Evaluation-driven design of a robust coreference resolution system

Popescu-Belis, Andréi

doi:10.1017/s135132490300319x

Cited by 3 publications

(6 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…". The resolution of references to documents offers the advantage of a restricted set of candidate entities, when compared to anaphora or coreference resolution [3][4][5][6].…”

Section: What Are References To Documents?mentioning

confidence: 99%

Detection and Resolution of References to Meeting Documents

Popescu-Belis

Lalanne

2006

Machine Learning for Multimodal Interaction

Self Cite

View full text Add to dashboard Cite

Abstract. This article describes a method for document/speech alignment based on explicit verbal references to documents and parts of documents, in the context of multimodal meetings. The article focuses on the two main stages of dialogue processing for alignment: the detection of the expressions referring to documents in transcribed speech, and the recognition of the documents and document elements that they refer to. The detailed evaluation of the implemented modules, first separately and then in a pipeline, shows that results are well above baseline values. The integration of this method with other techniques for document/speech alignment is finally discussed.

show abstract

“…". The resolution of references to documents offers the advantage of a restricted set of candidate entities, when compared to anaphora or coreference resolution [3][4][5][6].…”

Section: What Are References To Documents?mentioning

confidence: 99%

Detection and Resolution of References to Meeting Documents

Popescu-Belis

Lalanne

2006

Machine Learning for Multimodal Interaction

Self Cite

View full text Add to dashboard Cite

show abstract

“…Semantic Processing: identifying meaning; this is primarily done via Information Extraction (IE) which utilizes a set of rules for pattern matching, which in turn make use of various resources, such as gazetteers, lexicons, shallow syntax parsing, etc. 4. Discourse Integration: connecting meaning between sentences, handling coreference, managing other idiosyncrasies of the English language that make it useful for humans, but difficult for machines.…”

Section: Elementary Natural Language Processingmentioning

confidence: 99%

“…The discourse problem is still an open field of research, as attested to by Andrei Popescu-Belis: "As more and more language resources become available to computers, the numeric evaluation of coreference/anaphora resolution on unrestricted texts becomes possible, but perfect scores are still far from reach." [4] As such, it would make sense for the Lab to form a research team of software developers and natural language processors to investigate and develop DI tools.…”

Section: Discourse Integration (Di)mentioning

confidence: 99%

“…ARDA is a U.S. Government entity that sponsors and promotes IT research and development for the Intelligence Community." 4 ARDA has three focus areas: Quantum Information Science, Information Exploitation, and Global Information System Access.…”

Section: Nlp In the Government Sectormentioning

confidence: 99%

“…• Andrei Popescu-Belis [4] provides an approach similar to Shankaranarayana and Cyre in that he looks sequentially at each referring expression (RE) in a document and compares it to the REs previously considered. Instead of giving the first RE to reference a real-world entity the status of a "definition," he creates a separate object called a discourse entity (DE) to which the RE refers.…”

Section: B2 Discourse Integrationmentioning

confidence: 99%

See 2 more Smart Citations

Natural Language Processing as a Discipline at LLNL

Firpo¹

2005

View full text Add to dashboard Cite

The field of Natural Language Processing (NLP) is described as it applies to the needs of LLNL in handling free-text. The state of the practice is outlined with the emphasis placed on two specific aspects of NLP: Information Extraction and Discourse Integration. A brief description is included of the NLP applications currently being used at LLNL. A gap analysis provides a look at where the technology needs work in order to meet the needs of LLNL. Finally, recommendations are made to meet these needs.

show abstract

Anisotropic span embeddings and the negative impact of higher-order inference for coreference resolution: An empirical analysis

Hou,

Wang,

et al. 2024

Nat. Lang. Eng.

View full text Add to dashboard Cite

Coreference resolution is the task of identifying and clustering mentions that refer to the same entity in a document. Based on state-of-the-art deep learning approaches, end-to-end coreference resolution considers all spans as candidate mentions and tackles mention detection and coreference resolution simultaneously. Recently, researchers have attempted to incorporate document-level context using higher-order inference (HOI) to improve end-to-end coreference resolution. However, HOI methods have been shown to have marginal or even negative impact on coreference resolution. In this paper, we reveal the reasons for the negative impact of HOI coreference resolution. Contextualized representations (e.g., those produced by BERT) for building span embeddings have been shown to be highly anisotropic. We show that HOI actually increases and thus worsens the anisotropy of span embeddings and makes it difficult to distinguish between related but distinct entities (e.g., pilots and flight attendants). Instead of using HOI, we propose two methods, Less-Anisotropic Internal Representations (LAIR) and Data Augmentation with Document Synthesis and Mention Swap (DSMS), to learn less-anisotropic span embeddings for coreference resolution. LAIR uses a linear aggregation of the first layer and the topmost layer of contextualized embeddings. DSMS generates more diversified examples of related but distinct entities by synthesizing documents and by mention swapping. Our experiments show that less-anisotropic span embeddings improve the performance significantly (+2.8 F1 gain on the OntoNotes benchmark) reaching new state-of-the-art performance on the GAP dataset.

show abstract

Evaluation-driven design of a robust coreference resolution system

Cited by 3 publications

References 35 publications

Detection and Resolution of References to Meeting Documents

Detection and Resolution of References to Meeting Documents

Natural Language Processing as a Discipline at LLNL

Anisotropic span embeddings and the negative impact of higher-order inference for coreference resolution: An empirical analysis

Contact Info

Product

Resources

About