DOCSLIB.ORG

INTERNATIONAL COMPUTER SCIENCE INSTITUTE Image Schemas and Force-Dynamics in Framenet

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
INTERNATIONAL COMPUTER SCIENCE INSTITUTE Image Schemas and Force-Dynamics in Framenet INTERNATIONAL COMPUTER SCIENCE INSTITUTE 1947 Center St. • Suite 600 • Berkeley, California 94704-1198 • (510) 666-2900 • FAX (510) 666-2956 Image Schemas and Force-Dynamics in FrameNet∗ Klinton Bicknell† and Ellen Dodge‡ TR-04-00X December 2004 Abstract This report presents the early results of the Frame+Schema project, a project to repre- sent image and force-dynamic schemas in FrameNet. These structures are eventually intended to be used in leveraging the wealth of frame semantic knowledge available in the FrameNet database for situation-specific reasoning. We detail our image and force- dynamic schemas in the Embodied Construction Grammar formalism and give corre- spondence rules to represent these schemas in FrameNet. We also analyze much of the current force-dynamic structure in the FrameNet database and suggest changes to make it more compatible with image and force-dynamic schema-based inference. We note many of the issues raised by our representation as well as the unsolved problems and also make proposals for future work in the Frame+Schema project. ∗This would be far worse if not for the many suggestions of the others at the Frame+Schema meetings: Jerry Feld- man, Srini Narayanan, Michael Ellsworth, Collin Baker, and Josef Ruppenhofer. Johno Bryant was also more than helpful and generous with his time. †[email protected] ‡[email protected] ii 0 Introduction (3) investigating how FrameNet currently rep- resents force-dynamic and image schematic Schemas have been hypothesized to be cen- ideas such as causation, and changing these tral to the human capacity to reason (John- representations to use the newly represented son 1987; Lakoff & Johnson 1999), with schemas. For the first two pieces, we de- image schemas and force-dynamic schemas fine our schemas in the Embodied Construc- (Talmy 1988) playing central roles. The tion Grammar (ECG) formalism and give map- goal of making inferences from language in ping rules to represent the ECG schemas in a human-like way, then, would require at FrameNet. We use the ECG formalism to rep- the least a force-dynamic/image schematic resent our schemas independently of the de- analysis. (Hereafter, image/force-dynamic tails of their FrameNet representation because schemas will be collectively referred to sim- a sizable body of literature exists showing that ply as schemas.) To allow for situation-specific image schemas can be effectively represented inferences, these schemas must interface with in the formalism1 (Bergen & Chang In Press; real-world knowledge about the scenario un- Chang et al. 2002; Feldman 2002b). This litera- der discussion. Frame semantics character- ture provided us with some initial ideas about izes this real-world knowledge in units called how to represent schemas in terms of roles and frames, which are representations of gestalt sce- relations, since there is no complete ontology narios defined by a set of entities and states of of image schemas and force-dynamic schemas affairs which participate in the scenario, and available for use. relations between these frames (Fillmore 1975, This paper treats all three pieces of the task: 1985). The FrameNet project at the Interna- Section 1 presents a short overview of the ba- tional Computer Science Institute has built a sic method of representation, i.e., of mapping large database of such frame semantic knowl- from ECG schemas to FrameNet frames2 and edge and connects it to corpus text by a pro- Section 2 (providing the bulk of the paper) de- cess of annotation (Fillmore et al. 2003; Baker tails our schema ontology in ECG and high- et al. 2003). Because schemas can also be for- lights anything controversial about a schema mally represented as a list of participating ele- definition, anything not straightforward about ments and relations to other schemas, it seems its FrameNet representation, or any relevant natural that the FrameNet project can be ex- relation between that schema and the rest of tended to represent both kinds of structures the FrameNet database. required for reasoning, the frames and the 1In fact, the ECG literature works on the hypothe- schemas, as well as the interactions between sis that image schemas, force-dynamic schemas, action these two structures. As a first step toward us- schemas, and frames can all be represented in a simi- ing the extensive frame semantic knowledge lar way, using roles and relations, calling them all just available in the FrameNet database for rea- “schemas.” It thus works very nicely with our current project of representing schemas as frames. soning, this paper proposes a means of repre- 2This assumes basic familiarity with the ECG formal- senting image/force-dynamic schemas in the ism, which may be gained from Bergen & Chang (In FrameNet database and a method of relating Press) or Chang et al. (2002). these structures to each other and to the frames already present in the database. This basic proposal can be broken into two main pieces: (1) defining the relevant schemas and (2) representing these schemas in the FrameNet database. A third piece is 1 1 The Basic Representation DEFINES DEFINES Strategy PARENT CHILD ONLY FORCED INHERITANCE “POV-Uses” Inheritance Most of the mapping from ECG to FrameNet FREE INHERITANCE Subframing “Chatty-Uses” representation is straightforward. A schema is represented as a class of frame3. Roles in ECG Table 1: The four frame-to-frame relations in map to frame elements in FrameNet. The sub- FrameNet, separated by binary features case of operator corresponds to an Inheritance relation. Type-constraints on roles are trans- lated into semantic types on frame elements if possible, and otherwise are just mentioned in the text of the frame element’s description. ECG identity statements (i.e., unification) cor- “Chatty-Uses.” “POV-Uses” is a quite rare re- respond to pairing frame elements in frame-to- lation, and is the relation between a point-of- frame relations (more on this below). All this view-independent representation of a scenario seems to be quite straightforward, since struc- and the point-of-view-specified versions of it. tures are available in the FrameNet database So, for example, the Commerce buy and Com- basically equivalent to these ECG structures. merce sell frames both have “POV-Uses” rela- tions to Commerce goods-transfer, since buy- ing and selling describe the same event from evokes 1.1 Representing two different perspectives. “Chatty-Uses” is The evokes operator in ECG should clearly the relation between a frame and a background specify some sort of frame-to-frame relation in frame needed to make sense of it. This is, for FrameNet. In order to explain why we chose example, the relation between the Volubility the “Chatty-Using” relation as its correspon- frame (which is referred to by words such as dent, it seems helpful to describe the other op- chatty, loudmouth, and mum) and the Commu- tions. nication frame. The Volubility frame only rep- FrameNet has four types of frame-to-frame resents a valid concept given the Communica- relations: Inheritance, Subframing, “Chatty- tion frame. Using,” and “POV-Using.” Inheritance is ex- These four relations can be thought of in actly what one would expect, a requirement a matrix of binary features (Table 1). “POV- that every element in the parent frame have Uses” and Inheritance both require that ev- a counterpart in the child frame. The Sub- ery element in the parent be mapped to (i.e., frame relation is used for the event struc- unified with) an element of the child, while ture of frames within a larger scene. For in- the Subframe relation and “Chatty-Uses” do stance, the Arresting frame would be a sub- not have such a requirement. “POV-Uses” frame of the larger Criminal process frame, as and the Subframe relation also express rela- would the Arraignment frame and the Trial tions in which the child helps to define the frame. There is a only a single formal rela- parent. For example, the Criminal Process tion called Using in the FrameNet database, al- frame is defined mostly by the definitions of though it is used to represent two distinct re- its constituent parts (such as Arresting or Ar- lations, what I have called “POV-Uses” and raignment). Similarly, the Commerce goods- 3 transfer frame is defined by being what is left The way this class is marked in the FrameNet after the point-of-view is removed from the database has not yet been specified. Commerce buy and Commerce sell frames. 2 For Inheritance and “Chatty-Uses” relations, to ‘metaphoric’ and ‘non-metaphoric’ scenar- this is not the case. Communication is defined ios, while frames are more specified and are just fine without reference to Volubility, and domain-specific. For example, both “I am in obviously, an inheritance relation does not add Austin” and “I am in mourning” would use anything to the superclass. the same In schema, while only the first case Given these two formal features of the four would use the Locative relation frame. relations, the final relation, “Chatty-Uses,” Our solution to this problem is to add an- seems to correspond most closely to the ECG other mediating structure to the hierarchy evokes relation, since it is certainly not defini- called a “dependent frame.” Every dependent tional for the parent, and there is no require- frame inherits from both one schema and one ment that the child have the same elements frame; thus they the very bottom of the hi- as the parent. By representing evokes with erarchy, and all FrameNet annotation should “Chatty-Uses,” we are extending the meaning be done relative to them. For example, one of “Chatty-Uses.” For our purposes, it will dependent frame might inherit from the In often not be a scenario in which a larger con- schema and the LocativeRelation frame; this de- text frame must be present for the structure pendent frame would be used in phrases such to be made sense of (as with the Volubility as “the ball in the box.” Another dependent frame and Communication), but will often be frame might inherit from In schema as well a scenario in which a smaller frame is used as the TakingTime frame; this dependent frame as part of the frame’s structure (as the Jailing would be used in cases such as “Andy finished frame might use the Container schema).
