Research map / Label-efficient and robust learning / Recognition and long-tail
Microsummarization of Online Reviews: An Experimental Study
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Brown cluster, word2vec features to semi-CRF (ablation) | semi-CRF · Facet recognition phrase level, Food | F1: 0.69 -> 0.79 | Helped |
| add Brown cluster features (ablation) | semi-CRF basic · Facet recognition phrase level, Venue | F1: 0.17 -> 0.54 | Helped |
| add word2vec on top of Brown clusters (ablation) | semi-CRF · Facet recognition phrase level, Food | F1: 0.77 -> 0.79 | Helped |
| word2vec instead of spectral CCA embeddings (ablation) | semi-CRF w/ Brown · Facet recognition phrase level, Amenities | F1: 0.47 -> 0.46 | No clear effect |
| add unsupervised features (Brown) at word level (ablation) | semi-CRF basic · Facet recognition word level | recall | Hurt |
| supervised semi-CRF instead of unsupervised heuristics | facet recognition · Facet recognition phrase level, Food | F1: 0.49 -> 0.79 | Helped |
| relabel neutral training spans as positive (ablation) | sentiment classifier · Facet sentiment classification | accuracy: 77.8 -> 77.9 | No clear effect |
| resample POS at 1.5x NEG and NEU equal to NEG (ablation) | sentiment classifier · Facet sentiment classification | accuracy: 77.8 -> 74.1 | Hurt |
| microsummary instead of extractive summary (ClusterSum) | summary presentation · Crowd preference | % preferred: 80.0% vs 16.0% | Helped |
| FacetSum instead of ClusterSum/KLSum as source selection (ablation) | microsummary · Crowd preference | % preferring microsummary: 80.0% -> 68.0% | Hurt |
| ClusterSum vs SumBasic and KLSum as source selection (ablation) | microsummary · Crowd preference | % preferring microsummary: 78.4% / 79.2% / 80.0% | No clear effect |
Closest papers by meaning
- Weakly-Supervised Opinion Summarization by Leveraging External Information (AAAI 2020 · similarity 0.86)
- PersonalSum: A User-Subjective Guided Personalized Summarization Dataset for Large Language Models (NeurIPS 2024 · similarity 0.85)
- Tell Me What Is Good about This Property: Leveraging Reviews for Segment-Personalized Image Collection Summarization (AAAI 2024 · similarity 0.84)
- Practical Sentiment Analysis for Education: The Power of Student Crowdsourcing (AAAI 2024 · similarity 0.84)
- MultiSum: A Multi-Facet Approach for Extractive Social Summarization Utilizing Semantic and Sociological Relationships (AAAI 2024 · similarity 0.82)
- Intelligent Calibration for Bias Reduction in Sentiment Corpora Annotation Process (AAAI 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.