MétaCan
Menu
← Back to cohort
Record W6990989201

Entity Typing with Natural Language Inference for Fine-Grained Named Entity Recognition

2023· other· fr· W6990989201 on OpenAlexfundno aff

Bibliographic record

VenuePolyPublie (École Polytechnique de Montréal) · 2023
Typeother
Languagefr
Field
Topic
Canadian institutionsnot available
FundersAlliance de recherche numérique du Canada
KeywordsContext (archaeology)Filter (signal processing)Subject (documents)Identification (biology)Nucleofection
DOInot available

Abstract

fetched live from OpenAlex

RÉSUMÉ: Le centre d’attention de ce mémoire est la reconnaissance d’entités nommées à fin degré de granularité (FgNER), qui consiste à détecter les mentions d’entités nommées dans des textes en anglais et à classifier chacune d’entre elles avec un type précis provenant d’une taxonomie. Cette taxonomie décrit des types "fins" qui sont plus complexes à assigner aux mentions que les types traditionnels de la reconnaissance d’entités nommées. Dans ce mémoire, notre premier objectif est de proposer un modèle d’inférence en langue naturelle (NLI) basé sur un modèle de langue pré-entrainé pour effectuer la tâche de typage d’entités à fin degré de granularité (ET) données en entrée. Nous proposons des patrons qui se basent sur les types pour améliorer l’inférence. Notre deuxième objectif est d’intégrer notre modèle d’ET dans un modèle complet de reconnaissance d’entités nommées (FgNER) tout en conservant des propriétés d’adaptabilité. Nous proposons les trois intégrations suivantes: un modèle de détection de mentions d’entités nommées suivi de notre modèle d’ET, un modèle de reconnaissance d’entités nommées FgNER combiné avec notre modèle d’ET et enfin une approche d’apprentissage par transfert utilisant les poids de l’encodeur et du décodeur de notre modèle d’ET pour les poids d’un modèle de FgNER. Notre recherche démontre que notre modèle de typage d’entités est comparable aux approches de l’état de l’art avec nos patrons de NLI. Nous avons également intégré avec succès notre modèle d’ET dans un modèle encodeur-décodeur de FgNER. ABSTRACT: The centre of attention of this thesis is Fine-grained Named Entity Recognition (FgNER), which consists of detecting named entities in English texts and classifying each of them with a fine-grained type from a taxonomy. This taxonomy is composed of fine-grained types that are harder to assign to named entities than traditional types from the Named Entity Recognition task. In this thesis, our first objective is to train a Natural Language Inference (NLI) model based on a pre-trained language model to achieve the fine-grained Entity Typing (ET) task. We propose templates based on types to improve inference. Our second objective is to integrate our best ET model inside a FgNER model while keeping the complete system adaptable. We propose the three following integrations: using Named Entity Detection followed by ET, using a FgNER model combined with ET and finally a transfer learning from our ET model encoder and decoder’s weights to the weights of a FgNER model. Our research shows that our ET model is comparable to state-of-the-art approaches when using our NLI templates. We also successfully integrated ET in an encoder-decoder FgNER model.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.004
metaresearch head score (Gemma)0.016
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: none
Teacher disagreement score0.010
Threshold uncertainty score0.024

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0040.016
Meta-epidemiology (narrow)0.0010.001
Meta-epidemiology (broad)0.0010.003
Bibliometrics0.0020.002
Science and technology studies0.0010.002
Scholarly communication0.0040.010
Open science0.0030.003
Research integrity0.0020.003
Insufficient payload (model declined to judge)0.0070.003

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.014
GPT teacher head0.254
Teacher spread0.240 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations0
Published2023
Admission routes1
Has abstractyes

Explore more

Same venuePolyPublie (École Polytechnique de Montréal)→French-language works237,207→