Cohesion, Coherence and Temporal Reference from an Experimental Corpus Pragmatics Perspective

Front matter

Yearbook of Corpus Linguistics and Pragmatics

Cristina Grisot

Cohesion, Coherence and Temporal Reference from an Experimental Corpus Pragmatics Perspective

Introduction

Despite the considerable amount of published literature on temporal reference and its linguistic expression, no previous study has dealt with verbal tenses in general and more specifically the categories of Tense, lexical aspect (hereafter, Aktionsart) and grammatical aspect (hereafter, Aspect) as cohesion ties contributing to the temporal coherence of a discourse from an empirical and an experimental perspective. This work aims to provide new methodological and theoretical insights into temporal reference and its linguistic components, from an experimental corpus pragmatics approach. This book, published in the Yearbook of Corpus Linguistics and Pragmatics series, illustrates how the study of linguistic and pragmatic phenomena benefits from the combination of two approaches: on the one hand, the rigorous and meticulous methodology found in the domains of corpus linguistics and psycholinguistics and, on the other hand, the rich theoretical understanding of language and the interpretation of sentence meaning and intended meaning provided by the fields of theoretical linguistics and pragmatics. As such, it investigates the phenomenon of temporal reference at the interface between corpus linguistics, theoretical linguistics and pragmatics, experimental pragmatics, psycholinguistics, natural language processing and machine translation. The line of research adopted in this book shows how theoretical studies bring forth new hypotheses about language meaning and language use, which are tested in both naturally occurring data and carefully designed experiments. This empirical and experimental testing provides evidence which might lead to the revision, if necessary, of the initial theoretical models. This book will give readers insights into how they can develop solid, empirically and experimentally based theoretical models of linguistic phenomena.

Since Halliday and Hasan’s seminal work on cohesion in English (1976), the notions of coherence, cohesion and cohesive ties have been used extensively in reference to a series of phenomena, such as pronominal, demonstrative and comparative reference. In this book, I deal with temporal reference — the localization of eventualities (states and events) in time — in natural language and its role in establishing temporal cohesion and coherence. There are numerous ways in which temporal reference may be expressed, such as the grammatical categories of Tense and Aspect (generally referred to by the generic notion verbal tense), inherent temporal

Vii viii Introduction features of the verb phrase (known as lexical/situation aspect or Aktionsart), temporal adverbials and connectives (such as yesterday and before, special particles such as the Mandarin Chinese aspectual particles -le and -guo, and pragmatic inferential principles of discourse comprehension, among others). In tensed languages, the primary focus of research on temporal reference has been Tense, with Aspect and Aktionsart secondary considerations. Crosslinguistic research from formal semantics over the past 40 years has pointed out that there are languages without the grammatical category of Tense (the so-called tenseless languages, such as Mandarin Chinese and Yucatec, Mayan, Mexico) and mixed-tense languages (with optional tense marking alongside untensed clauses, such as Navajo, Japanese and Korean) as noted by Tonhauser (2015). These studies have argued that Tense does not fully determine the temporal reference of a sentence but merely locates eventualities relationally (Smith 2008, 232).

Despite a long tradition of research on verbal tenses in tensed languages, there is no general consensus among current theories, except on a certain number of basic points, such as the use of temporal coordinates (e.g. Reichenbach 1947; Klein 1994) to describe verbal tenses and the notion of temporal anaphor (such as Partee 1973, 1984; Webber 1988; Kamp and Reyle 1993). The challenge for current research on temporal reference is to propose a crosslinguistically valid model that holds for both tensed and tenseless languages, mainly by decentralizing the role played by Tense and increasing the attention paid to the other constituents of temporal reference. To date, numerous semantic and pragmatic studies have been dedicated to individual verbal tenses in languages such as English, French and Spanish (see e.g. Moeschler et al. 1998 and de Saussure 2003 on French verbal tenses) without completely discriminating between types of temporal information from the categories of Tense, Aspect and Aktionsart. The direct consequence of approaches in which the three categories are not completely distinguished is the lack of a crosslinguistically valid model of temporal reference, which goes beyond language specificities and which allows for consistent contrastive analyses. Additionally, studies in psychology, psycholinguistics and neurolinguistics (such as Carreirras et al. 1997; Todorova et al. 2000; de Vega et al. 2004; Therriault and Raney 2007; Dery and Koenig 2015; Magliano and Schleich 2000; Bastiaanse 2008; Bastiaanse et al. 2011) have demonstrated that the constituents of a verbal tense, namely the categories of Tense, Aspect and Aktionsart are cognitively motivated.

