Dynamic Named Entity Recognition model to distinguish authors’ positions relative to mentioned locations
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
When sending a tweet, Twitter authors refer either to their current locations (in-situ location) or to distant (remote locations). Although essential for several spatial analysis tasks, distinguishing between in-situ and remote location entities poses significant challenges due to the ambiguity of natural language, informal and unstructured tweets, and dynamic location entities. To address these challenges, we developed a Dynamic Named Entity Recognition (DNER) model to differentiate between the two types of location entities in Twitter data. We investigated three annotation approaches and set up an annotation guideline based on strict grammatical rules on a decision tree. Using our custom-trained DNER model, we validated the effectiveness of our model on a validation dataset representative of the most prominent Twitter sources. Although to a marginal extent, our results show the model's effectiveness in distinguishing between in-situ and remote location mentions. Posts generated via native Twitter applications (Twitter for Android, iPhone and iPad) returned 64% and 32%, respectively, for the in-situ and remote locations within the bounding box of the geocoded locations. For Instagram posts, the returned percentages were 87% and 77%, respectively, for the in-situ and remote locations. We conclude that the DNER model is more effective on native applications than on Instagram-generated posts. • A Dynamic Named Entity Recognition (DNER) model performs better on data from native Twitter applications than on Instagram-generated posts. • Decision trees aid in annotating highly ambiguous posts • Group-based annotations are effective for consistent and diverse results in complex NER tasks. • Data composition affects model and validation results
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.
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 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 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 it