Load more

Recommended publications
  • 4 the Semantics of Verbs
    UvA-DARE (Digital Academic Repository) Event structure, conceptual spaces and the semantics of verbs Warglien, M.; Gärdenfors, P.; Westera, M. DOI 10.1515/tl-2012-0010 Publication date 2012 Document Version Final published version Published in Theoretical Linguistics Link to publication Citation for published version (APA): Warglien, M., Gärdenfors, P., & Westera, M. (2012). Event structure, conceptual spaces and the semantics of verbs. Theoretical Linguistics, 38(3-4), 159-193. https://doi.org/10.1515/tl- 2012-0010 General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons). Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible. UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl) Download date:02 Oct 2021 DOI 10.1515/tl-2012-0010 Theoretical Linguistics 2012; 38(3-4): 159 – 193 Massimo Warglien, Peter Gärdenfors and Matthijs Westera Event structure, conceptual spaces and the semantics of verbs Abstract: The aim of this paper is to integrate spatial cognition with lexical se- mantics.
    [Show full text]
  • The Force of Language: How Children Acquire the Semantic Categories of Force Dynamics
    THE FORCE OF LANGUAGE: HOW CHILDREN ACQUIRE THE SEMANTIC CATEGORIES OF FORCE DYNAMICS A Dissertation Submitted to The Temple University Graduate Board In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY Nathan R. George May 2014 Examining Committee Members: Dr. Kathy Hirsh-Pasek, Advisory Chair, Developmental Psychology Dr. Nora Newcombe, Examining Chair, Developmental Psychology Dr. Peter Marshall, Developmental Psychology Dr. Thomas Shipley, Brain and Cognitive Sciences Dr. Roberta Michnick Golinkoff, University of Delaware Dr. Phillip Wolff, Emory University © Copyright 2014 by Nathan R. George ii ABSTRACT Nathan R. George Doctor of Philosophy Temple University, 2014 Dissertation Advisory Committee Chair: Dr. Kathy Hirsh-Pasek Dissertation Examination Chair: Dr. Nora Newcombe Verbs and prepositions encode relations within events, such as a child running towards the top of a hill or a second child pushing the first away from the top. These relational terms present significant challenges in language acquisition, requiring the mapping of the categorical system of language onto the continuous stream of information in events. This challenge is magnified when considering the complexities of events themselves. Events consist of part-whole relations, or partonomic hierarchies, in which events defined by smaller boundaries, such as the child running up the hill, can be integrated into broader categories, such as the second child preventing the first from reaching the top (Zacks & Tversky, 2001). This dissertation addresses how this partonomic hierarchy in events is paralleled in the structure of relational language. I examine the semantic category of force dynamics, or “how entities interact with respect to force” (Talmy, 1988, p. 49), which introduces broad categories (e.g., help, prevent) that incorporate previously independent relations in events, such as paths, goals, and causality.