For this reason, this book will investigate the semantics and pragmatics of Tense, Aspect and Aktionsart, which are considered as cohesive ties playing a role in the expression of temporal coherence relations. For Halliday and Hasan, the notion of cohesion is a semantic concept and a property of a text that occurs when the interpretation of an element in the discourse is dependent on the interpretation of a different element. As a semantic concept, cohesion is part of the system of the language and is established by the relation that is set up between two elements. They identify grammatical and lexical cohesion, with each type having its own cohesive ties and methods. Grammatical cohesion involves methods such as substitution, anaphora and ellipsis. Anaphora is exemplified in (1), where the pronoun she has Mary as

Introduction ix antecedent, and done refers back to send a picture of the children. Ellipsis is exemplified in (2), where it is the second verbal group which is elliptical. Grammatical cohesion consists of grammatical cohesive ties such as pronouns, discourse connectives and verbal tenses, whereas lexical cohesion occurs by the repetition of a word, such as in example (3), where the repetition of the word apple has a cohesive effect.

(1) Mary promised to send a picture of the children, but she hasn’t done. (2) Are you dieting? I have been for some time. (3) Wash and core six apples. Put the apples into a fireproof dish.

Hence, successive utterances in a cohesive text refer to the same entities, and their cohesion is provided by the links existing between various elements. However, example (4) is not coherent, even if “he” can refer to “John” (Hobbs 1979, 67).

(4) John took a train from Paris to Istanbul. He likes spinach.

Hobbs argues that there is an expectation of coherence, which is deeper than the notion of a discourse just being “about” some set of entities. Similarly, temporal cohesion can be ensured when two verbal tenses refer to the same point of reference (Reichenbach 1947), as in example (5), or to different points, as in examples (6) and (7). Additionally, these three examples illustrate the types of temporal relations that comprehenders are led to infer when interpreting these successive utterances: temporal simultaneity in (5), where the phone call occurs at the same time as the arrival; temporal precedence in (6), where the phone call occurs before the arrival; and temporal succession in (7), where the phone call occurs after the arrival. In this book, I will refer to the first type of temporal relation as synchronous, and to the second and third as sequential. I will treat these as coherence relations (Hobbs 1979).

(5) Mary was arriving home when her husband called her. (6) Mary arrived home. Her husband had called her. (T) Mary arrived home. Her husband called her.

Coherence relations have been investigated from three points of view: theoretical linguistics, computational linguistics, and psycholinguistics. Theoretical linguistics describes the factors that contribute to discourse coherence and has sought to categorize the different types of coherence relations whose role is to connect clauses and sentences. One of the best-known taxonomies is that proposed by Halliday and Hassan: additive, temporal, causal and adversative (contrast). As Kehler (2004, 244) suggests, all proposals for taxonomies of discourse relations are based on data analysis, but they do not pursue the goal of descriptive accuracy to the same extent. He points out that “an explanatory theory of coherence requires a set of externally driven principles to motivate and ultimately constrain the relation set”. This is the direction x Introduction taken by Sanders and colleagues (Sanders et al. 1992, 1993; Sanders and Noordman

2000; Sanders 1997, 2005) in advocating the cognitive approach to coherence rela-

tions, in which psychological plausibility is the primary motivating factor: Understanding a discourse means constructing a mental coherent representation of that discourse by the hearer. [...] An adequate account of the relations establishing coherence has to be psychologically plausible, because coherence relations are ultimately cognitive relations. (Sanders et al. 1992, 1).

