Showing papers on "Semantic similarity published in 1992"

A theory of focus interpretation

Abstract: According to the alternative semantics for focus, the semantic reflec of intonational focus is a second semantic value, which in the case of a sentence is a set of propositions. We examine a range of semantic and pragmatic applications of the theory, and extract a unitary principle specifying how the focus semantic value interacts with semantic and pragmatic processes. A strong version of the theory has the effect of making lexical or construction-specific stipulation of a focus-related effect in association-with-focus constructions impossible. Furthermore, while focus has a uniform import, the sources of meaning differences in association with focus are various.

Frames, Fields, and Contrasts : New Essays in Semantic and Lexical Organization

Abstract: Contents: E.F. Kittay, A. Lehrer, Introduction. Part I:Principles of Organization. L.W. Barsalou, Frames, Concepts, and Conceptual Fields. C.J. Fillmore, B.T. Atkins, Toward a Frame-Based Lexicon: The Semantics of RISK and Its Neighbors. R.E. Grandy, Semantic Fields, Prototypes, and the Lexicon. A. Lehrer, Names and Naming: Why We Need Fields and Frames. J. Ross, Semantic Contagion. E.V. Clark, Conventionality and Contrast: Pragmatic Principles with Lexical Consequences. Part II:Concepts and Relations. R. Jackendoff, What Is a Concept? A. Wierzbicka, Semantic Primitives and Semantic Fields. E.F. Kittay, Semantic Fields and the Individuation of Content. R. Chaffin, The Concept of a Semantic Relation. D.A. Cruse, Antonymy Revisited: Some Thoughts on the Relationship Between Words and Concepts. Part III:Specific Analyses. P. Kay, At Least. M.J. Powell, Folk Theories of Meaning and Principles of Conventionality: Encoding Literal Attitude via Stance Adverb. K. Allan, "Something that Rhymes with Rich." Part IV:Computational Processes in the Lexicon. M.F. Garrett, Lexical Retrieval Processes: Semantic Field Effects. Y. Ravin, Synonymy from a Computational Point of View. D.E. Walker, Developing Computational Lexical Resources.

Global shape cannot be attended without object identification.

Abstract: Whether the global shape of objects can be processed without accessing semantic or identity information was tested. Ss judged which of 2 fragmented forms had the same global shape as a reference stimulus. Matching stimuli could be physically identical, semantically related, or unrelated. The reference stimulus and nonmatching (distractor) form could be semantically related or unrelated. Similarity effects in the related condition were unconfounded with matches nameable and nonnameable forms. For nameable forms, related matching forms facilitated performance; a related distractor disrupted performance. Semantic interference was eliminated when nameable distractors were replaced with nonnameable partners; semantic similarity effects on matching were eliminated with a nonnameable reference stimulus and with inverted targets and distractors. Access to information concerning global shape does not normally occur without object identification.

Automatic Learning for Semantic Collocation

Abstract: The real difficulty in development of practical NLP systems comes from the fact that we do not have effective means for gathering "knowledge". In this paper, we propose an algorithm which acquires automatically knowledge of semantic collocations among "words" from sample corpora.The algorithm proposed in this paper tries to discover semantic collocations which will be useful for disambiguating structurally ambiguous sentences, by a statistical approach. The algorithm requires a corpus and minimum linguistic knowledge (parts-of-speech of words, simple inflection rules, and a small number of general syntactic rules).We conducted two experiments of applying the algorithm to diferent corpora to extract different types of semantic collocations. Though there are some unsolved problems, the results showed the effectiveness of the proposed algorithm.

Structural patterns vs. string patterns for extracting semantic information from dictionaries

Abstract: This chapter presents evidence for preferring to extract semantic information from a syntactic analysis of a dictionary definition rather than directly from the definition string itself when the information to be extracted is found in the differentiae. We present examples of how very complex information can be extracted from the differentiae of the definition using structural analysis patterns, and why string patterns would fail to do the same.

