Incorporating Unstructured Textual Knowledge Sources into Neural Dialogue Systems

By Ryan J. Lowe, Nissan Pow, Laurent Charlin, Joelle Pineau. 2015

How researchers see this paper

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

In Natural Language Processing

Citations in the corpus

Towards Exploiting Background Knowledge for Building Conversation Systems (2018, methodology): “However, most existing large scale datasets (Lowe et al., 2015b; Ritter et al., 2010; Serban et al., 2016) simply contain a sequence of utterances and responses without any explicit background knowledge associated with them.”
A Survey of Available Corpora For Building Data-Driven Dialogue Systems: The Journal Version (2015, methodology): “To leverage such data towards learning individual conversations, researchers have proposed methods to automatically estimate segmentations of conversations (Lowe et al., 2015a; Nio et al., 2014a).”
Relevance of Unsupervised Metrics in Task-Oriented Dialogue for Evaluating Natural Language Generation (2017, background): “Recently, deep-learning based approaches (Wen et al., 2015b; Sharma et al., 2017; Lowe et al., 2015; Serban et al., 2016) have shown promising results for dialogue response generation.”
Extending Neural Generative Conversational Model using External Knowledge Sources (2018, background): “Moving closer to unstructured domains, but still within task-driven conversations, (Lowe et al., 2015) proposes a way of retrieving relevant responses based on external knowledge sources.”
Augmenting End-to-End Dialogue Systems With Commonsense Knowledge (2018, unknown): “The model we propose in this work shares similarities with (Lowe et al. 2015a), which encoded unstructured textual knowledge with a recurrent neural network (RNN).”
A survey on empathetic dialogue systems (2020, methodology): “Since unstructured knowledge is usually in the form of plain texts, the suitable knowledge encoders for unstructured knowledge are mostly sequence encoders [96, 116] that were used to convert a sequence of words into a dense vector.”
Toxicity Detection can be Sensitive to the Conversational Context (2021, background): “Despite the wide use of context in other Natural Language Processing (NLP) tasks, such as dialogue systems (Lowe et al., 2015; Dušek and Jurčı́ček, 2016) and informational bias detection (van den Berg and Markert, 2020), very few researchers have focused on context-aware toxic language detection.”