Joshi, Aravind K
Email Address
ORCID
Disciplines
relationships.isProjectOf
relationships.isOrgUnitOf
Position
Introduction
Research Interests
Collection
21 results
Search Results
Now showing 1 - 10 of 21
Publication Using Entity Features to Classify Implicit Discourse Relations(2010-09-01) Joshi, Aravind K; Louis, Annie; Nenkova, Ani; Prasad, RashmiWe report results on predicting the sense of implicit discourse relations between adjacent sentences in text. Our investigation concentrates on the association between discourse relations and properties of the referring expressions that appear in the related sentences. The properties of interest include coreference information, grammatical role, information status and syntactic form of referring expressions. Predicting the sense of implicit discourse relations based on these features is considerably better than a random baseline and several of the most discriminative features conform with linguistic intuitions. However, these features do not perform as well as lexical features traditionally used for sense prediction.Publication Feature Structures Based Tree Adjoining Grammars(1988-10-01) Joshi, Aravind K; Shanker, K. VijayWe have embedded Tree Adjoining Grammars (TAG) in a feature structure based unification system. The resulting system, Feature Structure based Tree Adjoining Grammars (FTAG), captures the principle of factoring dependencies and recursion, fundamental to TAG's. We show that FTAG has an enhanced descriptive capacity compared to TAG formalism. We consider some restricted versions of this system and some possible linguistic stipulations that can be made. We briefly describe a calculus to represent the structures used by this system, extending on the work of Rounds, and Kasper [Rounds et al. 1986, Kasper et al. 1986)involving the logical formulation of feature structures.Publication Living Up to Expectations: Computing Expert Responses(2007-01-01) Joshi, Aravind K; Webber, Bonnie L; Weischedel, RalphIn cooperative man-machine interaction, it is necessary but not sufficient for a system to respond truthfully and informatively to a user's question. In particular, if the system has reason to believe that its planned response might mislead the user, then it must block that conclusion by modifying its response. This paper focuses on identifying and avoiding potentially misleading responses by acknowledging types of 'informing behavior' usually expected of an expert. We attempt to give a formal account of several types of assertions that should be included in response to questions concerning the achievement of some goal (in addition to the simple answer), lest the questioner otherwise be misled.Publication Parsing Strategies With 'Lexicalized' Grammars: Application to Tree Adjoining Grammars(1988-08-01) Schabes, Yves; Abeillé, Anne; Joshi, Aravind KIn this paper, we present a parsing strategy that arose from the development of an Earley-type parsing algorithm for TAGs (Schabes and Joshi 1988) and from some recent linguistic work in TAGs (Abeillé: 1988a). In our approach, each elementary structure is systematically associated with a lexical head. These structures specify extended domains of locality (as compared to a context-free grammar) over which constraints can be stated. These constraints either hold within the elementary structure itself or specify what other structures can be composed with a given elementary structure. The 'grammar' consists of a lexicon where each lexical item is associated with a finite number of structures for which that item is the head. There are no separate grammar rules. There are, of course, 'rules' which tell us how these structures are composed. A grammar of this form will be said to be 'lexicalized'. We show that in general context-free grammars cannot be 'lexicalized'. We then show how a 'lexicalized' grammar naturally follows from the extended domain of locality of TAGs and examine briefly some of the linguistic implications of our approach. A general parsing strategy for 'lexicalized' grammars is discussed. In the first stage, the parser selects a set of elementary structures associated with the lexical items in the input sentence, and in the second stage the sentence is parsed with respect to this set. The strategy is independent of nature of the elementary structures in the underlying grammar. However, we focus our attention on TAGs. Since the set of trees selected at the end of the first stage is not infinite, the parser can use in principle any search strategy. Thus, in particular, a top-down strategy can be used since problems due to recursive structures are eliminated. We then explain how the Earley-type parser for TAGs can be modified to take advantage of this approach.Publication Processing Crossed and Nested Dependencies: An Automaton Perspective on the Psycholinguistic Results(1988-09-01) Joshi, Aravind KThe clause-final verbal clusters in Dutch and German (and in general, in West Germanic languages) has been extensively studied in different syntactic theories. Standard Dutch prefers crossed dependencies (between verbs and their arguments) while Standard German prefers nested dependencies. Recently Bach, Brown, and Marslen-Wilson (1986) have investigated the consequences of these differences between Dutch and German for the processing complexity of sentences, containing either crossed or nested dependencies. Stated very simply, their results show that Dutch is 'easier' than German, thus showing that the push-down automaton (PDA) cannot be the universal basis for the human parsing mechanism. They provide an explanation for the inadequacy of PDA in terms of the kinds of partial interpretations the dependencies allow the listener to construct. Motivated by their results and their discussion of these results we introduce a principle of partial interpretation (PPI) and present an automaton, embedded push-down automaton (EPDA) which permits processing of crossed