# bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification architecture

> Architecture diagram of the GitHub repository bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-11-21). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification
- Repository: https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification

## Overview

This repository is a reproducible IMDb sentiment-classification research pipeline comparing six BERT pooling and fine-tuning strategies. The main workflow is configured experiment execution: load and split IMDb data, tokenize it, train/select models using validation results, evaluate on test data, and preserve metrics and predictions for analysis. Separate post-hoc tools examine gate weights, sequence lengths, and paired predictions. Sampled excerpts and the README support the principal orchestration; some model and data-loader internals were not sampled, so those links rely on documented capabilities.

Main entry points: [`experiments/run_experiments.py`](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/experiments/run_experiments.py) (Experiment runner), [`data/dataset_loader.py`](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/data/dataset_loader.py) (IMDb data prep), [`training/evaluate.py`](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/training/evaluate.py) (Evaluation), [`main.py`](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/main.py) (CLI entry)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_experiment["Experiment workflow"]
  node_entry["CLI entry<br/>[main.py]"]
  node_config["Run configuration<br/>[config.py]"]
  node_runner["Experiment runner<br/>[run_experiments.py]"]
  node_dataset["IMDb data prep<br/>[dataset_loader.py]"]
  node_artifacts[("Run artifacts")]
end

subgraph group_models["Classifier strategies"]
  node_finetune["Fine-tuned CLS<br/>[finetuned_bert.py]"]
  node_frozencls["Frozen CLS<br/>[frozen_cls.py]"]
  node_frozenmean["Frozen mean"]
  node_attention["Frozen attention"]
  node_static["Static control"]
  node_adaptive["Adaptive pooling"]
  node_pooling["Pooling utilities<br/>[pooling.py]"]
end

subgraph group_training["Training and evaluation"]
  node_trainer["Training loop<br/>[train.py]"]
  node_evaluator["Evaluation<br/>[evaluate.py]"]
  node_metrics["Classification metrics<br/>[metrics.py]"]
  node_parameters["Parameter counts<br/>[parameter_count.py]"]
end

subgraph group_analysis["Results analysis"]
  node_summaries["Run summaries"]
  node_gateanalysis["Gate analysis<br/>[analyze_gates.py]"]
  node_lengthanalysis["Length analysis<br/>[analyze_lengths.py]"]
  node_comparison["Prediction comparison"]
  node_statistics["Bootstrap statistics<br/>[statistics.py]"]
end

node_researcher(("Researcher"))
node_huggingface["Hugging Face"]

node_researcher -->|"runs"| node_entry
node_entry -->|"invokes"| node_runner
node_runner -->|"reads options"| node_config
node_runner -->|"loads data"| node_dataset
node_dataset -->|"downloads dataset"| node_huggingface
node_dataset -.->|"downloads tokenizer"| node_huggingface
node_runner -->|"selects model"| node_finetune
node_runner -->|"selects model"| node_frozencls
node_runner -->|"selects model"| node_frozenmean
node_runner -->|"selects model"| node_attention
node_runner -->|"selects model"| node_static
node_runner -->|"searches candidates"| node_adaptive
node_attention -->|"uses mask"| node_pooling
node_runner -->|"trains model"| node_trainer
node_trainer -->|"validates model"| node_evaluator
node_runner -->|"tests selected model"| node_evaluator
node_evaluator -->|"computes metrics"| node_metrics
node_runner -->|"counts parameters"| node_parameters
node_runner -->|"writes run files"| node_artifacts
node_evaluator -->|"writes predictions"| node_artifacts
node_researcher -->|"analyzes results"| node_summaries
node_researcher -->|"analyzes gates"| node_gateanalysis
node_researcher -->|"analyzes lengths"| node_lengthanalysis
node_lengthanalysis -->|"loads splits"| node_dataset
node_researcher -->|"compares predictions"| node_comparison
node_comparison -.->|"bootstraps differences"| node_statistics

