@inproceedings {6223, title = {A Gold Anaphora Annotation Layer on an Eye Movement Corpus}, year = {2018}, month = {05/2018}, pages = {3518-3522}, author = {Olga Seminck and Pascal Amsili} } @conference {6089, title = {The Time Course of Pronoun Resolution in Natural Text Reading}, year = {2018}, month = {02/2018}, address = {T{\"u}bingen}, author = {Olga Seminck and Pascal Amsili} } @inproceedings {5563, title = {A Computational Model of Human Preferences for Pronoun Resolution}, year = {2017}, month = {04/2017}, abstract = {

We present a cognitive computational model of pronoun resolution that reproduces the human interpretation preferences of the Subject Assignment Strategy and the Parallel Function Strategy. Our model relies on a probabilistic pronoun resolution system trained on corpus data. Factors influencing pronoun resolution are represented as features weighted by their relative importance. The importance the model gives to the preferences is in line with psycholinguistic studies. We demonstrate the cognitive plausibility of the model by running it on experimental items and simulating antecedent choice and reading times of human participants. Our model can be used as a new means to study pronoun resolution, because it captures the interaction of preferences.\ 

}, author = {Olga Seminck and Pascal Amsili} } @inproceedings {5616, title = {A Google-Proof Collection of French Winograd Schemas}, year = {2017}, month = {04/2017}, pages = {24-29}, abstract = {

This article presents the first collection of French Winograd Schemas. Winograd Schemas form anaphora resolution problems that can only be resolved with extensive world knowledge. For this reason the Winograd Schema Challenge has been proposed as an alternative to the Turing Test. A very important feature of Winograd Schemas is that it should be impossible to resolve them with statistical information about word co-occurrences: they should be Google-proof. We propose a measure of Google-proofness based on Mutual Information, and demonstrate the method on our collection of French Winograd Schemas.

}, author = {Pascal Amsili and Olga Seminck} } @conference {5635, title = {Predicting Processing Cost of Anaphora Resolution}, year = {2017}, month = {03/2017}, address = {Saarbr{\"u}cken}, author = {Olga Seminck and Pascal Amsili} } @inproceedings {5900, title = {Schémas Winograd en français: une étude statistique et comportementale}, volume = {2}, year = {2017}, month = {06/2017}, pages = {28-35}, publisher = {Association pour le Traitement Automatique des Langues}, edition = {24}, address = {Orl{\'e}ans}, abstract = {

Nous présentons dans cet article une collection de schémas Winograd en français, adaptée de la liste proposée par Levesque et al. (2012) pour l\’anglais. Les schémas Winograd sont des problèmes de résolution d\’anaphore conçus pour être IA-complets. Nous montrons que notre collection vérifie deux propriétés cruciales : elle est robuste vis-à-vis de méthodes statistiques simples (\“Google-proof\”), tout en étant largement dépourvue d\’ambiguïté pour les sujets humains que nous avons testés.

Winograd schemas in French : a statistical and behavioral study

We present in this paper a collection of Winograd schemas in French, adapted from the English collection proposed by Levesque et al. (2012). Winograd schemas constitute anaphora resolution problems meant to be AI-complete. We show that our collection has two crucial properties : it is robust regarding simple statistical techniques of resolution (\“Google-proof\”) and basically non ambiguous for the human participants that were tested.

}, author = {Pascal Amsili and Pascal Amsili and Olga Seminck} } @conference {4785, title = {A cognitively plausible model for anaphora resolution}, year = {2016}, month = {09/2016}, address = {Bilbao}, abstract = {

Anaphora resolution a is complex problem, as it deals with syntax, semantics and discourse. The subject is well studied in field of psycholinguistics, where multiple preferences were discovered, and in the field of computational linguistics, where many systems have been developed to perform anaphora resolution in documents. Nevertheless, the work done in the two fields remains disconnected.\ 

We investigate how we can bridge the gap by exploiting the options for making a cognitively plausible model for anaphora resolution. Such model can be beneficent for both fields as it can inspire computational linguistics with findings about how humans process anaphora, and help the psycholinguistic community developing large coverage, incremental models simulating the human processing of anaphora.\ 

Inspired by the surprisal framework that uses incremental probabilistic parsing to predict processing cost coming from syntax for each word in a corpus, we focus on incremental probabilistic systems of anaphora resolution from the field of computational linguistics, for example the largely spread pair-wise model, and turn them into a cognitive model of processing of anaphora. We will present some preliminary results on the measures of processing cost of anaphora we developed and the perspective of evaluating the model on eye-tracking data.\ 

}, author = {Olga Seminck and Pascal Amsili} } @inproceedings {4783, title = {Un mod{\`e}le simple du co{\^u}t cognitif de la r{\'e}solution des anaphores}, year = {2016}, month = {07/2016}, address = {Paris}, abstract = {

Nous pr{\'e}sentons un travail en cours sur un projet de recherche en TAL et en psycholinguistique. Le but de notre projet est de mod{\'e}liser le co{\^u}t cognitif que repr{\'e}sente la r{\'e}solution d\’anaphores. Nous voulons obtenir une mesure du co{\^u}t cognitif continue et incr{\'e}mentale qui peut, {\`a} un stade de recherche plus avanc{\'e}, {\^e}tre corr{\'e}l{\'e}e avec des mesures d\’occulom{\'e}trie sur corpus. Pour cela, nous proposons une mod{\'e}lisation inspir{\'e}e par des techniques venues du TAL. Nous utilisons un solveur d\’anaphores probabiliste bas{\'e} sur l\’algorithme couples de mentions et la notion d\’entropie pour {\'e}tablir une mesure du co{\^u}t cognitif des anaphores. Ensuite, nous montrons par des visualisations quelles sont les pr{\'e}dictions de cette premi{\`e}re mod{\'e}lisation pour les pronoms personnels de troisi{\`e}me personne dans le corpus ANCOR Centre.\ 

}, author = {Olga Seminck} }