    [Show full text]
  • An Analysis of Trump's 2016 Campaign Speeches from The
    ISSN 1799-2591 Theory and Practice in Language Studies, Vol. 9, No. 9, pp. 1118-1124, September 2019 DOI: http://dx.doi.org/10.17507/tpls.0909.07 An Analysis of Trump’s 2016 Campaign Speeches from the Perspective of Social Force in Force-dynamics Theory Yunyou Wang Department of International Studies, Southwest University, Chongqing, China Abstract—Using Donald Trump’s presidential campaign speeches from 2016 as its corpus, this paper takes Talmy’s (1988) force-dynamics theory as a framework for deconstructing and analyzing political discourse from a cognitive perspective. In view of Newton’s second law of classical mechanics, this study analyzes the source and composition of social force within force-dynamics theory and constructs a cognitive model of discourse: F (social force) = m (quantification of discourse information) * a (discourse strategy), applicable to analysis of the relations among language, rights, and ideology in political discourse. In Trump’s campaign speeches, the quantification of discourse information and strategies using cognitive mechanisms such as metaphor, metonymy, and categorization greatly strengthens the social force of the discourse. This plays an important role in conveying Trump’s intention to voters, influencing the public’s emotions and winning their support—thus changing the original psychological state of voters and finally even influencing the election result. Index Terms—presidential campaign speeches, force-dynamics theory, Newton’s second law of classical mechanics, quantification of discourse information, discourse strategy I. INTRODUCTION The study of political discourse originated in Aristotle's Politics and is interdisciplinary, typically a combination of psychology and anthropology. The study of political discourse from the perspective of linguistics dates to the 1980s, and combines the methodology of critical discourse analysis (CDA) focusing on the relations among language, ideology, and rights with systematic functional grammar.
    [Show full text]
  • Guided Generation of Cause and Effect
    Guided Generation of Cause and Effect Zhongyang Li1,2∗, Xiao Ding1, Ting Liu1, J. Edward Hu2 and Benjamin Van Durme2 1Harbin Institute of Technology, China 2Johns Hopkins University, USA fzyli,xding,[email protected], fedward.hu,[email protected] Abstract because they are hungry because they are lonely We present a conditional text generation frame- because they are in pain work that posits sentential expressions of possible cause because they want to be loved causes and effects. This framework depends on because they want to go home … two novel resources we develop in the course of babies cry this work: a very large-scale collection of English will lead to sleep problems sentences expressing causal patterns (CausalBank); can lead to depression and a refinement over previous work on construct- effect can lead to a bad marriage ing large lexical causal knowledge graphs (Cause can lead to bad food habits Effect Graph). Further, we extend prior work in result in tears to the eyes lexically-constrained decoding to support disjunc- … tive positive constraints. Human assessment con- Figure 1: Possible causes and effects generated by our model, con- firms that our approach gives high-quality and di- ditioned on the input sentence “babies cry”. Tokens in blue are verse outputs. Finally, we use CausalBank to per- constraint keywords derived from our Cause Effect Graph, which are form continued training of an encoder supporting forced to be included in the outputs by constrained decoding. a recent state-of-the-art model for causal reason- ing, leading to a 3-point improvement on the COPA tures causal relations as a mapping across lexical types, lemma- challenge set, with no change in model architecture.
    [Show full text]
  • Book of Abstracts
    The Association for Literary and Lingustic Computing The Association for Computers and the Humanities Society for Digital Humanities — Société pour l’étude des médias interactifs Digital Humanities 2008 The 20th Joint International Conference of the Association for Literary and Linguistic Computing, and the Association for Computers and the Humanities and The 1st Joint International Conference of the Association for Literary and Linguistic Computing, the Association for Computers and the Humanities, and the Society for Digital Humanities — Société pour l’étude des médias interactifs University of Oulu, Finland 24 – 29 June, 2008 Conference Abstracts International Programme Committee • Espen Ore, National Library of Norway, Chair • Jean Anderson, University of Glasgow, UK • John Nerbonne, University of Groningen, The Netherlands • Stephen Ramsay, University of Nebraska, USA • Thomas Rommel, International Univ. Bremen, Germany • Susan Schreibman, University of Maryland, USA • Paul Spence, King’s College London, UK • Melissa Terras, University College London, UK • Claire Warwick, University College London, UK, Vice Chair Local organizers • Lisa Lena Opas-Hänninen, English Philology • Riikka Mikkola, English Philology • Mikko Jokelainen, English Philology • Ilkka Juuso, Electrical and Information Engineering • Toni Saranpää, English Philology • Tapio Seppänen, Electrical and Information Engineering • Raili Saarela, Congress Services Edited by • Lisa Lena Opas-Hänninen • Mikko Jokelainen • Ilkka Juuso • Tapio Seppänen ISBN: 978-951-42-8838-8 Published by English Philology University of Oulu Cover design: Ilkka Juuso, University of Oulu © 2008 University of Oulu and the authors. _____________________________________________________________________________Digital Humanities 2008 Introduction On behalf of the local organizers I am delighted to welcome you to the 25th Joint International Conference of the Association for Literary and Linguistic Computing (ALLC) and the Association for Computers and the Humanities (ACH) at the University of Oulu.