Towards a probabilistic modal logic for semantic-based information retrieval

Abstract: Semantic-based approaches to Information Retrieval make a query evaluation similar to an inference process based on semantic relations. Semantic-based approaches find out hidden semantic relationships between a document and a query, but quantitative estimation of the correspondence between them is often empiric. On the other hand, probabilistic approaches usually consider only statistical relationships between terms. It is expected that improvement may be brought by integrating these two approaches. This paper demonstrates, using some particular probabilistic models which are strongly related to modal logic, that such an integration is feasible and natural. A new model is developed on the basis of an extended modal logic. It has the advantages of : (1) augmenting a semantic-based approach with a probabilistic measurement, and (2) augmenting a probabilistic approach with finer semantic relations than just statistical ones. It is shown that this model verifies most of the conditions for an absolute probability function.

Gpsm: a generalized probabilistic semantic model for ambiguity resolution

Abstract: In natural language processing, ambiguity resolution is a central issue, and can be regarded as a preference assignment problem. In this paper, a Generalized Probabilistic Semantic Model (GPSM) is proposed for preference computation. An effective semantic tagging procedure is proposed for tagging semantic features. A semantic score function is derived based on a score function, which integrates lexical, syntactic and semantic preference under a uniform formulation. The semantic score measure shows substantial improvement in structural disambiguation over a syntax-based approach.

Finding Semantic Similarity in Raw Text: the Deese Antonyms

Abstract: As more and more text becomes readily available in electronic form, much interest is being generated by finding ways of automatically extracting information from subsets of this text. While manual indexing and automatic keyword indexing are well known, both have drawbacks. Recent research on robust syntactic analysis and statistical correlations promises that some of the intuitive advantages of manual indexing can be retained in a fully automatic system. Here I present an experiment performed with my system SEXTANT which extracts semantically similar words from raw text. Using statistical methods combined with robust syntactic analysis, SEXTANT was able to find many of the intuitive pairings between semantically similar words studied by Deese [Deese, 1954].

The picture superiority effect in categorization: visual or semantic?

Abstract: Two experiments are reported whose aim was to replicate and generalize the results presented by Snodgrass and McCullough (1986) on the effect of visual similarity in the categorization process. For pictures, Snodgrass and McCullough's results were replicated because Ss took longer to discriminate elements from 2 categories when they were visually similar than when they were visually dissimilar. However, unlike Snodgrass and McCullough, an analogous increase was also observed for word stimuli. The pattern of results obtained here can be explained most parsimoniously with reference to the effect of semantic similarity, or semantic and visual relatedness, rather than to visual similarity alone. Language: en

Semantic distance in conceptual graphs

Abstract: Given two concepts C1 and C2 with types T1 and T2, Garner and Tsui (1987) have proposed a modification of Sowa’s semantic distance between C1 and C2 as follows. Find the concept C3 which generalizes C1 and C2 with type T3 such that T3 is the most specific type which subsumes T1 and T2; the semantic distance between C1 and C2 is the sum of the distances from C1 to C3 and C2 to C3. It should be clear that this is indeed a metric, satisfying reflexivity, symmetry and the triangle inequality.

Incremental Semantic Analysis

Abstract: Semantic analysis is a central part of the compilation process. The main subproblems include name analysis, type checking, and detection of static-semantic errors. In an interactive programming environment it is useful to perform the semantic analysis incrementally, keeping the staticsemantic information up to date while the program is edited. This allows advanced browsing and editing facilities to be implemented, based on the semantic information. Furthermore, incremental semantic analysis is a prerequisite for making also the rest of the compilation process incremental in order to reduce the turnaround time between editing and execution. This work is directed towards incremental semantic analysis for object-oriented programming languages. These languages have comparatively complex static-semantics which could not be adequately handled with earlier techniques such as attribute grammars. The main contribution of this work is a new technique for developing incremental semantic analyzers: Door Attribute Grammars . This technique extends standard attribute grammars by allowing objects and references to be specified as part of the attribution of a syntax tree. This extension results in space-efficient attributions for which incremental updates can be performed efficiently. In particular, the complex naming semantics of object-oriented languages can be handled in a straight forward way by attributing the tree with explicit visibility graphs built using objects and references. The price for using objects and references in an attribution is that non-local attribute dependencies are introduced which prevent incremental attribute evaluators to be generated completely automatically from the grammar. We solve this problem by splitting the grammar in two parts: one part (the main grammar) which can be treated by automatic methods, and another part (the door package) for which a manual, but systematic, implementation technique is developed. A door package can implement general aspects of a family of programming languages. To specify a new language in the supported family it suffices to write a main grammar, using the door package as a tool box. The techniques have been developed and tested in practice. A complete incrementally compiling environment has been built: Mjolner/Orm , which currently supports the major part of Simula.

