A Discourse Coherence Analysis Method Combining Sentence Embedding and Dimension Grid

Discourse coherence is strongly associated with text quality, making it important to natural language generation and understanding. However, existing coherence models focus on measuring individual aspects of coherence, such as lexical overlap, entity centralization, rhetorical structure, etc., lacking measurement of the semantics of text. In this paper, we propose a discourse coherence analysis method combining sentence embedding and the dimension grid, we obtain sentence-level vector representation by deep learning, and we introduce a coherence model that captures the fine-grained semantic transitions in text. Our work is based on the hypothesis that each dimension in the embedding vector is exactly assigned a stated certainty and specific semantic. We take every dimension as an equal grid and compute its transition probabilities. The document feature vector is also enriched to model the coherence. Finally, the experimental results demonstrate that our method achieves excellent performance on two coherence-related tasks.

A Study of Zero Anaphora Resolution in Chinese Discourse: From the Perspective of Psycholinguistics

Anaphora is the main linguistic means to establish discourse coherence, and anaphora resolution is the psychological process to maintain this coherence. Chinese discourse is characterized with providing multiplicity of linguistic clues to readers by employing various referential apparatuses such as pronoun anaphora, zero anaphora, and so on. As a way of avoiding repeated reference to an object that is mentioned beforehand, zero anaphora is frequently employed in discourse. The production and resolution of zero anaphora largely concerns some constraints underlying psychological mechanisms. We particularly focus on zero anaphora resolution in the present study to try to discover some specific aspects of the underlying mechanism, hoping to find out some factors unique to the resolution process. We designed the first two experiments to probe into the psychological reality when participants were presented with sentences containing either pronoun anaphora or zero anaphora or both under discourse condition with topic continuity in Experiment 1a and topic discontinuity in Experiment 1b. We did not find any significant difference in the reaction time between zero anaphora resolution and pronoun anaphora resolution, indicating that zero anaphora possibly works within the processing mechanism on which pronoun anaphora resolution depends. However, we found significantly longer time in reading the first sentence in any of the discourse, showing that the first-mention effect exists in anaphora resolution. We further explored the time course of zero anaphora resolution by measuring the reaction time during the period when participants read sentences that varied according to the location where zero anaphora occurred under two conditions: topic continuity (Experiment 2a) vs. topic discontinuity (Experiment 2b). The strategies of searching for the referential information were found divergent: the exhaustive searching strategy was adopted when the topics within a discourse were kept continuous and the heuristic searching strategy was employed when the topics were discontinuous. The design of Experiment 5 took the factor of voice type and situation consistency into consideration, investigating in what way do those factors influence the resolution of zero anaphora. The voice type, according to the results, plays a significant role for its exclusively close relationship with the first-mention effect.

Modeling hypotactic structure for Chinese-English neural machine translation of complex sentences

The hypotactic structural relation between clauses plays an important role in improving the discourse coherence of document-level translation. However, the standard neural machine translation (NMT) models do not explicitly model the hypotactic relationship between clauses, which usually leads to structurally incorrect translations of long and complex sentences. This problem is particularly noticeable on Chinese-to-English translation task of complex sentences due to the grammatical form distinction between English and Chinese. English is rich in grammatical form (e.g. verb morphological changes and subordinating conjunctions) while Chinese is poor in grammatical form. These linguistic phenomena make it a challenge for NMT to learn the hypotactic structure knowledge from Chinese as well as the structure alignment between Chinese and English. To address these issues, we propose to model the hypotactic structure for Chinese-to-English complex sentence translation by introducing hypotactic structure knowledge. Specifically, we annotate and build a hypotactic structure aligned parallel corpus that provides rich hypotactic structure knowledge for NMT. Moreover, we further propose a structure-infused neural framework to combine the hypotactic structure knowledge with the NMT model through two integrating strategies. In particular, we introduce a specific structure-aware loss to encourage the NMT model to better learn the structure knowledge. Experimental results on WMT17, WMT18 and WMT19 Chinese-to-English translation tasks demonstrate the effectiveness of the proposed methods.

Salience and shift in salience as means of creating discourse coherence

The Chipaya language, an endangered isolate of the Bolivian highlands, has a set of three enclitics, =l, =m and =ʐ, which are coreferential with the subject of a clause but are not necessarily attached to it and are not obligatory. In this paper, I investigate the pragmatic function of these forms. The salience-marking enclitics (henceforth SMEs) occur at paratactic and hypotactic discourse transitions, where they indicate a shift in salience, thereby contributing to creating discourse coherence. Discourse transitions without a shift in salience are not accompanied by the enclitics. Those enclitics that occur at paratactic transitions have scope over at least the segment whose beginning and/or end they occur in, whereas SMEs at hypotactic transitions have scope over the clause they appear in. Use of the SMEs is genre-specific.

