Geo-Spatial Market Segmentation & Characterization Exploiting User Generated Text Through Transformers & Density-Based Clustering
Bibliographic record
Abstract
In data analysis, context information plays a significant role in enhancing the quality of the insight obtained. Furthermore, spatial analysis helps understand spatial relationships among entities. Nevertheless, findings of a comprehensive literature review show that the characterization of geographic areas based on user generated content, such as text messages, has not been sufficiently explored. This paper focuses on investigating how to combine and exploit geographic information with user generated text content to detect geographic clusters of textual events, and infer relationships between each cluster and a fixed set of retail product categories, which we consider as an insightful way to perform spatial market segmentation. We propose a workflow composed of several machine learning models incorporating Transformers as an attention mechanism and BERT-based data augmentation capable of predicting product classes from Amazon product reviews and Twitter message corpora, and then characterizing the obtained geographic clusters based on their aggregated scores. The output of our system is an effective visualization of the geographic areas with their corresponding relevance score against a fixed set of categories. We trained a product document classifier achieving an F1-Score of 86% in the test set for product reviews, and of 76% in the test set for tweets; and validated our approach by manually annotating a subset of Twitter data with respect to ten product categories. Our approach provides practitioners with a mechanism to combine location context, a Transformer encoder, and transfer learning to derive insights from geo-spatial and text data; and researchers with opportunities to continue advancing the field.
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 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.004 | 0.003 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.002 | 0.002 |
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".