Following Hobbs (1979), Sanders and colleagues (1992) argued that coherence relations point to coherence in the cognitive representation of discourse, and they see coherence relations as cognitive entities (Beaugrande and Dressler 1981; Mann and Thompson 1986; Sanders et al. 1992, 1993). At a more general level, they hoped to shed light on human cognition by investigating the mechanisms underlying discourse coherence. In this book, I endore this view of coherence relations and speak about temporal coherence both at the discursive and cognitive levels, with respect to the coherence established by comprehenders within and between the mental representations built during the comprehension process (Chap. 6).

A slightly different account of temporal relations has been proposed in the relevance-theoretic framework (Sperber and Wilson 1986; Wilson and Sperber 2012), where it is assumed that linguistic expressions underdetermine the content that a speaker explicitly and implicitly communicates. The hearer must therefore recover inferentially the speaker’s intended meaning, with the interpretative process guided by the expectation of relevance and the quest for cognitive effects. Regarding temporal relations holding between eventualities, Carston (1988) and Wilson and Sperber (1998) convincingly argued that they should be treated as pragmatically determined aspects of “what is said”. In other words, they are explicatures — that is, enriched forms of truth-functional propositional content. More specifically, linguistic expressions encode conceptual and procedural information (i.e. instructions for manipulating conceptual representations), respectively contributing to and constraining the interpretative process. As for verbal tenses, some scholars have argued they encode instructions on how to relate eventualities temporally, with respect to the moment of speech S, as well as to one another. In this book, I conduct annotation experiments, based on which I postulate the necessity of distinguishing between the temporal information from the categories of Tense, Aktionsart and Aspect. Consequently, I will attempt to propose an alternative model of the role that verbal tenses play in utterance and discourse comprehension, which is based on the conceptual and procedural content encoded by these three categories. I propose that Tense has a mixed conceptual-procedural nature, Aktionsart encodes conceptual information, whereas Aspect encodes procedural information (Chap. 5). Put differently, language encodes procedures on how to manipulate conceptual mental representations in order to ensure cognitive temporal coherence.

Why an Empirical Study?

It is easy to see linguists’ growing desire to use robust and objective findings in addition to intuitive and subjective acceptability judgments or built examples. This is maybe due to a perceived dissatisfaction with the methods of so-called armchair

Introduction xi linguistics (in the sense of Fillmore 1992). Empirical linguistics aims to support or challenge current theories by way of authentic and reliable data and to propose new models for the interpretation of linguistic phenomena. McEnery and Wilson (1996) highlight that, broadly speaking, linguists have tended to favour the use of data which are either introspective (i.e. language data constructed by linguists) or naturally occurring (i.e. examples of actual language usage). Most linguists see these two types of data as complementary approaches, and not as mutually exclusive. Gries (2002) argue that, although intuition may be poor as a methodology for investigating mental representations, linguists’ intuitions are useful for the formulation of testable hypotheses about linguistic structure and behaviour. Introspective and corpus data were the two main sources of data for theoretical linguistics until the mid-1990s. Other sources have since been considered, such as experimentation, language acquisition, language pathologies, neurolinguistics, etc. Kepser and Reis (2005) argue that linguistic evidence coming from different domains of data sheds more light on the issues investigated than where data are drawn from a unique source. Multisource converging evidence can either validate the theory or bring contradictory results, therefore opening up new perspectives.

