Promptagator: Few-shot Dense Retrieval From 8 Examples

Much recent research on information retrieval has focused on how to transfer from one task (typically with abundant supervised data) to various other tasks where supervision is limited, with the implicit assumption that it is possible to generalize from one task to all the rest. However, this overlooks the fact that there are many diverse and unique retrieval tasks, each targeting different search intents, queries, and search domains. In this paper, we suggest to work on Few-shot Dense Retrieval, a setting where each task comes with a short description and a few examples. To amplify the power of a few examples, we propose Prompt-base Query Generation for Retriever (Promptagator), which leverages large language models (LLM) as a few-shot query generator, and creates task-specific retrievers based on the generated data. Powered by LLM's generalization ability, Promptagator makes it possible to create task-specific end-to-end retrievers solely based on a few examples {without} using Natural Questions or MS MARCO to train %question generators or dual encoders. Surprisingly, LLM prompting with no more than 8 examples allows dual encoders to outperform heavily engineered models trained on MS MARCO like ColBERT v2 by more than 1.2 nDCG on average on 11 retrieval sets. Further training standard-size re-rankers using the same generated data yields another 5.0 point nDCG improvement. Our studies determine that query generation can be far more effective than previously observed, especially when a small amount of task-specific knowledge is given.

ED2LM: Encoder-Decoder to Language Model for Faster Document Re-ranking Inference

State-of-the-art neural models typically encode document-query pairs using cross-attention for re-ranking. To this end, models generally utilize an encoder-only (like BERT) paradigm or an encoder-decoder (like T5) approach. These paradigms, however, are not without flaws, i.e., running the model on all query-document pairs at inference-time incurs a significant computational cost. This paper proposes a new training and inference paradigm for re-ranking. We propose to finetune a pretrained encoder-decoder model using in the form of document to query generation. Subsequently, we show that this encoder-decoder architecture can be decomposed into a decoder-only language model during inference. This results in significant inference time speedups since the decoder-only architecture only needs to learn to interpret static encoder embeddings during inference. Our experiments show that this new paradigm achieves results that are comparable to the more expensive cross-attention ranking approaches while being up to 6.8X faster. We believe this work paves the way for more efficient neural rankers that leverage large pretrained models.

Large Dual Encoders Are Generalizable Retrievers

It has been shown that dual encoders trained on one domain often fail to generalize to other domains for retrieval tasks. One widespread belief is that the bottleneck layer of a dual encoder, where the final score is simply a dot-product between a query vector and a passage vector, is too limited to make dual encoders an effective retrieval model for out-of-domain generalization. In this paper, we challenge this belief by scaling up the size of the dual encoder model {\em while keeping the bottleneck embedding size fixed.} With multi-stage training, surprisingly, scaling up the model size brings significant improvement on a variety of retrieval tasks, especially for out-of-domain generalization. Experimental results show that our dual encoders, \textbf{G}eneralizable \textbf{T}5-based dense \textbf{R}etrievers (GTR), outperform %ColBERT~\cite{khattab2020colbert} and existing sparse and dense retrievers on the BEIR dataset~\cite{thakur2021beir} significantly. Most surprisingly, our ablation study finds that GTR is very data efficient, as it only needs 10\% of MS Marco supervised data to achieve the best out-of-domain performance. All the GTR models are released at https://tfhub.dev/google/collections/gtr/1.

Sentence-T5: Scalable Sentence Encoders from Pre-trained Text-to-Text Models

We provide the first exploration of text-to-text transformers (T5) sentence embeddings. Sentence embeddings are broadly useful for language processing tasks. While T5 achieves impressive performance on language tasks cast as sequence-to-sequence mapping problems, it is unclear how to produce sentence embeddings from encoder-decoder models. We investigate three methods for extracting T5 sentence embeddings: two utilize only the T5 encoder and one uses the full T5 encoder-decoder model. Our encoder-only models outperforms BERT-based sentence embeddings on both transfer tasks and semantic textual similarity (STS). Our encoder-decoder method achieves further improvement on STS. Scaling up T5 from millions to billions of parameters is found to produce consistent improvements on downstream tasks. Finally, we introduce a two-stage contrastive learning approach that achieves a new state-of-art on STS using sentence embeddings, outperforming both Sentence BERT and SimCSE.