Semantic Similarity Relations in Schema Integration

Abstract: To find similarities between objects of different schemata at the semantic level is a crucial problem in schema integration. To identify such similarities it is necessary to form a set of semantic characteristics the objects may have. In this paper, we present a set of such characteristics and a set of semantic similarity relations. The relations are classified into four groups, weak semantic relation, compatible semantic relation, equivalence semantic relation, and mergeable semantic relation. We also propose a schema integration tool, which makes use of the semantic relations to integrate objects of different schemata.

Logical structures in the lexicon

Abstract: The lexical entry for a word must contain all the information needed to construct a semantic representation for sentences that contain the word. Because of that requirement, the formats for lexical representations must be as detailed as the semantic forms. Simple representations, such as features and frames, are adequate for resolving many syntactic ambiguities. However, since those notations cannot represent all of logic, they are incapable of supporting all the function needed for semantics. Richer semantic-based approaches have been developed in both the model-theoretic tradition and the more computational AI tradition. Although superficially in conflict, these two traditions have a great deal in common at a deeper level. Both of them have developed semantic structures that are capable of representing a wide range of linguistic phenomena. The paper compares these approaches, and evaluates their adequacy for various kinds of semantic information that must be stored in the lexicon. It presents conceptual graphs as a synthesis of the logicist and AI representations designed to support the requirements of both.

Learning Useful Horn Approximations.

Abstract: While the task of answering queries from an arbitrary propositional theory is intractable in general, it can typically be performed eeciently if the theory is Horn. This suggests that it may be more eecient to answer queries using a \Horn approximation"; i.e., a horn theory that is semantically similar to the original theory. The utility of any such approximation depends on how often it produces answers to the queries that the system actually encounters; we therefore seek an approximation whose expected \coverage" is maximal. Unfortunately, there are several obstacles to achieving this goal in practice: (i) The optimal approximation depends on the query distribution, which is typically not known a priori; (ii) identifying the optimal approximation is intractable, even given the query distribution; and (iii) the optimal approximation might be too large to guarantee tractable inference. This paper presents an approach that overcomes (or side-steps) each of these obstacles. We deene a learning process, AdComp, that uses observed queries to estimate the query distribution \online", and then uses these estimates to hill-climb, eeciently, in the space of size-bounded Horn approximations , until reaching one that is, with provably high probability, eeectively at a local optimum.

Lack of semantic activation from unattended text during passage reading

Abstract: Two-line passages were read while eye movements were measured. The first line contained a primary target; directly below the primary target, the second line contained the secondary target. The semantic relatedness of primary and secondary targets was varied. The results showed no reliable effects of semantic relatedness when viewing time on the primary target was examined but some effects of relatedness when viewing time on the secondary target was examined. The results suggest that readers do not obtain semantic information from unattended lines of text during reading.

Effects of semantic similarity, omission probability and number of alternatives in computer menu search

Abstract: An experiment was conducted to assess the influence of semantic relatedness, omission probability and number of alternatives on search strategy and response accuracy in computer menu selection. Search strategies were defined as either self-terminating, exhaustive, or redundant and a direct measure of search type was provided in a condition employing sequential presentation of menu alternatives. A simultaneous condition was included to test the generality of results obtained with sequential presentation. Regression analyses indicated that semantic relatedness, omission probability and number of alternatives were all significant predictors of search strategy and response accuracy. Mode of presentation, sequential or simultaneous, was not significant in any of the analyses.

Word Sense Disambiguation With Sublexical Representations

Abstract: This paper introduces a new representational scheme for word sense disambiguation. Drawing on work in information retrieval (Latent Semantic Indexing as proposed by [Deerwester et al. 1990]) an efficient method for learning sublezical representations is described: Words and contexts are represented as vectors in a multidimensional space which approximates similarity of collocational patterns. Closeness of words in the space is equivalent to occurrence in similar contexts, thus giving a rough approximation of semantic similarity. The Bayesian classification system AutoClass was then used to perform an unsupervised classification of sublexical representations of contexts of the three ambiguous words interest, suit and plant in a training text. In applying this classification to a test text, AutoClass disamhiguated 90% of all occurrences correctly. Unsupervised classification failed for tank, but a more sophisticated algorithm also achieved a disambiguation rate of 90%.

