Detecting Drug-Drug Interaction (DDI) over the Social Media using Convolution Neural Network Deep Learning
Bibliographic record
Abstract
Drug-Drug Interaction (DDI) detection is a challenging problem for drug manufacturers, drug regulatory authorities, and medical professionals alike.It is impossible to run trials or be aware of every single case involving an entire population.Research in the use of social media data is currently gaining attention, and with the application of machine learning techniques has been successfully applied in businesses.This paper presents a project extracting DDI from biomedical text using a Convolutional Neural Network (CNN) classifier.The classifier is trained on the SemEval 2013 DDIExtraction challenge dataset and aims to automatically learn the best feature representation on the input of the given task.Different models have been proposed, which make use of position embeddings in combination with word embeddings trained on the machine learning model to learn features.Word embeddings are necessary for providing dense vector representation of words that can be trained, but a large amount of data is required to train an effective vector representation of words.To compensate for the lack shortage of data, the CNN model is trained on a pre-trained PubMed word embedding, which provides a vector dimension of size 200 for the representation of each word.This project aims to provide a trained CNN model for which vector representation of words is provided by weights that have been trained for medical text classification purposes.
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 imitationNot 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.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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 source (direct Gemma or distilled Codex), 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".