Recursive deep models for semantic compositionality over a sentiment treebank¶
Authors: R. Socher; A. Perelygin; J. Wu; J. Chuang; C. D. Manning; A. Ng; C. Potts
Year: 2013
type: dataset collection: sok-references status: auto
Links: link
Summary¶
Auto-extracted summary (unreviewed)
Socher et al. introduce the Stanford Sentiment Treebank, a corpus of 11,855 movie-review sentences with fine-grained sentiment labels on all 215,154 phrases of their parse trees, together with the Recursive Neural Tensor Network (RNTN) for compositional sentiment prediction. Trained on the treebank, the RNTN raises sentence-level positive/negative accuracy from 80% to 85.4% and reaches 80.7% fine-grained accuracy over all phrases, and it is the only model that correctly captures negation and its scope.
Abstract¶
Semantic word spaces have been very useful but cannot express the meaning of longer phrases in a principled way. Further progress towards understanding compositionality in tasks such as sentiment detection requires richer supervised training and evaluation resources and more powerful models of composition. To remedy this, we introduce a Sentiment Treebank. It includes fine grained sentiment labels for 215,154 phrases in the parse trees of 11,855 sentences and presents new challenges for sentiment compositionality. To address them, we introduce the Recursive Neural Tensor Network. When trained on the new treebank, this model outperforms all previous methods on several metrics. It pushes the state of the art in single sentence positive/negative classification from 80% up to 85.4%. The accuracy of predicting fine-grained sentiment labels for all phrases reaches 80.7%, an improvement of 9.7% over bag of features baselines. Lastly, it is the only model that can accurately capture the effects of negation and its scope at various tree levels for both positive and negative phrases.
Source: pdf
Extracted details¶
| Field | Value | Source | Confidence |
|---|---|---|---|
| contribution | Stanford Sentiment Treebank with fine-grained labels for 215,154 phrases in 11,855 sentences | llm | 0.90 |
| contribution | Recursive Neural Tensor Network (RNTN) model for compositional sentiment | llm | 0.90 |
| contribution | new state of the art in binary and fine-grained sentence sentiment classification | llm | 0.90 |
| contribution | analysis of contrastive conjunction and negation handling | llm | 0.90 |
| models | RNTN | llm | 0.90 |
| models | RNN (recursive) | llm | 0.90 |
| models | MV-RNN | llm | 0.90 |
| models | Naive Bayes | llm | 0.90 |
| models | SVM | llm | 0.90 |
| models | bigram Naive Bayes | llm | 0.90 |
| models | VecAvg | llm | 0.90 |
| datasets | Stanford Sentiment Treebank (SST) | llm | 0.90 |
| datasets | Pang and Lee (2005) movie reviews | llm | 0.90 |
| tasks | sentiment classification | llm | 0.90 |
| tasks | fine-grained 5-class sentiment | llm | 0.90 |
| tasks | binary sentiment classification | llm | 0.90 |
| tasks | semantic compositionality | llm | 0.90 |
| key result | accuracy (binary, sentence-level): 85.4% (RNTN on SST, positive/negative full sentences (previous SOTA 80%)) | llm | 0.90 |
| key result | accuracy (fine-grained, all phrases): 80.7% (RNTN on SST, 5-class, +9.7% over bag-of-features baselines) | llm | 0.90 |
| key result | accuracy (fine-grained, sentence-level): 45.7% (RNTN on SST root nodes) | llm | 0.90 |
| key result | accuracy (contrastive conjunction): 41% (RNTN on 131 'X but Y' test cases vs MV-RNN 37, RNN 36, biNB 27) | llm | 0.90 |
SoK reference [42]¶
R. Socher, A. Perelygin, J. Wu, J. Chuang, C. D. Manning, A. Ng, and C. Potts, “Recursive deep models for semantic compositionality over a sentiment treebank,” in Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, D. Yarowsky, T. Baldwin, A. Korhonen, K. Livescu, and S. Bethard, Eds. Seattle, Washington, USA: Association for Computational Linguistics, Oct. 2013, pp. 1631–1642. [Online]. Available: https://aclanthology.org/D13-1170/
Listed on the reference page of the site.
Keyword mentions¶
Counts of vocabulary terms in the full text (first page of occurrence in brackets).
- model: RNN / LSTM 27 [p.2], SVM 4 [p.2], Naive Bayes 3 [p.2], Autoencoder 2 [p.5]
- technique: Softmax 9 [p.4]
- task: Sentiment analysis 4 [p.3]
Files¶
| Path (under papers/) | Collection | Pages | Size | SHA-256 | Note |
|---|---|---|---|---|---|
references/sok-references/042_recursive-deep-models-for-semantic-compositionality-over-a-s.pdf |
sok-references | 12 | 1.3 MB | c08f19ffc32b |
Database key: recursive-deep-models-for-semantic-compositionality-over-a-sentiment · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set recursive-deep-models-for-semantic-compositionality-over-a-sentiment COLUMN VALUE