In this research, both corpus data and experimental data are used. Corpus data were randomly selected from four stylistic registers (literary, journalistic, legislation and political discourse from the Europarl corpus), in order to draw generalized conclusions. Experimental data consist of natural data (i.e. corpus data) and controlled experimental items built for the specific purposes of the experiments. Corpus data come from four languages (English, and three Romance languages, French, Italian and Romanian) and are analysed contrastively, with the aim of proposing a crosslinguistic valid tertium comparationis. Qualitative and quantitative analyses are carried out both for corpus data and for experimental data. Experimental data are drawn from three types of investigation: annotation experiments, in which comprehenders are asked to evaluate consciously the meaning or the contextual usage of a linguistic category; offline experiments with acceptability judgement tasks, in which comprehenders are asked to evaluate consciously the acceptability of different versions of the experimental items; and online experiments with self-paced reading tasks, in which readers have to read experimental items in different experimental conditions at their own pace and which allow direct access to the participants’ processing of the experimental items.

These types of investigations and their subsequent types of data are complementary and necessary in pragmatic research. Regarding the complementarity of corpus and experimental data, Gilquin and Gries (2009) argue that a corpus has a fourfold purpose in experimentation: (a) validator, where the corpus serves as a validator of the experiment; (b) validatee, where the corpus is validated by the experiment; (c) equal, where corpus and experimental data are used on an equal footing; and (d) stimulus composition, where the corpus serves as a database for the items used in experiments. Compared to experimentation, corpus work deals with a larger range of phenomena that can be investigated. Experiments, however, allow for the study of phenomena that are infrequent in corpora. Moreover, corpus data are naturally occurring data, whereas experimental data are artificial data. Finally, manipulation of variables is only possible by way of the experimental design. Corpora and xii Introduction experiments have advantages and limits that are complementary. Nowadays, for these reasons, linguists tend to use both of these empirical methods.

Furthermore, this book proposes a modelization of the semantics and pragmatics of verbal tenses and of the categories Tense, Aspect and Aktionsart, that was implemented in the fields of Natural Language Processing and Machine Translation. The innovative model of temporal reference developed in this research consists of features that were shown to be valid and effective by natural language processing techniques serving to label raw corpus data automatically. The statistical machine translation systems trained on these automatically labelled data, with the features provided by the model of temporal reference developed in this book, gave better results when translating verbal tenses than did other systems which do not make use of these features. These improvements represent a solid empirical validation of the model proposed. More generally, this research highlights the need to develop linguistic and pragmatic models of language that are empirically and experimentally derived, not only for the testing of hypotheses issued from existing solid theories but for their validation by successful implementation in applied fields, such as natural language processing, machine translation and second language teaching, among others.

The Research Questions

The focus of this research on time and verbal tenses comes from an apparently simple empirical question: how can one improve the translation of verbal tenses by statistical machine translation systems? This question was asked in the context of two inter-disciplinary Swiss research projects, COMTIS and MODERN,! which aim to improve the quality of machine-translated texts in terms of their overall coherence. The coherence of a text depends on several cohesive ties, such as pronouns, discourse connectives and verbal tenses. Crosslinguistic analyses of bilingual and multilingual translation corpora pointed to one frequent and problematic translation divergency: the translation of the English Simple Past into French, Italian and Romanian (Chap. 3). For this translation divergency, four verbal tenses are most frequently used in these Romance languages: the simple past, the imperfect, the compound past and the present (as described in Chap. 1). In order to solve this empirical issue, other related issues had to be addressed, from theoretical, empirical and experimental perspectives. So, the research presented in this book can be organized according to four main research questions.

Firstly, what is the basis of the role of a verbal tense and its constituent categories as cohesive ties? Chaps. | and 2 present a comprehensive discussion of previous attempts to answer this question, as included in grammar books, and general linguistic, semantic and pragmatic studies. Where grammar books and a considerable amount of pragmatic studies give accounts of individual verbal tenses in English, French, Italian and Romanian (such as the English Simple Past, the French Passé Composé or Imparfait, the imperfect in Romance languages), general linguis-

'COMTIS (Improving coherence of machine translation output by modelling intersentential relations, SNSF CRSI22_127510, 2010-2013, https://www.idiap.ch/project/comtis) and MODERN (Modeling discourse entities and relations for coherence machine translation, SNSF CRSII2_ 147653, 2013-2016, https://www.idiap.ch/project/modern/front-page)