and nested dependencies consistent with PPI. We show that there are appropriate complexity measures (motivated by the discussion in Bach, Brown, and Marslen-Wilson (1986) according to which the processing of crossed dependencies is easier than the processing of nested dependencies. This EPDA characterization of the processing of crossed and nested dependencies is significant because EPDAs are known to be exactly equivalent to Tree Adjoining Grammars (TAG), which are also capable of providing a linguistically motivated analysis for the crossed dependencies of Dutch (Kroch and Santorini 1988). This significance is further enhanced by the fact that two other grammatical formalisms, (Head Grammars (Pollard, 1984) and Combinatory Grammars (Steedman, 1987), also capable of providing analysis for crossed dependencies of Dutch, have been recently shown to be equivalent to TAGS in their generative power. We have also briefly discussed some issues concerning the degree to which grammars directly encode the processing by automata, in accordance with PPI.Publication Relative compositionality of multi-word expressions: a study of verb-noun (V-N) collocations(2005-10-11) Venkatapathy, Sriram; Joshi, Aravind KRecognition of Multi-word Expressions (MWEs) and their relative compositionality are crucial to Natural Language Processing. Various statistical techniques have been proposed to recognize MWEs. In this paper, we integrate all the existing statistical features and investigate a range of classifiers for their suitability for recognizing the non-compositional Verb-Noun (V-N) collocations. In the task of ranking the V-N collocations based on their relative compositionality, we show that the correlation between the ranks computed by the classifier and human ranking is significantly better than the correlation between ranking of individual features and human ranking. We also show that the properties ‘Distributed frequency of object’ (as defined in [27] ) and ‘Nearest Mutual Information’ (as adapted from [18]) contribute greatly to the recognition of the non-compositional MWEs of the V-N type and to the ranking of the V-N collocations based on their relative compositionality.Publication Sense Annotation in the Penn Discourse Treebank(2008-02-10) Miltsakaki, Eleni; Lee, Alan; Joshi, Aravind K; Robaldo, LivioAn important aspect of discourse understanding and generation involves the recognition and processing of discourse relations. These are conveyed by discourse connectives, i.e., lexical items like because and as a result or implicit connectives expressing an inferred discourse relation. The Penn Discourse TreeBank (PDTB) provides annotations of the argument structure, attribution and semantics of discourse connectives. In this paper, we provide the rationale of the tagset, detailed descriptions of the senses with corpus examples, simple semantic definitions of each type of sense tags as well as informal descriptions of the inferences allowed at each level.Publication Characterizing Structural Descriptions Produced by Various Grammatical Formalisms(1988-09-01) Vijay-Shanker, K.; Weir, David; Joshi, Aravind KWe consider the structural descriptions produced by various grammatical formalisms in terms of the complexity of the paths and the relationship between paths in the sets of structural descriptions that each system can generate. In considering the relationships between formalisms, we show that it is useful to abstract away from the details of the formalism, and examine the nature of their derivation process as reflected by properties of their derivation trees. We find that several of the formalisms considered can be seen as being closely re1aled since they have derivation tree sets with the same structure as those produced by Context-Free Grammars. On the basis of this observation, we describe a class of formalisms which we call Linear Context Free Rewriting Systems, and show they are recognizable in polynomial time and generate only semilinear languages.Publication Flexible Margin Selection for Reranking with Full Pairwise Samples(2004-03-22) Shen, Libin; Joshi, Aravind KPerceptron like large margin algorithms are introduced for the experiments with various margin selections. Compared to the previous perceptron reranking algorithms, the new algorithms use full pairwise samples and allow us to search for margins in a larger space. Our experimental results on the data set of [1] show that a perceptron like ordinal regression algorithm with uneven margins can achieve Recall/Precision of 89.5/90.0 on section 23 of Penn Treebank. Our result on margin selection can be employed in other large margin machine learning algorithms as well as in other NLP tasks.Publication Unification-Based Tree Adjoining Grammars(1991-03-01) Vijay-Shanker, K.; Joshi, Aravind KMany current grammar formalisms used in computational linguistics take a unification-based approach that use structures (called feature structures) containing sets of feature-value pairs. In this paper, we describe a unification-based approach to Tree Adjoining Grammars (TAG). The resulting formalism (UTAG) retains the principle of factoring dependencies and recursion that is fundamental to TAGs. We also extend the definition of UTAG to include the lexicalized approach to TAGs (see [Schabes et al., 1988]). We give some linguistic examples using UTAG and informally discuss the descriptive capacity of UTAG, comparing it with other unificationbased formalisms. Finally, based on the linguistic theory underlying TAGs, we propose some stipulations that can be placed on UTAG grammars. In particular, we stipulate that the feature structures associated with the nodes in an elementary tree are bounded ( there is an analogous stipulation in GPSG). Grammars that satisfy these stipulations are equivalent to TAG. Thus, even with these stipulations, UTAGs have more power than CFG-based unification grammars with the same stipulations.