Bibliographic record
Abstract
In recent years we have witnessed many successes of neural networks in the information \nretrieval community with lots of labeled data. Yet it remains unknown whether the same \ntechniques can be easily adapted to search social media posts where the text is much \nshorter. In addition, we find that most neural information retrieval models are compared \nagainst weak baselines. In this thesis, we build an end-to-end neural information retrieval \nsystem using two toolkits: Anserini and MatchZoo. In addition, we also propose a novel \nneural model to capture the relevance of short and varied tweet text, named MP-HCNN. \nWith the information retrieval toolkit Anserini, we build a reranking architecture based \non various traditional information retrieval models (QL, QL+RM3, BM25, BM25+RM3), \nincluding a strong pseudo-relevance feedback baseline: RM3. With the neural network \ntoolkit MatchZoo, we offer an empirical study of a number of popular neural network \nranking models (DSSM, CDSSM, KNRM, DUET, DRMM). Experiments on datasets from \nthe TREC Microblog Tracks and the TREC Robust Retrieval Track show that most \nexisting neural network models cannot beat a simple language model baseline. How- \never, DRMM provides a significant improvement over the pseudo-relevance feedback baseline \n(BM25+RM3) on the Robust04 dataset and DUET, DRMM and MP-HCNN can provide \nsignificant improvements over the baseline (QL+RM3) on the microblog datasets. Further \ndetailed analyses suggest that searching social media and searching news articles exhibit \nseveral different characteristics that require customized model design, shedding light on \nfuture directions.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
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".