What it does

This is a hand-maintained bibliography of deep-learning papers for dialog systems, split between task-oriented bots (booking flights, slot-filling) and open-domain chat bots. Each entry links to arXiv or conference PDFs. The maintainer, Tiancheng Zhao at CMU’s LTI, invites pull requests for missing work.

The interesting bit

The taxonomy itself is the artifact. The README predates the transformer era—everything here is RNNs, seq2seq, memory networks, and early RL—so it reads like an archaeological slice of what researchers thought mattered before attention became the only game in town. The “Inteprebility” typo in the Chat Bots section is still there.

Key highlights

• Covers ~80+ papers from 2015–2018, with sections on user simulators, adversarial training, and cross-domain transfer
• Explicitly splits task bots (goal-oriented, reinforcement learning-heavy) from chat bots (retrieval, diversity, persona modeling)
• Includes now-classic early work: Vinyals’s neural conversational model, Li’s adversarial dialogue generation, Wen’s end-to-end task-oriented systems
• Curated by a dialogue researcher at CMU, not a generic aggregator
• Still accepting issues and PRs for missing papers

Caveats

• Effectively frozen in time: no transformer-era papers (no BERT, no GPT, no instruction tuning)
• README is truncated in the source; full list may be longer than visible
• Some links point to arXiv preprints that may have later journal versions

Verdict

Worth a skim if you’re building a historical literature review or teaching a course on dialog systems. Skip it if you need current SOTA—this is a pre-2018 time capsule, not a living survey.