate here once more. For a long time, the Newtonian theory had excluded all other methods of interpretation of phenomena in mechanics. Such exclusivity can be explained by the great power of purely mathematical description of natural phenomena in physics, where theories describe well-known facts and predict with good accuracy the other facts that have not yet been observed.
In general linguistics, the phenomena under investigation are much more complicated and variable from one object i. Therefore, the criteria for accuracy of description and prediction of new facts are not so clear-cut in this field, allowing different approaches to coexist, affect each other, compete, or merge. Because of this, linguistics has a rich history with many different approaches that formed the basis for the current linguistic theories. Let us give now a short retrospective of the development of general linguistics in the twentieth century. The reader should not be worried if he or she does not know many terms in this review not yet introduced in this book.
There will be another place for strict definitions in this book. He considered natural language as a structure of mutually linked elements, similar or opposed to each other. Later, several directions arose in general linguistics and many of them adhered to the same basic ideas about language. This common method was called structuralism , and the corresponding scientific research not only in linguistics, but also in other sciences belonging to the humanities was called structuralist.
Most of them worked in Europe, and European structuralism kept a significant affinity to the research of the previous periods in its terminology and approaches. The order of words in a sentence was considered the main tool to become aware of word grouping and sentence structures. At this period, almost every feature of English seemed to confirm this postulate.
The sentences under investigation were split into the so-called immediate constituents , or phrases , then these constituents were in their turn split into subconstituents, etc. Such a method of syntactic structuring was called the phrase structure , or con stituency , approach. Among the formal tools developed by Chomsky and his followers, the two most important components can be distinguished:.
The generative grammars produce strings of symbols, and sets of these strings are called formal languages , whereas in general linguistics they could be called texts. Chomskian hierarchy is taught to specialists in computer science, usually in a course on languages and automata. This redeems us from necessity to go into any details. The con text-free grammars constitute one level of this hierarchy. Just examples of these first attempts are extensively elaborated in the manuals on artificial intelligence.
It is a good approach unless a student becomes convinced that it is the only possible. Let us consider an example of a context-free grammar for generating very simple English sentences. It uses the initial symbol S of a sentence to be generated and several other non-terminal symbols: the noun phrase symbol NP , verb phrase symbol VP , noun symbol N , verb symbol V , and determinant symbol D. All these non-terminal symbols are interpreted as grammatical categories.
In our simple case, let the set of the rules be the following:. Each symbol at the right side of a rule is considered a constituent of the entity symbolized at the left side. An additional set of rules is taken to convert all these non-terminal symbols to the terminal symbols corresponding to the given grammatical categories.
The terminals are usual words of Spanish, English, or any other language admitting the same categories and the same word order. Let the rules be the following:. With more complicate rules, some types of ungrammaticality can be eliminated. However, to fully get rid of potentially meaningless sentences is very difficult, since from the very beginning the initial symbol does not contain any specific meaning at all.
It merely presents an abstract category of a sentence of a very vast class, and the resulting meaning or nonsense is accumulated systematically, with the development of each constituent. Syntactic structure of a sentence was identified with the so-called con stituency tree. In other words, this is a nested structure subdividing the sentence into parts, then these parts into smaller parts, and so on. This decomposition corresponds to the sequence of the grammar rules applications that generate the given sentence. Example of constituency tree. Further research revealed great generality, mathematical elegance, and wide applicability of generative grammars.
They became used not only for description of natural languages, but also for specification of formal languages, such as those used in mathematical logic , pattern recognition, and programming languages. During the next three decades after the rise of mathematical linguistics, much effort was devoted to improve its tools for it to better correspond to facts of natural languages.
At the beginning, this research stemmed from the basic ideas of Chomsky and was very close to them. However, it soon became evident that the direct application of simple context-free grammars to the description of natural languages encounters great difficulties. They were mainly English-oriented and explained how to construct an interrogative or negative sentence from the corresponding affirmative one, how to transform the sentence in active voice to its passive voice equivalent, etc. For example, an interrogative sentence such as Does John see Mary?
This is shown in the following figure:. Not nested:. A transformational grammar is a set of rules for such insertions, permutations, movements, and corresponding grammatical changes. Such a set of transformational rules functions like a program. It takes as its input a string constructed according to some context-free grammar and produces a transformed string.
The application of transformational grammars to various phenomena of natural languages proved to be rather complicated. The theory has lost its mathematical elegance, though it did not acquire much of additional explanatory capacity. After the introduction of the Chomskian transformations, many conceptions of language well known in general linguistics still stayed unclear. Nearly all of them were based on the cfg s, but used different methods for description of some linguistic facts. One very important linguistic idea had been suggested already by Chomsky and adopted by the newer theories. It is the subcategorization of verbs according to their ability to accept specific sets of complements.
The term valency is also used in chemistry, and this is not by accident. The point is that each specific verb has its own standard set of actants usually nouns. Within a sentence, the actants are usually located close to the predicative verb and are related with it semantically. For example, the Spanish verb dar has three actants reflecting 1 donator who gives? In texts, the valencies are given in specific ways depending on the verb, e. All three actants of the verb dar can be seen in the sentence Juan 1 dio muchas flores 2 a Elena 3.
