MétaCan
Menu
Back to cohort
Record W4394905126

Systèmes de questions-réponses interactifs à grande échelle

2022· dissertation· en· W4394905126 on OpenAlexaboutno aff
Oralie Cattan

Bibliographic record

Venuetheses.fr (ABES) · 2022
Typedissertation
Languageen
FieldComputer Science
TopicTopic Modeling
Canadian institutionsnot available
Fundersnot available
KeywordsScale (ratio)Computer scienceInformation retrievalData scienceGeographyCartography
DOInot available

Abstract

fetched live from OpenAlex

Informational search has evolved with our need for immediacy and intuitiveness into a form of natural language querying, no longer solely focused on the use of relevant keywords. The study of these interactions raises major issues in the field of machine understanding with regard to the contextualization of questions. Indeed, questions are rarely asked in isolation. Grouped together, they form a dialogue that is built and structured over the course of the conversation. In the following series of questions: “How much does a hotel room cost in Montreal? », « how to prepare a Basque cake », « what are black bees? », « do they sleep? », the interpretation of some questions depends on the questions and answers previously asked. In this context, designing an interactive question-answering system capable of sustaining a conversation that is not limited to a simple succession of sporadic questions and answers constitutes a challenge in terms of contextual modeling and high-performance computing. The evolution of intensive computing techniques and solutions, the availability of large volumes of raw data (in the case of unsupervised learning) or enriched with linguistic or semantic information (in the case of supervised learning) have allowed machine learning methods to experience significant development, with considerable applications in the industrial sector. Despite their success, these domain and language models, learned from a massive amount of data with a large number of parameters, raise questions of usability and today appear less than optimal, given the new challenges of digital sobriety. In a real business scenario, where systems are developed rapidly and are expected to work robustly for an increasing variety of domains, tasks and languages, fast and efficient learning from a limited number of examples is essential. In this thesis we deepen each of the aforementioned issues and propose approaches based on the knowledge transfer from latent and contextual representations to optimize performance and facilitate a cost-effective large-scale deployment of systems.

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)0.009
metaresearch head score (Gemma)0.030
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 score0.030
Threshold uncertainty score0.099

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0090.030
Meta-epidemiology (narrow)0.0030.002
Meta-epidemiology (broad)0.0030.003
Bibliometrics0.0030.002
Science and technology studies0.0020.002
Scholarly communication0.0090.011
Open science0.0050.006
Research integrity0.0050.004
Insufficient payload (model declined to judge)0.0300.019

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 head0.018
GPT teacher head0.290
Teacher spread0.272 · 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
Published2022
Admission routes1
Has abstractyes

Explore more

Same venuetheses.fr (ABES)Same topicTopic ModelingFrench-language works237,207