A Neuro-Symbolic Learning System for Analyzing Listing Images in the Short-Term Rental Industry
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
In this paper, we propose the automation of listing image related tasks in the short-term rental industry using neuro-symbolic AI system. The tasks performed by the system are the selection of main “hero” images from the pool of images available for each listing, and the recommendation of content-based image enhancement such as reducing clutter, incorporating accent colors, etc. Automating these tasks using approaches that rely exclusively on deep learning (end-to-end trained neural networks) are unable to produce accurate, explainable models due to two main issues: first, the lack of labelled training data available across the many segments (different geographical locations and listing types/sizes) that comprise the market. Second, the black box nature of neural networks makes it difficult to leverage knowledge that has been previously learnt and apply it to new rental market segments. To overcome these limitations, we used a hybrid system with a neural component for identifying features (symbols/objects) in images, and a symbolic component for reasoning over those symbols to produce a logic knowledgebase. The inclusion of a symbolic reasoning component produces a more explainable and market segment transferable model due to the creation of a knowledgebase that captures the abstract concepts amongst image features that drive listing click-through performance. This logic can be inspected, decomposed, and queried to produce explainable image recommendations, predict the image that will perform best in the market as hero images, and provide useful background knowledge when operating the system in new market segments.
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 | 0.001 | 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