The last two actants and their position in a sentence after the verb dar can be expressed with the pattern:. As to the generative grammar approach, it operates only with constituents and related grammar categories.
We have induced only one of the possible subcategorization frames for the verb dar. Direct and indirect complements swap over, while their semantic roles stay the same, bit it is not clear in such a subcategorization frame.
University of the Saarland, Saarbrücken
Categorization and subcategorization are kinds of classification. Any consistent classification implies separation of entities to several non-intersecting groups. However, in the case under our study, the verb dar should be considered belonging to two different subcategories. Or else two verbs dar should be introduced, with equal meanings and equal semantic valency sets, but with different subcategorization frames.
In fact, the situation in languages with the free word order is even more complicated. Indeed, the verb dar can have their donation and receptor actants staying before the subject, like in the sentence A Elena 3 le dio Juan 1 muchas flores 2. Such an order requires even more subcategorization frames obligatorily including the subject of the sentence, i. The actants are used with verbs more frequently than the so-called circonstants. However, the way they are expressed in text does not usually depend on a specific verb.
Only one, obligatory and usually the most important, participant of the situation is expressed by many languages in a quite standard form, namely the subject of the sentence. In Spanish, English, and several other languages but not in all of them! Since the subject is expressed in the same manner with all verbs, it is not specified explicitly in the subcategorization frames.
However, it is efficient only for languages with strict word order. It can only distinguish which noun phrase within a sentence is subject, or direct complement, or indirect complement, etc. Each verb has its own set of semantic cases, and the whole set of verbs in any language supposedly has a finite and rather limited inventory of all possible semantic cases.
Just among them, we can see semantic cases of donator, donation, and receptor sufficient for interpretation of the verb dar. To filter out such incorrect combinations, this rule can be specified in a form similar to an equation:. The gender Gen and the number Num of a determinant should be the same as i. Such a notation is shorthand of the following more complete and more frequently used notation:. The following variant is also used, where the same value is not repeated, but is instead assigned a number and then referred to by this number.
Each feature of a constituent can be expressed with a pair: its name and then its value, e. The rule above determines which values are to be equal. With each constituent, any number of features can be expressed, while different constituents within the same rule possibly can have different sets of specified features.
The featured notation permits to diminish the total number of rules. A generalized rule covers several simple rules at once. Such approach paves way to the method of unification to be exposed later. In addition to the advanced traits of the gpsg , it has introduced and intensively used the notion of head. In most of the constituents, one of the sub-constituents called daughters in hpsg is considered as the principal, or its head called also head daughter. For example, in the rule:. Another example: in the rule:. In the previous examples, the features of the predicate determine features of the whole sentence, and the features of the noun determine the corresponding features of the whole noun phrase.
Such formalism permits to easier specify the syntactic structure of sentences and thus facilitates syntactic analysis parsing. As it was already said, the interpretation in early generative grammars was always of syntactic nature. By contrast, each word in the hpsg dictionary is supplied with semantic information that permits to combine meanings of separate words into a joint coherent semantic structure. The novel rules of the word combining gave a more adequate method of construing the semantic networks. Meantime, Chomskian idea of transformations was definitely abandoned by this approach.
Having in essence the same initial idea of phrase structures and their context-free combining, the hpsg and several other new approaches within Chomskian mainstream select the general and very powerful mathematical conception of unification. The purpose of unification is to make easier the syntactic analysis of natural languages. The unification algorithms are not linguistic proper.
Rather they detect similarities between parts of mathematical structures strings, trees, graphs, logical formulas labeled with feature sets. A priori, it is known that some features are interrelated, i. Thus, some feature combinations are considered compatible while met in analysis, whereas the rest are not.
Two sets of features can be unified, if they are compatible. Then the information at an object admitting unification i. Unification allows filtering out inappropriate feature options, while the unified feature combination characterizes the syntactic structure under analysis more precisely, leading to the true interpretation of the sentence. As the first example of unification operations, let us compare feature sets of two Spanish words, el and muchacho, staying in a text side by side. Hence, the condition of unification is satisfied, and this pair of words can form a unifying constituent in syntactic analysis.
Another example is the adjacent Spanish words las estudiantes. Hence, this pair can form a unifying mother constituent las estudiantes, which inherits the feature set from the head daughter estudiantes. The gender of the particular word occurrence estudiantes is feminine, i. Therefore this particular word occurrence of quisiera is of the third person.
The whole sentence inherits this value, since the verb is its head daughter. The European linguists went their own way, sometimes pointing out some oversimplifications and inadequacies of the early Chomskian linguistics. For more than 30 years, this theory has been developed by the scientific teams headed by I. Apresian in Russia, as well as by other researchers in various countries. One very important feature of the mtt is considering the language as multistage, or multilevel, transformer of meaning to text and vice versa. The transformations are comprehended in a different way from the theory by Chomsky.
Some inner representation corresponds to each level, and each representation is equivalent to representations of other levels. Namely, surface morphologic , deep morphologic, surface syntactic, deep syntactic, and semantic levels, as well as the corresponding representations, were introduced into the model. The government patterns were introduced not only for verbs, but also for other parts of speech. For a verb, gp has the shape of a table of all its possible valency representations.
The table is preceded by the formula of the semantic interpretation of the situation reflected by the verb with all its valencies. The table is succeeded by information of word order of the verb and its actants.