Neural Passage Retrieval with Improved Negative Contrast

In this paper we explore the effects of negative sampling in dual encoder models used to retrieve passages for automatic question answering. We explore four negative sampling strategies that complement the straightforward random sampling of negatives, typically used to train dual encoder models. Out of the four strategies, three are based on retrieval and one on heuristics. Our retrieval-based strategies are based on the semantic similarity and the lexical overlap between questions and passages. We train the dual encoder models in two stages: pre-training with synthetic data and fine tuning with domain-specific data. We apply negative sampling to both stages. The approach is evaluated in two passage retrieval tasks. Even though it is not evident that there is one single sampling strategy that works best in all the tasks, it is clear that our strategies contribute to improving the contrast between the response and all the other passages. Furthermore, mixing the negatives from different strategies achieve performance on par with the best performing strategy in all tasks. Our results establish a new state-of-the-art level of performance on two of the open-domain question answering datasets that we evaluated.

RRF102: Meeting the TREC-COVID Challenge with a 100+ Runs Ensemble

In this paper, we report the results of our participation in the TREC-COVID challenge. To meet the challenge of building a search engine for rapidly evolving biomedical collection, we propose a simple yet effective weighted hierarchical rank fusion approach, that ensembles together 102 runs from (a) lexical and semantic retrieval systems, (b) pre-trained and fine-tuned BERT rankers, and (c) relevance feedback runs. Our ablation studies demonstrate the contributions of each of these systems to the overall ensemble. The submitted ensemble runs achieved state-of-the-art performance in rounds 4 and 5 of the TREC-COVID challenge.

Zero-shot Neural Retrieval via Domain-targeted Synthetic Query Generation

Deep neural scoring models have recently been shown to improve ranking quality on a number of benchmarks (Guo et al., 2016; Daiet al., 2018; MacAvaney et al., 2019; Yanget al., 2019a). However, these methods rely on underlying ad-hoc retrieval systems to generate candidates for scoring, which are rarely neural themselves (Zamani et al., 2018). Re-cent work has shown that the performance of ad-hoc neural retrieval systems can be competitive with a number of baselines (Zamani et al.,2018), potentially leading the way to full end-to-end neural retrieval. A major road-block to the adoption of ad-hoc retrieval models is that they require large supervised training sets to surpass classic term-based techniques, which can be developed from raw corpora. Previous work shows weakly supervised data can yield competitive results, e.g., click data (Dehghaniet al., 2017; Borisov et al., 2016). Unfortunately for many domains, even weakly supervised data can be scarce. In this paper, we pro-pose an approach to zero-shot learning (Xianet al., 2018) for ad-hoc retrieval models that relies on synthetic query generation. Crucially, the query generation system is trained on general domain data, but is applied to documents in the targeted domain. This allows us to create arbitrarily large, yet noisy, query-document relevance pairs that are domain targeted. On a number of benchmarks, we show that this is an effective strategy for building neural retrieval models for specialised domains.

QURIOUS: Question Generation Pretraining for Text Generation

Recent trends in natural language processing using pretraining have shifted focus towards pretraining and fine-tuning approaches for text generation. Often the focus has been on task-agnostic approaches that generalize the language modeling objective. We propose question generation as a pretraining method, which better aligns with the text generation objectives. Our text generation models pretrained with this method are better at understanding the essence of the input and are better language models for the target task. When evaluated on two text generation tasks, abstractive summarization and answer-focused question generation, our models result in state-of-the-art performances in terms of automatic metrics. Human evaluators also found our summaries and generated questions to be more natural, concise and informative.

The Practical Challenges of Active Learning: Lessons Learned from Live Experimentation

We tested in a live setting the use of active learning for selecting text sentences for human annotations used in training a Thai segmentation machine learning model. In our study, two concurrent annotated samples were constructed, one through random sampling of sentences from a text corpus, and the other through model-based scoring and ranking of sentences from the same corpus. In the course of the experiment, we observed the effect of significant changes to the learning environment which are likely to occur in real-world learning tasks. We describe how our active learning strategy interacted with these events and discuss other practical challenges encountered in using active learning in the live setting.