    [Show full text]
  • ACCT 2000 Accounting Concepts for Nonbusiness Students Fall, Spring, Summer. Accounting Concepts and Procedures and Their Contri
    Subject Catalog Long Title Descr Min Units Max Units ACCT 2000 Accounting Concepts for Nonbusiness Students Fall, Spring, Summer. Accounting concepts and procedures and their 3.00 3.00 contribution to administrative processes. Enterprise analysis, relevant data, its uses and limitations. Not applicable to preprofessional core requirements in the College of Business Administration. No credit allowed toward BSBA degree. Prerequisite: sophomore standing or consent of department. Approved for distance education. ACCT 2210 Accounting and Business Concepts I Fall, Spring, Summer. Concepts and issues of organizational reporting are 3.00 3.00 introduced within the context of financial and managerial accounting, systems, taxation and auditing, and are illustrated through the use of examples involving international and domestic businesses, non-profit and government organizations. The course is designed to enhance group dynamics, communications skills, use of electronic media and inquiries into ethics and values within the accounting environment. Prerequisite: sophomore standing or consent of the department. Approved for distance education. ACCT 2220 Accounting and Business Concepts II Fall, Spring, Summer. ACCT 2210 continued. Prerequisite: ACCT 2210. 3.00 3.00 Approved for distance education. ACCT 3210 Intermediate Financial Accounting I Fall, Spring. Development and application of financial accounting 3.00 3.00 concepts and generally accepted accounting principles. Preparation of financial statements and accounting for changes in accounting principles. Emphasis on valuation and cost allocation methods for assets and related effects on income statements. Prerequisite: (Pre-BSBA or BSBA) and (C or better in ACCT 2210 and ACCT 2220) and (Pre-Accounting or admission to the accounting or finance or information systems auditing and control specialization) or consent of department.
    [Show full text]
  • Semplates: a New Concept in Lexical Semantics?
    SHORT REPORT Semplates: A new concept in lexical semantics? STEPHEN C. LEVINSON NICLAS BURENHULT Max Planck Institute Max Planck Institute for Psycholinguistics for Psycholinguistics and Lund University This short report draws attention to an interesting kind of configuration in the lexicon that seems to have escaped theoretical or systematic descriptive attention. These configurations, which we dub SEMPLATES, consist of an abstract structure or template, which is recurrently instantiated in a number of lexical sets, typically of different form classes. A number of examples from different language families are adduced, and generalizations made about the nature of semplates, which are contrasted to other, perhaps similar, phenomena.* Keywords: semplate, lexical semantics, space, landscape lexicon, metaphor, analogy, cultural models, sense relations, toponyms 1. INTRODUCTION. The purpose of this short report is to draw attention to an unre- marked type of patterning in the lexicon, and to suggest that these patterns are of sufficient importance to warrant the introduction of a new descriptive concept in lexical 1 semantics, which we dub a SEMPLATE (a blend of semantic template). The observations that motivate the new descriptive concept are of the following kind. Within a language, some semantic or conceptual template—for example, a three- or four-way spatial opposition—surfaces again and again in distinct lexical sets, say prepo- sitions, spatial nouns, verbs of motion, and the like. This template typically involves not just a single parameter or dimension of opposition, but rather a structured set of opposing distinctions. To take a simple example, in Ye´lıˆ Dnye (the Papuan language of Rossel Island), there are three intransitive stative positional verbs of location glossing ‘sit’, ‘stand’, and ‘hang’—all physical objects have conventional collocations with the positionals according (especially) to their shape, canonical position, and rigidity (Levinson 2000).