Introduction xiii tic studies focus on meaning and role at the discursive level of Tense, Aktionsart and Aspect. However, as formal semantic-discursive models in particular have pointed out, the role of Tense, Aktionsart and Aspect is not only sentential but mainly relational — that is, it goes beyond sentence boundaries. A detailed review of the relevance-theoretic pragmatic account shows that this theory is well placed to account for the role of these cohesive ties, for two reasons. The first is its assumption that language underderminates the speaker’s intended meaning, which must be recovered contextually. The second is the characterization of encoded meaning, in terms of conceptual or procedural types of information. During comprehension, hearers built conceptual mental representations, which are manipulated according to the procedural information encoded by certain linguistic expressions or linguistic categories. The comprehensive discussion in Chaps. 1 and 2 will lay the foundations for formulating a certain number of hypotheses with respect to the role of Tense, Aktionsart and Aspect in discourse comprehension. These hypotheses are put forward and tested in annotation experiments in Chap. 4.

Secondly, can temporal relations be treated as cognitive coherence relations linking the mental representations of eventualities built by comprehenders? This question is addressed in Chap. 6, which explores the cognitive bases of the temporal cohesive ties investigated in this research and of the temporal relations holding between eventualities. Building on Hobbs’ (1979, 1985), Sanders et al.’s (1992, 1993) and Evers-Vermeul et al.’s (2017) cognitive approaches to discourse relations, I will show that temporal relations are cognitively motivated for two reasons. The first is that they affect both processing and language acquisition (Mandler 1986; Segal et al. 1991; Murray 1995, 1997; Grisot and Blochowiak 2015, 2017; Evers- Vermeul et al.’s 2017). The role of two French temporal connectives (ensuite and puis) in expressing the sequential temporal relations holding between events expressed with the Passé Composé and the Passé Simple is thus assessed in Chap. 6. The second reason is that the linguistic categories triggering temporal relations (Tense, Aspect and Aktionsart), along with temporal connectives and temporal adverbials, are themselves cognitively motivated, as has been found by numerous experimental studies carried out in psychology and neurolinguistics.

In the light of this, I discuss the notion of temporal cognitive coherence, which is linked to the coherence established within the multithreaded mental representations that comprehenders build during language comprehension (Gernsbacher and Givon 1995; Graesser et al. 1997). In this model, language is seen as encoding processing instructions on how to construct mental representations of the situations described (Zwaan and Radvansky 1998). According to the innovative model of temporal reference put forward in this book, the category of Tense both contributes to and constrains the construction of mental representations. Furthermore, aspectual information from Aktionsart and Aspect also contributes to and constrains the construction of mental representations respectively. In particular, Aktionsart provides the type of eventuality to be included in the conceptual mental representation (state, activity, achievement or accomplishment), whereas Aspect constrains this process by instructing the comprehender to represent the eventuality as completed or in progress.

xiv Introduction

Thirdly, how do hearers consciously evaluate temporal information from Tense, Aspect and Aktionsart during the comprehension of naturally occurring data? This question is tackled in Chaps. 3 and 4. In Chap. 3, I describe the bilingual and other multilingual corpora on which the current research is based. The monolingual and crosslinguistic analyses of these corpora reveal a source language’s most frequent and most problematic verbal tenses with regard to their translation into a target language. In particular, the English Simple Past presents a unique challenge when it comes to its translation into French, Italian and Romanian. In Chap. 4, I assess comprehenders’ capacity to evaluate the meaning of a verbal tense consciously, with respect to a series of features drawn from the discussions provided in Chaps. 1 and 2. These features are linked to the semantics and pragmatics of Tense (i.e. the localization of eventualities with respect to the moment of speech S, as well as to one another, operationalized as the [+narrativity] feature), of Aktionsart (i.e. the actual realization of an eventuality in a context, operationalized as the [+boundedness] feature) and of Aspect (i.e. the speaker’s perfective or imperfective viewpoint, operationalized as the [+perfectivity] feature).

