Internet-Augmented Dialogue Generation

By Mojtaba Komeili, Kurt Shuster, Jason Aaron Edward Weston. 2021

325 citations recorded by the source corpus.

The largest store of continually updating knowledge on our planet can be accessed via internet search. In this work we study giving access to this information to conversational agents. Large language models, even though they store an impressive amount of knowledge within their weights, are known to hallucinate facts when generating dialogue (Shuster et al., 2021); moreover, those facts are frozen in time at the point of model training. In contrast, we propose an approach that learns to generate an internet search query based on the context, and then conditions on the search results to finally generate a response, a method that can employ up-to-the-minute relevant information. We train and evaluate such models on a newly collected dataset of human-human conversations whereby one of the speakers is given access to internet search during knowledgedriven discussions in order to ground their responses. We find that search-query based access of the internet in conversation provides superior performance compared to existing approaches that either use no augmentation or FAISS-based retrieval (Lewis et al., 2020b).

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

Toolformer: Language Models Can Teach Themselves to Use Tools (2023, methodology): “Tool Use Several approaches aim to equip LMs with the ability to use external tools such as search engines (Komeili et al., 2022; Thoppilan et al., 2022; Lazaridou et al., 2022; Shuster et al., 2022; Yao et al., 2022), web browsers (Nakano et al.”
LaMDA: Language Models for Dialog Applications (2022, positive): “[45] compare different types of pre-trained models and retrieval methods, and reach a similar conclusion that augmenting language models with a search engine provides more factually grounded responses.”
Improving language models by retrieving from trillions of tokens (2021, background): “In the open-domain dialogue setting, BlenderBot 2.0 (Komeili et al., 2021) learns to issue textual internet queries, outperforming dense retrieval methods when evaluated on a task measuring how close model responses are to those of humans.”
OPT: Open Pre-trained Transformer Language Models (2022, methodology): “3.2 Dialogue Given that LLMs are known to be an integral component of modern dialogue models (Adiwardana et al., 2020; Roller et al., 2021; Thoppilan et al., 2022; Rae et al., 2021; Chowdhery et al., 2022), we additionally evaluate OPT-175B on several open source dialogue datasets.”
Sustainable AI: Environmental Implications, Challenges and Opportunities (2021, background): “There are four major phases: manufacturing , transport , product use , and recycling 2 .”
A General Language Assistant as a Laboratory for Alignment (2021, methodology): “This might include calibration training on factual claims and claims about the internal state of the model, and the use of search [KSW21] to augment accuracy.”
Autoregressive Search Engines: Generating Substrings as Document Identifiers (2022, background): “…Google” Recently, language models have been shown to be able to directly generate search queries for modern web search engines either with ﬁnetuning on demonstrations (Komeili et al., 2021; Shuster et al., 2022) and human preferences (Nakano et al., 2021) or via prompting (Lazaridou et al., 2022).”
Language Models that Seek for Knowledge: Modular Search & Generation for Dialogue and Prompt Completion (2022, methodology): “studied in separate work that augments generations with internet search (Komeili et al., 2021).”
Internet-augmented language models through few-shot prompting for open-domain question answering (2022, background): “We expect that “learning to search” approaches could boost performance of the overall system [22, 10, 11], where complex queries could be decomposed into simpler sub-queries, akin to approaches in question decomposition [40, 41].”
The Web Is Your Oyster - Knowledge-Intensive NLP against a Very Large Web Corpus (2021, methodology): “A CCNet snapshot has been considered as a knowledge source in dialog research by Komeili et al. (2021), where authors use it together with a Wikipedia snapshot.”
When Life Gives You Lemons, Make Cherryade: Converting Feedback from Bad Responses into Good Labels (2022, methodology): “These tasks include the original reply correction task, the task with context as the input and free-form textual feedback as the target, and the dialogue task of Wizard of Internet (Komeili et al., 2021).”
BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage (2022, methodology): “We observe the best performance from BB3-175B across almost all metrics, including Good Response % and overall Rating compared to BB1, BB2, SeeKeR and variants of OPT-175B.”
WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents (2022, background): “These include: 1) generation of good search queries [22, 59] and reformulation [37, 51], 2) strategic exploration for navigating through the website [55, 56, 29], 3) robust language understanding for textual state and action spaces [3, 7, 17, 44], and 4) long-term”
Teaching language models to support answers with verified quotes (2022, background): “task “self-supported question-answering” (SQA), and intend it as a sub-task that can be embedded into other generative language modelling tasks such as open-ended dialogue or debate (Askell et al., 2021; Irving et al., 2018; Komeili et al., 2021; Rae et al., 2021; Thoppilan et al., 2022).”
Towards Continual Knowledge Learning of Language Models (2021, background): “0 (Xu et al., 2021; Komeili et al., 2021), is also one of the explicit methods that search the internet for recent knowledge and saves recent conversations in external long-term memory.”
Generative Spoken Dialogue Language Modeling (2022, background): “More recent research focused on utilizing retrieval augmented generation methods (Lewis et al., 2020) for long-context, multi-session conversations (Xu et al., 2021), and grounding responses on fresh information from the internet (Komeili et al., 2021; Shuster et al., 2022).”
Reason first, then respond: Modular Generation for Knowledge-infused Dialogue (2021, methodology): “Recent advances in neural retrieval models have made some inroads into this problem (Lee et al., 2019; Lewis et al., 2020b; Shuster et al., 2021a; Komeili et al., 2021) on both the dialogue context and by learning to retrieve documents containing relevant knowledge.”
Taxonomy of Risks posed by Language Models (2022, background): “In additionto increasing representation of marginalised groups and addressing specific needs in downstream applications [159], approaches to expanding or up-dating LMs in real-time also include modes that continue to learn online, as changes occur to a training corpus [8, 43, 106, 112, 179].”