    [Show full text]
  • On Pragmatic and Cognitive Processes in Meaning Variation
    Linguistica Silesiana 31, 2010 ISSN 0208-4228 NADĚŽDA KUDRNÁČOVÁ Masaryk University ON PRAGMATIC AND COGNITIVE PROCESSES IN MEANING VARIATION This paper first looks into mechanisms that license the formation of transitive caus- ative constructions with the motion verbs run and walk. As a further step, it takes into consideration intransitive constructions with these verbs and, in doing so, it contrasts the meaning of run with walk as its most natural counterpart. The paper provides evidence in favour of positing one of the verb’s senses as core, representing a kind of starting point against which some of the other motion senses are established. In this way, arguments are offered in favour of the lexical network model of polysemy. At the same time, it is shown that the extensive usability of run (and, by the same token, the restricted usability of walk) is closely related to the degree of the verb’s context- -sensitivity, which, in its turn, points to the conception of the verb’s meaning as repre- senting a dynamic potential. 1. Principled Mechanisms in Meaning Variation Deriving from the basic assumption that the various senses of a polysemous word are interrelated in some way (on the evidence verifying the interrelations cf. Apresjan 1973, Panman 1982 and Williams 1992, among others), two notorious questions arise: (a) What are the mechanisms that underlie the variation? (b) If the senses are related in principled ways, can we posit one of the senses as core? Needless to say, the relat- edness of meaning is, by virtue of its nature, a scalar concept and, as such, is open to subjective evaluation (Lyons 1977: 552).
    [Show full text]
  • The Linguistic Representation of Conceptual Structure: Applications for Teaching Reading Comprehension
    The linguistic representation of conceptual structure: applications for teaching reading comprehension Mary Elaine Meagher Sebesta Escuela Nacional Preparatoria No. 6 La Joya 228 Casa 4 Col. Tepepan C. P. 14600, México D.F. [email protected] Resumen Este artículo utiliza los resultados de las investigaciones de Meagher (1999, 2004) para desarrollar estrategias que aumenten la comprensión de lectura en áreas científicas. Es­ tudian la manera en que los esquemas semánticos se relacionan con procesos discursivos. El objetivo general de Meagher (1999), basado en Jackendoff (1983, 1990), es elaborar una metodología para la apreciación de los conceptos que tienen los novatos y los exper­ tos sobre el aprendizaje a través de una aprehensión de esquemas subyacentes. Meagher (2004), desde la perspectiva de la dinámica de fuerzas de Talmy (1985, 2000), examina cómo las distintas nociones semánticas se combinan con el sistema de la dinámica de fuerzas para brindar dirección al discurso. Ambos proyectos se refieren a la construcción del significado a nivel de discurso y estudian la manera en que la información gramatical subléxica estructura el significado en corpora auténticos. Por lo tanto, los resultados terminan validando los postulados teóricos de Jackendoff y Talmy. La comprensión de estas estructuras conceptuales subyacentes es de gran utilidad para hispano­parlantes confrontados con la tarea de leer textos científicos en inglés. Palabras clave: comprensión de lectura, esquemas semánticos, procesos discursivos, dinámica de fuerzas, inglés para propósitos específicos Fecha de recepción del artículo: 19 de septiembre de 2005 Fecha de aceptación de versión revisada: 8 de enero de 2007 ela48ok.indd 197 14/2/11 14:16:36 198 Mary Elaine Meagher Sebesta Abstract This article applies the research results of Meagher (1999, 2004) to the development of strategies for increasing reading comprehension in English for Specific Purposes (E SP ) courses.
