A taxonomy of NLP interview trauma, question by question

What it does This repository is a curated index of natural-language-processing interview questions compiled from the authors’ own hiring experience in Chinese tech. It catalogs questions across information extraction—named entity recognition, relation extraction, event extraction—and pre-training algorithms such as TF-IDF, Word2vec, and FastText. The repository functions as a structured syllabus of what gets asked, with most detailed answers hosted externally.

The interesting bit The value is in the taxonomy. Each topic is decomposed into standard interview archetypes: definition, motivation, workflow, and pairwise comparison. There is even a dedicated “trick” section for Chinese NER that catalogs edge cases like nested entities, long spans, and annotation noise—precisely the gritty details candidates stumble over.

Key highlights

• Covers classical probabilistic graphical models (HMM, MEMM, CRF) alongside neural architectures (BiLSTM-CRF, IDCNN-CRF).
• Includes document-level relation extraction and event extraction sub-tasks such as trigger detection and argument identification.
• Enumerates practical engineering tricks: domain dictionary matching, handling data imbalance, and choosing between word-level and character-level embeddings for Chinese text.
• Maintains a WeChat discussion group for community updates.

Caveats

• The repo is largely a table of contents; detailed answers live behind links to a paid third-party platform (zsxq.com).
• All content is in Chinese with no English translation.
• No runnable code, notebooks, or self-contained explanations are provided.

Verdict Worth bookmarking if you are preparing for NLP engineering interviews in China and want a curated checklist of likely questions. Skip it if you need hands-on implementations or standalone tutorials.