7–11 August 2006, Málaga, Spain.
A workshop held as part of ESSLLI 2006, the Eighteenth European Summer School in Logic, Language and Information.
This paper presents novel psycholinguistic evidence on the factors governing pronoun resolution in German. To determine the relative influence of linear order versus grammatical function of potential referents in the German Mittelfeld on pronoun interpretation, an interpretation-preference task was run on 23 German natives. Subjects chose the preferred referent for fully ambiguous personal pronouns. The results across different verb types show that grammatical function, yet not linear order, predicts pronoun resolution in German.
Psycholinguistic experiments show that pronouns tend to be resolved differently depending on whether they occur in main or subordinate clauses. If a pronoun in a subordinate clause has more than one potential antecedent in the main clause, then the pronoun tends to refer to the antecedent which has a certain thematic role (depending on the verb and on the subordinating conjunction). In contrast, pronouns in main clauses tend to refer back to the subject of the previous main clause, and this tendency is not affected by any verbs or conjunctions. In natural language processing, these findings have recently led to a proposal that pronoun resolution systems should have a split architecture, i.e. that they should use different mechanisms for pronoun resolution in the two cases.
With the help of two parsed and coreference-annotated corpora, this paper estimates the impact of the split-architecture proposal. The findings of this work are as follows: (1) Subject pronouns in authentic texts behave the same way in main and subordinate clauses. (2) The number of sentences in which a split architecture would behave differently than a system that treats both cases the same way is close to zero. Therefore, a separate treatment of resolution within and across units is unlikely to improve the performance of any system. This result casts a doubt on the split-architecture proposal, and more generally on approaches that directly incorporate psycholinguistic results into performance-oriented algorithms for anaphora resolution without assessing the relative importance of the phenomena that underlie them.
Accessibility hierarchies (Ariel, 1990, 2001; Givón, 1992; Gundel, Hedberg, & Zacharski, 1993) assume that the form of anaphoric expressions signals the relative saliency of the antecedent. We argue that the form of relative pronouns in relative clauses has a similar function and therefore influences attachment preferences. We conducted two questionnaire experiments in which we investigated whether attachment preferences for ambiguous relative clauses are affected by the type of relative pronoun that is used. Experiment 1 showed a difference in attachment preference between qui and lequel, indicating that the form of the relative pronoun affects attachment preferences. Experiment 2 demonstrated that the difference observed in Experiment 1 is not due to differences in informativity between qui and lequel, suggesting that instead, it is due to a difference in markedness (qui is more frequent and shorter).
We present a model for resolution of plural references on groupings based on Reference Domains Theory. While the original theory doesn’t take plural reference into account, this paper shows how several entities can be grouped together by building a new domain and how they can be accessed later on. We introduce the notion of super-domain representing the access structure to all the plural referents of a given type.
This paper analyses one aspect of the so-called two-pronoun puzzle (or Dahl’s puzzle). I argue that the two-pronoun puzzle hides a double puzzle; I call them Puzzle A and Puzzle B. Puzzle A is why one logically possible interpretation disappears in certain cases of ellipsis of pronominal dependencies; and the other way around, Puzzle B is why this reading is still possible when ellipsis doesn’t apply. Puzzle A can be accounted for in terms of conditions on parallelism and semantic binding (cf. Fox 2000, Reinhart 2000, Büring 2005). Puzzle B has not yet been resolved in the same terms. In this paper I propose an explanation that is compatible with the analysis in Fox (2000). First, based on Basque data I show that the two-pronoun puzzle phenomenon is not restricted to VP ellipsis: it also concerns silent arguments. Then, I show how this Basque data help understanding better what is under Puzzle B: the facts appear to be less uniform once information structure is taken into account, and focus is shown to play a central role in the derivation that allows the crucial data.