click node_entry "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/main.py"
click node_config "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/config.py"
click node_runner "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/experiments/run_experiments.py"
click node_dataset "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/data/dataset_loader.py"
click node_artifacts "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/tree/main/results"
click node_finetune "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/finetuned_bert.py"
click node_frozencls "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/frozen_cls.py"
click node_frozenmean "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/frozen_mean_pooling.py"
click node_attention "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/frozen_attention_pooling.py"
click node_static "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/nonadaptive_control.py"
click node_adaptive "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/adaptive_frozen_bert.py"
click node_pooling "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/pooling.py"
click node_trainer "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/training/train.py"
click node_evaluator "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/training/evaluate.py"
click node_metrics "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/utils/metrics.py"
click node_parameters "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/utils/parameter_count.py"
click node_summaries "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/scripts/summarize_results.py"
click node_gateanalysis "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/scripts/analyze_gates.py"
click node_lengthanalysis "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/scripts/analyze_lengths.py"
click node_comparison "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/scripts/compare_predictions.py"
click node_statistics "https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/utils/statistics.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_entry,node_config,node_runner,node_dataset,node_artifacts toneBlue
class node_finetune,node_frozencls,node_frozenmean,node_attention,node_static,node_adaptive,node_pooling toneAmber
class node_trainer,node_evaluator,node_metrics,node_parameters toneMint
class node_summaries,node_gateanalysis,node_lengthanalysis,node_comparison,node_statistics toneRose
class node_researcher,node_huggingface toneIndigo
```

## Components

### Experiment workflow

- [CLI entry](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/main.py): `main.py`
- [Run configuration](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/config.py): `config.py`
- [Experiment runner](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/experiments/run_experiments.py): `experiments/run_experiments.py`
- [IMDb data prep](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/data/dataset_loader.py): `data/dataset_loader.py`
- [Run artifacts](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/tree/main/results): `results`

### Classifier strategies

- [Fine-tuned CLS](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/finetuned_bert.py): `models/finetuned_bert.py`
- [Frozen CLS](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/frozen_cls.py): `models/frozen_cls.py`
- [Frozen mean](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/frozen_mean_pooling.py): `models/frozen_mean_pooling.py`
- [Frozen attention](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/frozen_attention_pooling.py): `models/frozen_attention_pooling.py`
- [Static control](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/nonadaptive_control.py): `models/nonadaptive_control.py`
- [Adaptive pooling](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/adaptive_frozen_bert.py): `models/adaptive_frozen_bert.py`
- [Pooling utilities](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/models/pooling.py): `models/pooling.py`

### Training and evaluation

- [Training loop](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/training/train.py): `training/train.py`
- [Evaluation](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/training/evaluate.py): `training/evaluate.py`
- [Classification metrics](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/utils/metrics.py): `utils/metrics.py`
- [Parameter counts](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/utils/parameter_count.py): `utils/parameter_count.py`

### Results analysis

- [Run summaries](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/scripts/summarize_results.py): `scripts/summarize_results.py`
- [Gate analysis](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/scripts/analyze_gates.py): `scripts/analyze_gates.py`
- [Length analysis](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/scripts/analyze_lengths.py): `scripts/analyze_lengths.py`
- [Prediction comparison](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/scripts/compare_predictions.py): `scripts/compare_predictions.py`
- [Bootstrap statistics](https://github.com/bert-aiml/evaluating-sentence-embedding-techniques-for-efficient-bert-based-sentiment-classification/blob/main/utils/statistics.py): `utils/statistics.py`

### Other components

- **Researcher**
- **Hugging Face**

## Connections

- Researcher → CLI entry: runs (evidence: `README.md`)
- CLI entry → Experiment runner: invokes (evidence: `main.py`)
- Experiment runner → Run configuration: reads options (evidence: `experiments/run_experiments.py`)
- Experiment runner → IMDb data prep: loads data (evidence: `experiments/run_experiments.py`)
- IMDb data prep → Hugging Face: downloads dataset (evidence: `README.md`)
- IMDb data prep → Hugging Face: downloads tokenizer (evidence: `README.md`)
- Experiment runner → Fine-tuned CLS: selects model (evidence: `README.md`)
- Experiment runner → Frozen CLS: selects model (evidence: `README.md`)
- Experiment runner → Frozen mean: selects model (evidence: `README.md`)
- Experiment runner → Frozen attention: selects model (evidence: `README.md`)
- Experiment runner → Static control: selects model (evidence: `README.md`)
- Experiment runner → Adaptive pooling: searches candidates (evidence: `README.md`)
- Frozen attention → Pooling utilities: uses mask (evidence: `models/frozen_attention_pooling.py`)
- Experiment runner → Training loop: trains model (evidence: `experiments/run_experiments.py`)
- Training loop → Evaluation: validates model (evidence: `training/train.py`)
- Experiment runner → Evaluation: tests selected model (evidence: `experiments/run_experiments.py`)
- Evaluation → Classification metrics: computes metrics (evidence: `training/evaluate.py`)
- Experiment runner → Parameter counts: counts parameters (evidence: `experiments/run_experiments.py`)
- Experiment runner → Run artifacts: writes run files (evidence: `README.md`)
- Evaluation → Run artifacts: writes predictions (evidence: `README.md`)
- Researcher → Run summaries: analyzes results (evidence: `README.md`)
- Researcher → Gate analysis: analyzes gates (evidence: `README.md`)
- Researcher → Length analysis: analyzes lengths (evidence: `README.md`)
- Length analysis → IMDb data prep: loads splits (evidence: `scripts/analyze_lengths.py`)
- Researcher → Prediction comparison: compares predictions (evidence: `README.md`)
- Prediction comparison → Bootstrap statistics: bootstraps differences
