Pascal Denis

Specialized Models and Ranking for Coreference Resolution

This paper investigates two strategies for improving coreference resolution: (1) training separate models that specialize in particular types of mentions (e.g., pronouns versus proper nouns) and (2) using a ranking loss function rather than a classification function. In addition to being conceptually simple, these modifications of the standard single-model, classification-based approach also deliver significant performance improvements. Specifically, we show that on the ACE corpus both strategies produce f-score gains of more than 3% across the three coreference evaluation metrics (MUC, B3, and CEAF).

Predicting globally-coherent temporal structures from texts via endpoint inference and graph decomposition

An elegant approach to learning temporal orderings from texts is to formulate this problem as a constraint optimization problem, which can be then given an exact solution using Integer Linear Programming. This works well for cases where the number of possible relations between temporal entities is restricted to the mere precedence relation [Bramsen et al., 2006; Chambers and Jurafsky, 2008], but becomes impractical when considering all possible interval relations. This paper proposes two innovations, inspired from work on temporal reasoning, that control this combinatorial blow-up, therefore rendering an exact ILP inference viable in the general case. First, we translate our network of constraints from temporal intervals to their endpoints, to handle a drastically smaller set of constraints, while preserving the same temporal information. Second, we show that additional efficiency is gained by enforcing coherence on particular subsets of the entire temporal graphs. We evaluate these innovations through various experiments on TimeBank 1.2, and compare our ILP formulations with various baselines and oracle systems.

Comparing Word Representations for Implicit Discourse Relation Classification

This paper presents a detailed comparative framework for assessing the usefulness of unsupervised word representations for identifying so-called implicit discourse relations. Specifically, we compare standard one-hot word pair representations against low-dimensional ones based on Brown clusters and word embeddings. We also consider various word vector combination schemes for deriving discourse segment representations from word vectors, and compare representations based either on all words or limited to head words. Our main finding is that denser representations systematically outperform sparser ones and give state-of-the-art performance or above without the need for additional hand-crafted features.

Extracting and visualizing quotations from news wires

We introduce SAPIENS, a platformfor extracting quotations fromnews wires, associated with their author and context. The originality of SAPIENS is that it relies on a deep linguistic processing chain, which allows for extracting quotations with a wide coverage and an extended definition, including quotations which are only partially quotes-delimited verbatim transcripts. We describe the architecture of SAPIENS and how it was applied to process a corpus of French news wires from the AFP news agency.

Leveraging different meronym discovery methods for bridging resolution in french

This paper presents a statistical system for resolving bridging descriptions in French, a language for which current lexical resources have a very low coverage. The system is similar to that developed for English by [22], but it was enriched to integrate meronymic information extracted automatically from both web queries and raw text using syntactic patterns. Through various experiments on the DEDE corpus [8], we show that although still mediocre the performance of our system compare favorably to those obtained by [22] for English. In addition, our evaluation indicates that the different meronym extraction methods have a cumulative effect but that the text pattern-based extraction method is more robust and leads to higher accuracy than the Web-based approach.

Learning Connective-based Word Representations for Implicit Discourse Relation Identification

We introduce a simple semi-supervised approach to improve implicit discourse relation identification. This approach harnesses large amounts of automatically extracted discourse connectives along with their arguments to construct new distributional word representations. Specifically, we represent words in the space of discourse connectives as a way to directly encode their rhetorical function. Experiments on the Penn Discourse Treebank demonstrate the effectiveness of these task-tailored representations in predicting implicit discourse relations. Our results indeed show that, despite their simplicity, these connective-based representations outperform various off-the-shelf word embeddings, and achieve state-of-the-art performance on this problem.

Fast Gaussian pairwise constrained spectral clustering

We consider the problem of spectral clustering with partial supervision in the form of must-link and cannot-link constraints. Such pairwise constraints are common in problems like coreference resolution in natural language processing. The approach developed in this paper is to learn a new representation space for the data together with a distance in this new space. The representation space is obtained through a constraint-driven linear transformation of a spectral embedding of the data. Constraints are expressed with a Gaussian function that locally reweights the similarities in the projected space. A global, non-convex optimization objective is then derived and the model is learned via gradient descent techniques. Our algorithm is evaluated on standard datasets and compared with state of the art algorithms, like [14,18,31]. Results on these datasets, as well on the CoNLL-2012 coreference resolution shared task dataset, show that our algorithm significantly outperforms related approaches and is also much more scalable.