This experimental research has led me to put forward an innovative proposal regarding the role of the cohesive tie attributed to verbal tenses and its constituent categories (Chap. 5). I suggest a relevance-theoretic pragmatic account of Tense, Aktionsart and Aspect, according to which these categories encode conceptual and procedural types of information. This account is entailed by the Highly Discriminatory (HD) model of temporal reference, which aims to discriminate between the categories and principles that play a role in determining temporal reference, disregarding the type of language explored — that is, tensed, tenseless or mixed. This model predicts that the global interpretation of temporal reference at the discursive level is determined, on the one hand, by the linguistic means existent in a language and on the other hand, by their ad hoc inferential contextual determination. It distinguishes between temporal information supplied by Tense, Aktionsart, Aspect, Mood, temporal adverbials, temporal connectives, aspectual markers and markedness, among others (cf. Binnick 1991, 2012). In this chapter, I also revisit the verbal tenses explored in this book according to the HD model and re-analyse their meaning and their contextual usages in terms of their conceptual and procedural content.

Fourthly, how can one predict the verbal tense used in a target language? This question is directly linked to the applicative purpose of this research that is, to improve the results of statistical machine translation systems in terms of coherence. This issue is assessed both in the crosslinguistic analyses of annotated data, where the annotators’ answers are compared with the verbal tense used in the target language, and in a generalized mixed model in which the fixed factors tested are the features of [+narrativity], [+boundedness] and [+perfectivity]. The first two features are also assessed using methods relevant to the fields of Natural Language Processing and Statistical Machine Translation (Chap. 7). In this chapter, I describe automatic annotation experiments using classifiers and translation experiments performed by statistical machine translation systems, as implemented by my colleagues Thomas Meyer and Andrei Popescu-Belis for the [+narrativity] feature (Meyer et al. 2013;

Introduction Xv

Grisot and Meyer 2014; Meyer 2014) and by Sharid Lodiciga for the [+boundedness] feature (Lodiciga and Grisot 2016; Lodiciga 2017).

With all this in mind, this book critically reviews the limitations of the current models of temporal reference in general and of verbal tenses in particular and incorporates insights and contributions from different fields (corpus linguistics, contrastive analysis, theoretical and experimental pragmatics, psychology and psycholinguistics) in order to advance an innovative comprehensive model of temporal reference, based on relevance-theoretic notions such as the conceptual/procedural distinction. This research gives a new account of Tense, Aktionsart and Aspect as temporal cohesive ties, founded on a detailed corpus study and empirical findings which point to the need to account for corpus and experimental data when making claims about tendencies and regularities in language. Furthermore, it addresses how these cohesive ties play a crucial role in establishing temporal discourse coherence and cognitive coherence. This account has significant implications in the fields of natural language processing and machine translation and could in future be implemented in other fields, such as language teaching and translation studies.

The Structure of the Book

Chapter 1, “The Linguistic Expression of Temporal Reference”, consists of two main sections and ends with a summary. First, I address the description of four types of verbal tense (the simple past, the imperfect, the compound past and the present) in English, French, Italian and Romanian, as presented by grammar books and pragmatic studies focusing on individual verbal tenses. Second, I review the semantics of Tense, Aktionsart and Aspect, which are the constituent categories of the generic notion verbal tense, as they are discussed in general linguistics.

Chapter 2, “Formal Semantic-Discursive and Pragmatic Assessments of Temporal Reference”, contains three main sections and ends with a summary. First, I consider the contribution of verbal tenses to discourse interpretation and the calculation of temporal relations in several formal sematic-discursive theories. Second, I discuss Grice’s treatment of temporal relations as conversational implicatures. Third, having briefly introduced the basic tenets of Relevance Theory and the conceptual/procedural distinction, I pay particular attention to the relevance-theoretic account of temporal relations as “pragmatically determined aspects of what is said” and to the ongoing debate on the purely procedural vs. mixed nature of the meaning encoded by verbal tenses.

