Efficient Natural Language Response Suggestion for Smart Reply
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:QT26B7RWrecord.jsonopen to challenge →
read the original abstract
This paper presents a computationally efficient machine-learned method for natural language response suggestion. Feed-forward neural networks using n-gram embedding features encode messages into vectors which are optimized to give message-response pairs a high dot-product value. An optimized search finds response suggestions. The method is evaluated in a large-scale commercial e-mail application, Inbox by Gmail. Compared to a sequence-to-sequence approach, the new system achieves the same quality at a small fraction of the computational requirements and latency.
This paper has not been read by Pith yet.
Forward citations
Cited by 17 Pith papers
-
SkMTEB: Slovak Massive Text Embedding Benchmark and Model Adaptation
SkMTEB is the first comprehensive text embedding benchmark for Slovak, and vocabulary-trimmed E5 adaptations achieve competitive performance with much smaller models.
-
SimCSE: Simple Contrastive Learning of Sentence Embeddings
SimCSE achieves 76.3% unsupervised and 81.6% supervised Spearman's correlation on STS tasks with BERT-base, improving prior best results by 4.2% and 2.2% via simple contrastive learning.
-
Dense Passage Retrieval for Open-Domain Question Answering
Dense dual-encoder retrievers outperform BM25 by 9-19% absolute in top-20 passage retrieval accuracy across open-domain QA datasets and enable new state-of-the-art end-to-end QA results.
-
ProtSent: Protein Sentence Transformers
Contrastive fine-tuning of protein language models on Pfam, structural, interaction, and mutational datasets produces embeddings that improve kNN performance on 15-16 of 23 downstream tasks including remote homology d...
-
Field Order Should Not Matter: Permutation-Invariant Embedding Model Fine-Tuning for Structured Metadata Retrieval
Permutation-invariant fine-tuning (PI-FT) randomizes field order and applies dropout during embedding model training to eliminate sensitivity to serialization order, reducing order-change penalty from 7.4 to 0.2 nDCG@...
-
LCSHBench: A Multilingual, Consensus-Grounded Benchmark for Library of Congress Subject Heading Assignment
LCSHBench supplies a consensus-filtered multilingual dataset for LCSH assignment evaluation together with exact and concept-level metrics, and reports that fine-tuning a 300M embedder yields higher exact recall@200 th...
-
To MRL or not to MRL: Text Embeddings are Robust to Truncation Without Matryoshka Learning, Except In Heavy Truncation Scenarios
Text embeddings are robust to truncation without MRL except when reducing size by at least 80%.
-
ORPHEAS: A Cross-Lingual Greek-English Embedding Model for Retrieval-Augmented Generation
ORPHEAS, a Greek-English embedding model created with knowledge graph fine-tuning, outperforms state-of-the-art multilingual models on monolingual and cross-lingual retrieval benchmarks.
-
REPLUG: Retrieval-Augmented Black-Box Language Models
REPLUG improves frozen black-box LMs by prepending LM-supervised retrieved documents, delivering 6.3% better language modeling on GPT-3 and 5.1% better five-shot MMLU on Codex.
-
To MRL or not to MRL: Text Embeddings are Robust to Truncation Without Matryoshka Learning, Except In Heavy Truncation Scenarios
Truncated embeddings from non-MRL models perform comparably to or better than MRL-trained models for most truncation levels, except heavy truncation of 80% or more.
-
Domain-Adapted Retrieval for In-Context Annotation of Pedagogical Dialogue Acts
Domain-adapted utterance-level retrieval raises Cohen's kappa for tutoring dialogue act annotation to 0.526-0.580 on TalkMoves and 0.659-0.743 on Eedi, beating no-retrieval baselines by large margins across three LLMs.
-
Don't Retrieve, Generate: Prompting LLMs for Synthetic Training Data in Dense Retrieval
LLM-generated synthetic hard negatives for training dense retrievers consistently underperform corpus-mined negatives from BM25 and cross-encoders across 10 BEIR datasets, with non-monotonic gains from scaling the gen...
-
Search for Coverage: Learning Coverage-Aware Retrieval with Augmented Sub-Question Answerability
CoveR improves nugget coverage by 10% over dense baselines in long-form RAG via coverage-aware contrastive training on LLM-generated sub-question signals without losing relevance performance.
-
The Multilingual Curse at the Retrieval Layer: Evidence from Amharic
Zero-shot multilingual retrievers underperform monolingual Amharic models by 23% relative MRR@10; fine-tuned multilingual models still trail the best monolingual baseline.
-
Reinforced Preference Optimization for Reasoning-Augmented Recommendations
RPORec unifies LLM reasoning with a recommendation head through reasoning-augmented modeling and reinforced preference optimization to improve recommendation accuracy and interpretability.
-
Unified Supervision for Walmart's Sponsored Search Retrieval via Joint Semantic Relevance and Behavioral Engagement Modeling
A hybrid supervision method for bi-encoder retrievers combines graded relevance from teacher models, production retrieval priors, and selective engagement to improve relevance and NDCG over Walmart's current sponsored...
-
OCC: A Smart Reply System for Efficient In-App Communications
Uber's OCC system uses unsupervised embeddings plus nearest-neighbor intent detection followed by historical reply retrieval, reporting 76% intent accuracy and 71% production usage in English-speaking countries.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.