An Exploration into Semantic Distance

Abstract: The notion of semantic distance or similarity measurement is closely related to the process of categorization. Any theory of measurement must take into account some real-world results from cognitive science: it is not possible to make a priori judgments about how similar are two random concepts.

Query optimization by using knowledge about data semantics

Abstract: In this paper we address the problem of using semantic properties of data within the process of query optimization. The discussion is in terms of the relational data model. We present query transformation rules, called semantic transformation rules (or simply semantic rules), that are based on the database properties expressed by the integrity constraints. The semantic rules presented in the paper allow join elimination, clustering index introduction and empty query test according to the content of the integrity constraints. We provide a formal proof of the correction of such transformation rules. We also investigate the problem of using semantic rules within transactions, where any arbitrary sequence of queries and modification operations may occur, and semantic integrity can be violated during intermediate steps of processing. Conditions are provided under which the semantic rules presented in the paper can be correctly applied to transform queries occurring within complied transactions.

Similarity in analogical reasoning

Abstract: A computational model of similarity assessment in the context of analogical reasoning is proposed. Three types of similarity are defined: associative, semantic and structural and their specific role in the process of analogical reasoning is discussed. Mechanisms for similarity computation are proposed on the basis of a hybrid cognitive architecture (DUAL). The interaction between the three types of similarity is discussed. Finally, a number of experimental facts is explained in terms of the model. In particular, the dynamic and context-dependent aspects of similarity as well as why it is not a transitive and symmetric relation are discussed.

Semantic information connected with data

Abstract: Informational capability of an attribute set is defined as the set of generated propositions. General data structures (comprising the set construct, tuple construct, functional as well as relational view) are compared for their informational capability using the definability relation. Relations of distinguishing capability (

Some Concepts and Mathematical Consideration of Similarity System Theory

Abstract: This paper introduces a new discipline called Similarity System Theory. Some new concepts, such as Similar Elements, Similarity Unit, Similar Systems and Similarity Entropy are presented. The numerical method and dynamic analysis of similarity system theory are studied.

Document retrieving device

Abstract: PURPOSE:To realize the method for executing a retrieval by contriving semantic similarity of an input keyword and a registration keyword by calculating automatically the degree of relation between each registration keyword, based on semantic information given to the registration keyword CONSTITUTION:A keyword to which semantic information is given is accumulated in a semantic dictionary part 5 When generation of a semantic dictionary is finished, matching of each set consisting of the whole explanatory word given to one keyword is executed by a registration assisting part 1, and a result of matching is shot by a real number value The degree of relation between the keywords shown by this numerical value is accumulated in a keyword network part 6, and when an unregistered retrieval keyword is inputted, a keyword related to the retrieval key-word is generated, and a retrieval is executed, based on the generated keyword According to this constitution, the retrieval can be executed by contriving semantic similarity of the keyword inputted at the time of retrieval and the registration keyword

The QA system

Abstract: In the GA system, semantic information is combined with keywords to measure similarity between natural language queries and documents. A combination of keyword relevance and semantic relevance is achieved by treating keyword weights and semantic weights alike using the vector processing model as a basis for the combination. The approach is based on (1) the database concept of semantic modeling and (2) the linguistic concept of thematic roles. Semantic information is stored in a lexicon built manually using information found in Roget's thesaurus

Syntactic/semantic coupling in the BBN DELPHI system

Abstract: We have recently made significant changes to the BBN DELPHI syntactic and semantic analysis component. These goal of these changes was to maintain the tight coupling between syntax and semantics characteristic of earlier versions of DELPHI, while making it possible for the system to provide useful semantic interpretations of input for which complete syntactic analysis is impossible. Semantic interpretation is viewed as a process operating on a sequence of messages characterizing local grammatical relations among phrases, rather than as a recursive tree walk over a globally complete and coherent parse tree. The combination of incremental semantic interpretation and statistical control of the parsing process makes it feasible to reconstruct local grammatical relations with substantial accuracy, even when a global parse cannot be obtained. Grammatical relations provide the interface between syntactic processing and semantic interpretation, and standard global parsing is viewed as merely one way to obtain evidence for the existence of such grammatical relations in an input string. This focus on grammatical relations leads to substantial simplification of both grammar and semantic rules, and will facilitate our ultimate aim of acquiring syntactic, semantic and lexical knowledge by largely automatic means.