Chapter 3, “Corpus-Based Contrastive Study of Verbal Tenses”, consists of four main sections and ends with a summary. First, I give an account of the importance of using authentic, naturally occurring and systematic data, as well as of the advantages and limitations of using corpora in linguistic and pragmatic research. Second, I describe the three sets of translation corpora that were compiled for this research: bilingual English-French, bilingual French-English, multilingual English-French- Italian and Romanian. In these sets of data, I assessed the frequency of verbal tenses in the source language and their translations into a target language.

Chapter 4, “Experimental Study Using Annotation Experiments”, includes four main sections and ends with a summary. First, I discuss several issues linked to xvi Introduction using annotation data, such as reliability, validity and the measurement of inter-annotator agreement. Following my proposal in Grisot (2017a), I interpret inter-annotator agreement rates, measured with the K coefficient, as dependent on the degree of accessibility to consciousness and the degree of availability to conscious thought and, as such, on their conceptual or procedural nature. Second, I advance a series of hypotheses regarding the meaning of Tense, Aktionsart and Aspect, and their predictions with respect to comprehenders’ behaviour when they have to evaluate it consciously in annotation experiments. Third, I describe the annotation experiments and discuss their results. Fourth, in order to assess the role of Tense, Aktionsart and Aspect in predicting the verbal tense used in a target language, I discuss the results of a generalized mixed model suited to the data.

Chapter 5, “A Pragmatic Model of Temporal Cohesive Ties”, comprises four main sections and ends with a summary. First, I put forward the HD model, which is an innovative model of temporal reference distinguishing between the temporal information from Tense, Aktionsart, Aspect, Mood, temporal adverbials, temporal connectives, aspectual markers and markedness, among others. Second, I develop a mixed conceptual-procedural account of Tense, by specifying that the notion of context, referred to as ConText, is understood as a cognitive construct consisting of a set of assumptions selected during the interpretation process, rather than determined in advance and expanded during the interpretation process when the expectation of relevance is satisfied or abandoned. In the ConText, the comprehender inferentially constructs the conceptual content of Tense and Aktionsart and makes use of the procedural information encoded by Tense and Aspect in order to manipulate the conceptual representations built. Third, I revisit the verbal tenses investigated in this book according to the HD model.

Chapter 6, “Temporal Coherence”, consists of four main sections and ends with a summary. First, I provide a general account of coherence relations as cognitive entities and focus on the status of temporal relations. Second, in order to support the proposal that temporal relations are cognitively motivated, I address the results of a series of online and offline experiments carried out in order to test the processing and conscious evaluation of temporal relations, as expressed by the French Passé Composé versus the Passé Simple and by the temporal connectives ensuite and puis. Third, I tackle the notion of cognitive temporal coherence, by arguing that the cohesive ties investigated in this research are cognitively motivated categories and by exploring a psycholinguistic account of coherence according to which mental representations of discourse segments are structured and coherent.

Chapter 7, “Application to Natural Language Processing and Machine Translation”, contains two main sections and ends with a summary. First, I review the literature on Natural Language Processing and Machine Translation which has investigated temporal information. Second, I describe the automatic implementation of [+narrativity] and [+boundedness] and show that these are effective at improving the results of statistical machine translation systems, in terms of lexical choices of verbs and of verbal tenses in automatically translated texts.

The book ends with a conclusion, in which I summarize the main contributions of this research and propose areas of further study.

Contents

2.1 The Formal Semantic-Discursive Account PEENE ET 65 22. The Gricean Account; ssc ridoc irass EEE TEET EAT s+ 76 23 The Relevance- Theoretic Account... 202.064 sees sea nenna ns 79 2.3.1 Basic Relevance-Theoretic Tenets... 0.0.6 60000e ees hea T9