Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TeachPro: Multi-Label Qualitative Teaching Evaluation via Cross-View Graph Synergy and Semantic Anchored Evidence Encoding

Xiangqian Wang, Yifan Jia, Yang Xiang, Yumin Zhang, Yanbin Wang, Ke Liu

TL;DR

TeachPro reframes student feedback analysis as a multi-label, multi-dimensional task that links open-ended comments to five actionable teaching dimensions. It introduces a Cross-View Graph Synergy Network to capture syntactic and semantic relations in comments and a Dimension-Anchored Evidence Encoder that grounds each facet with learnable prompts and prototype-guided evidence, all under a parameter-efficient low-rank prediction head. A new TeachScope dataset provides expert-annotated, evidence-grounded supervision across five dimensions, enabling interpretable, facet-level diagnostics. Across extensive experiments, TeachPro outperforms strong baselines, with ablations confirming the contributions of dual-graph encoding, prototype-guided query refinement, and efficient low-rank classification, demonstrating robust, granular, and evidence-supported teaching assessments.

Abstract

Standardized Student Evaluation of Teaching often suffer from low reliability, restricted response options, and response distortion. Existing machine learning methods that mine open-ended comments usually reduce feedback to binary sentiment, which overlooks concrete concerns such as content clarity, feedback timeliness, and instructor demeanor, and provides limited guidance for instructional improvement.We propose TeachPro, a multi-label learning framework that systematically assesses five key teaching dimensions: professional expertise, instructional behavior, pedagogical efficacy, classroom experience, and other performance metrics. We first propose a Dimension-Anchored Evidence Encoder, which integrates three core components: (i) a pre-trained text encoder that transforms qualitative feedback annotations into contextualized embeddings; (ii) a prompt module that represents five teaching dimensions as learnable semantic anchors; and (iii) a cross-attention mechanism that aligns evidence with pedagogical dimensions within a structured semantic space. We then propose a Cross-View Graph Synergy Network to represent student comments. This network comprises two components: (i) a Syntactic Branch that extracts explicit grammatical dependencies from parse trees, and (ii) a Semantic Branch that models latent conceptual relations derived from BERT-based similarity graphs. BiAffine fusion module aligns syntactic and semantic units, while a differential regularizer disentangles embeddings to encourage complementary representations. Finally, a cross-attention mechanism bridges the dimension-anchored evidence with the multi-view comment representations. We also contribute a novel benchmark dataset featuring expert qualitative annotations and multi-label scores. Extensive experiments demonstrate that TeachPro offers superior diagnostic granularity and robustness across diverse evaluation settings.

TeachPro: Multi-Label Qualitative Teaching Evaluation via Cross-View Graph Synergy and Semantic Anchored Evidence Encoding

TL;DR

TeachPro reframes student feedback analysis as a multi-label, multi-dimensional task that links open-ended comments to five actionable teaching dimensions. It introduces a Cross-View Graph Synergy Network to capture syntactic and semantic relations in comments and a Dimension-Anchored Evidence Encoder that grounds each facet with learnable prompts and prototype-guided evidence, all under a parameter-efficient low-rank prediction head. A new TeachScope dataset provides expert-annotated, evidence-grounded supervision across five dimensions, enabling interpretable, facet-level diagnostics. Across extensive experiments, TeachPro outperforms strong baselines, with ablations confirming the contributions of dual-graph encoding, prototype-guided query refinement, and efficient low-rank classification, demonstrating robust, granular, and evidence-supported teaching assessments.

Abstract

Standardized Student Evaluation of Teaching often suffer from low reliability, restricted response options, and response distortion. Existing machine learning methods that mine open-ended comments usually reduce feedback to binary sentiment, which overlooks concrete concerns such as content clarity, feedback timeliness, and instructor demeanor, and provides limited guidance for instructional improvement.We propose TeachPro, a multi-label learning framework that systematically assesses five key teaching dimensions: professional expertise, instructional behavior, pedagogical efficacy, classroom experience, and other performance metrics. We first propose a Dimension-Anchored Evidence Encoder, which integrates three core components: (i) a pre-trained text encoder that transforms qualitative feedback annotations into contextualized embeddings; (ii) a prompt module that represents five teaching dimensions as learnable semantic anchors; and (iii) a cross-attention mechanism that aligns evidence with pedagogical dimensions within a structured semantic space. We then propose a Cross-View Graph Synergy Network to represent student comments. This network comprises two components: (i) a Syntactic Branch that extracts explicit grammatical dependencies from parse trees, and (ii) a Semantic Branch that models latent conceptual relations derived from BERT-based similarity graphs. BiAffine fusion module aligns syntactic and semantic units, while a differential regularizer disentangles embeddings to encourage complementary representations. Finally, a cross-attention mechanism bridges the dimension-anchored evidence with the multi-view comment representations. We also contribute a novel benchmark dataset featuring expert qualitative annotations and multi-label scores. Extensive experiments demonstrate that TeachPro offers superior diagnostic granularity and robustness across diverse evaluation settings.
Paper Structure (24 sections, 25 equations, 5 figures, 5 tables)

This paper contains 24 sections, 25 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Datasets
  4. Methodology
  5. Cross-View Graph Synergy Network
  6. Contextual Initialization with BERT
  7. Graph Induction and Relational Propagation
  8. Syntactic graph.
  9. Semantic graph.
  10. BiAffine Cross-View Fusion and Readout
  11. Dimension-Anchored Evidence Encoder
  12. Dimension Word Encoding and Review-Based Prototype Segmentation
  13. Prototype-Guided Interaction for Query Refinement
  14. Cross-Attention Fusion for Dimension Evidence Extraction
  15. Parameter-Efficient Multi-Dimensional Evaluation Prediction Head
  16. ...and 9 more sections

Figures (5)

  • Figure 1: Heatmap of correlation between different dimensions in the dataset
  • Figure 2: The overall framework of the model in this article.
  • Figure 3: How inter-dimensional correlation changes as model depth increases
  • Figure 4: Evidence alignment analysis across training epochs. (a) Precision measures the proportion of model-highlighted tokens that correspond to annotated evidence; (b) Recall measures the proportion of annotated evidence covered by model attention; (c) IoU quantifies the overlap between predicted and gold evidence spans; and (d) entropy reflects the concentration of attention, with lower entropy indicating more focused evidence attribution.
  • Figure 5: Visualization of token-level attention weights from each instructional dimension embedding to the input comment. Each panel shows, for a fixed epoch, how attention mass is allocated over tokens for the five dimensions, with warmer colors indicating higher normalized attention weights.