We describe two pronoun interpretation experiments in which a Coherence Hypothesis is tested against preference-based accounts. The Coherence Hypothesis holds that apparent preferences in antecedent selection are actually byproducts of the inferencing processes used to establish different types of coherence. In Experiment 1 we show that preferences can be systematically disrupted through the manipulation of coherence and that when the relevant factors are balanced, preferences disappear. In Experiment 2 we show that the coherence effect is not disrupted by voice alternations (active/passive), providing evidence for a strong semantic model of coherence-driven interpretation. We speculate on the adequacy of this strong semantic model and propose additional online experiments to examine the interaction between propositional content and information structure influences on pronoun interpretation.
In this paper I present different types of ambiguity that occur in annotating and resolving anaphoric adverbials. My analysis concentrates on the German temporal adverbial danach (“after that” or “thereafter”). However, issues discussed here go beyond the analysis of this particular German word. I assume that the types of ambiguity presented in this paper occur not only with other anaphoric adverbials like therefore or besides but also with other types of anaphoric expressions (e.g. personal pronouns). At least, the “Justified Sloppiness Hypothesis” proposed by Poesio et al. (2005) leads to this conclusion. First, I present an account of the meaning and anaphoric character of the German pronominal adverb danach. By means of a pilot study, I then show that the resolution of danach involves several types of ambiguity, namely structural and referential ambiguity. Regarding referential ambiguity, we can distinguish the following sub-types: 1. competition between temporal and situation referents, 2. competition between simple and structured situation referents and 3. competition between different situation referents. Finally, I make some suggestions for handling these types of ambiguity.
Recent accounts of anaphor resolution propose a two-route architecture for the interpretation of pronominals (e.g., Grodzinsky & Reinhart, 1993; Reuland, 2001). A pronoun can either be resolved by a grammatical operation in logical syntax (i.e., variable binding) or through value assignment in discourse (i.e., co-reference). Reuland (2001) proposes that an interpretation through variable binding requires less processing resources and is therefore preferred over a co-reference interpretation. Rule I compares variable binding with co-reference interpretations to decide whether a co-reference dependency is allowed (Grodzinsky & Reinhart, 1993; Reinhart 2000). This rule prevents discourse processes from by-passing logical syntax where the latter rules out an interpretation as ungrammatical. The question is whether Rule I always compares variable binding with co-reference interpretations, or is only executed if a dependency is initially ruled out by logical syntax. In an eye-tracking experiment we manipulated the interpretation of ambiguous and unambiguous pronouns in an attempt to explore how variable binding, co-reference and Rule I influence the way readers resolve pronouns. The results show that if a pronoun was ambiguous between a variable binding and co-reference antecedent, the variable binding antecedent was initially preferred even if discourse information clearly favoured the co-reference antecedent. Therefore, we argue that logical syntactic processes function independently from discourse processes, indicating that it may be warranted to consider logical syntax and discourse as distinct modules of the language system. Furthermore, the language system does not seem to apply Rule I in cases where both variable binding and co-reference lead to the same grammatical interpretation suggesting that Rule I is only relevant in cases that are potentially ungrammatical.
The goal of this paper is to explore whether game theory can account for the distribution of pronouns in Catalan, a null-subject language. I extend the game-theoretical account proposed by Clark and Parikh (2006), which uses games of partial information to account for anaphora production and resolution in English. Their proposal needs to be extended to account for the fact that null-subject languages make extensive use of null pronouns in subject position, apart from overt pronouns, proper names and definite descriptions. The main idea is that speaker and hearer choose the most efficient strategy with the highest payoff, given the type of entity they want to refer to (more or less salient) and the kind of information they want to convey. Game theory provides an elegant framework to capture anaphora production and resolution, by giving payoffs to each option and probabilities to information states. However, anaphora production and resolution in Catalan constitute a challenge for this type of approach. While this approach elegantly captures the use of an overt pronoun to mark a special information structure, its use in intrasentential anaphora situations is more problematic to model.
Using a qualitative analysis of disagreements from a referentially annotated newspaper corpus, we show that, in coreference annotation, vague referents are prone to greater disagreement. We show how potentially problematic cases can be dealt with in a way that is practical even for larger-scale annotation, considering a real-world example from newspaper text.
Inquiries: anaphora@essex.ac.uk
The organizers are involved in the Arrau project (Anaphora Resolution and Underspecification).