How focus particles and accents affect attachment

This project shows that focus and information structure, as indicated by the focus particle “only” and pitch accents, influence syntactic attachment, in contrast to the well-known effects of prosodic boundaries on attachment. One written questionnaire, one completion study, and several auditory questionnaires show that the position of “only” strongly affects attachment preferences in ambiguous sentences, while contrastive pitch accents have smaller effects. The two types of focus marking do not interact but independently impact attachment. These results support a modified version of the Focus Attraction Hypothesis, with ambiguous material drawn to attach to the most important information in a sentence. This research shows that information structure can affect sentence structure as well as discourse coherence.

Discourse Relations and Connectives in Higher Text Structure

The present article investigates possibilities and limits of local (shallow) analysis of discourse coherence with respect to the phenomena of global coherence and higher composition of texts. We study corpora annotated with local discourse relations in Czech and partly in English to try and find clues in the local annotation indicating a higher discourse structure. First, we classify patterns of subsequent or overlapping pairs of local relations, and hierarchies formed by nested local relations. Special attention is then given to relations crossing paragraph boundaries and their semantic types, and to paragraph-initial discourse connectives. In the third part, we examine situations in which annotators incline to marking a large argument (larger than one sentence) of a discourse relation even with a minimality principle annotation rule in place. Our analyses bring (i) new linguistic insights regarding coherence signals in local and higher contexts, e.g. detection and description of hierarchies of local discourse relations up to 5 levels in Czech and English, description of distribution differences in semantic types in cross-paragraph and other settings, identification of Czech connectives only typical for higher structures, or the detection of prevalence of large left-sided arguments in locally annotated data; (ii) as another type of contribution, some new reflections on methodologies of the approaches under scrutiny.

Anaphoricity in emoji: An experimental investigation of face and non-face emoji

Emoji are widely used, but have received relatively little attention in psycholinguistic research. Upon encountering a message consisting of both text and emoji, readers presumably construct some link between emoji and text. Based on a psycholinguistic study on text-emoji relations, we argue for (at least) two types of emoji-text dependencies, related to referential dependencies known to exist in the linguistic domain, namely (i) the dependency between an expressive (e.g. wow, damn, f*king) and the individual whose opinion it expresses, and (ii) the dependency between a pronoun (or other pro-form) and its antecedent. We extend the discussion of these dependencies to emoji, and provide experimental data that face emoji resemble expressives in that they tend to be interpreted as expressing the opinion of a salient experiencer, while action emoji are interpreted based on principles of discourse coherence (e.g. discourse relations like explanation), similar to what coherence-based accounts of pronoun resolution predict.

Clause Structure and Illocutionary Force in Medieval Gallo-Romance: Clitic Position in Old Occitan and Early Old French Sentential Coordination

This paper analyzes the position of object and adverbial clitic pronouns in coordinated affirmative verb-first declaratives introduced by e(t) “and” in Old Occitan and early Old French, a context in which clitics are variably preverbal or postverbal. An empirical study reveals that this variation is principled and reflects semantico-discursive properties in the same way in these two related and grammatically similar medieval Gallo-Romance varieties. On a theoretical level, I posit that preverbal clitics occur when conjunction occurs at the TP level, and postverbal clitics occur when conjunction occurs at the CP level, and that the choice of clause structure (TP vs. CP) for second conjunct clauses depends on illocutionary force, which in turn depends on discourse coherence relations and the semantics of verba dicendi (verbs of utterance).

Clause Structure and Illocutionary Force in Medieval Gallo-Romance: Clitic Position in Old Occitan and Early Old French Sentential Coordination

This paper analyzes the position of object and adverbial clitic pronouns in coordinated affirmative verb-first declaratives introduced by e(t) “and” in Old Occitan and early Old French, a context in which clitics are variably preverbal or postverbal. An empirical study reveals that this variation is principled and reflects semantico-discursive properties in the same way in these two related and grammatically similar medieval Gallo-Romance varieties. On a theoretical level, I posit that preverbal clitics occur when conjunction occurs at the TP level, and postverbal clitics occur when conjunction occurs at the CP level, and that the choice of clause structure (TP vs. CP) for second conjunct clauses depends on illocutionary force, which in turn depends on discourse coherence relations and the semantics of verba dicendi (verbs of utterance).