    [Show full text]
  • Constructions, Frames and Event Structure
    The AAAI 2017 Spring Symposium on Computational Construction Grammar and Natural Language Understanding Technical Report SS-17-02 Constructions, Frames and Event Structure William Croft and Meagan Vigus Department of Linguistics MSC 03 2130 1 University of New Mexico Albuquerque NM 87131-0001 Abstract Commercial_transaction Charles Fillmore developed both the theory of construction Commerce_goods_transfer Commerce_money_transfer grammar and the theory of frame semantics. Fillmore’s orig- inal frame semantic model included broad frames such as Commercial Transaction and Risk. FrameNet II (the current Commerce_buy Commerce_sell Commerce_collect Commerce_pay incarnation) has a complex lattice of frame-to-frame rela- tions, in which the original frames are often abstract frames, and/or have been decomposed into distinct concrete frames. Figure 1: Commercial transaction frames in FrameNet The frame-to-frame relations are of different types beyond simple taxonomic relations. The concrete frames are often more restricted in their argument structure construction pos- (location of the Goods), as well as non-core FEs such as In- sibilities than the original, more general frames. We argue strument, Manner, Means, Purpose, Explanation, Frequency, that some frame-to-frame relations effectively represent the Place and Time. An example is Leslie stole the watch from internal structure of the core events in the frame; the event Kim. structures are associated with different argument structure However, the nature of frames in FrameNet differs from constructions. Introducing event structure into the frame se- that in Fillmore’s earlier work, and in earlier versions of mantic representation offers the potential to simplify the FrameNet. For instance, in Fillmore (1977b, 59), all of the FrameNet lattice.
    [Show full text]
  • A New Resource for German Causal Language
    Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5968–5977 Marseille, 11–16 May 2020 c European Language Resources Association (ELRA), licensed under CC-BY-NC A New Resource for German Causal Language Ines Rehbein and Josef Ruppenhofer Data and Web Science Group Archive for Spoken German University of Mannheim, Leibniz Institute for the German Language Mannheim, Germany Mannheim, Germany [email protected] [email protected] Abstract We present a new resource for German causal language, with annotations in context for verbs, nouns and adpositions. Our dataset includes 4,390 annotated instances for more than 150 different triggers. The annotation scheme distinguishes three different types of causal events (CONSEQUENCE,MOTIVATION,PURPOSE). We also provide annotations for semantic roles, i.e. of the cause and effect for the causal event as well as the actor and affected party, if present. In the paper, we present inter-annotator agreement scores for our dataset and discuss problems for annotating causal language. Finally, we present experiments where we frame causal annotation as a se- quence labelling problem and report baseline results for the prediciton of causal arguments and for predicting different types of causation. Keywords: Annotation of causal language, causal tagger, German 1. Introduction these pre-defined contructions in text, Dunietz et al. (2015) Understanding causality is crucial for making sense of the obtain high agreement scores for human annotation. In the world around us. Thus, understanding causal expressions paper, we adapt their approach and present a new dataset in text is an important prerequisite for Natural Language for German causal language, with annotations in context Understanding (NLU).
    [Show full text]
  • Force Dynamics Phillip Wolff Robert Thorstad
    Force Dynamics Phillip Wolff Robert Thorstad Emory University Please address correspondence to: Phillip Wolff Emory University Department of Psychology 36 Eagle Row Atlanta, GA 30322 Office #: 404-727-7140 Fax #: 404-727-0372 [email protected] In press in Oxford Handbook of Causal Reasoning (Michael Waldmann, Ed.) Abstract Force dynamics is an approach to knowledge representation that aims to describe how notions of force, resistance, and tendency enter into the representation of certain kinds of words and concepts. As a theory of causation, it specifies how the concept of CAUSE may be grounded in people’s representations of force and spatial relations. In this chapter, we review theories of force dynamics that have recently emerged in the linguistic, psychological, and philosophical literatures. In discussing these theories we show how a force dynamic approach to causation is able to account for many of the key phenomena in causal cognition, including the representation of individual causal events, the encoding of causal relations in language, the encoding of causal chains, and causation by omission. Recent accounts of force dynamics have shown how this perspective can be instantiated in computational terms, which not only helps clarify the force dynamic approach to various phenomena, but also helps explain how the units of cognition hypothesized in this approach may be grounded in people’s perceptual experience. In addition to reviewing the strengths of this approach, the chapter identifies lines of inquiry for possible future research. Keywords: Force Dynamics; Causal understanding; Causal Perception; Language and Causation; Causal Reasoning; Causation by omission; Causal composition; Causal chains; Singular causation Causal events typically involve a large set of factors, many of which may be necessary for the occurrence of an effect (Mill, 1973/1872).
    [Show full text]