Explainable NLP Across Seven Domains: A Survey of Methods and Evaluation
A recent study published on arXiv examines the implementation of explainable Natural Language Processing (XNLP) across seven sectors, including healthcare, finance, and customer service. Researchers stress the necessity for transparency in environments where AI models like GPT-4o and BERT impact decision-making. The study evaluates the explanation types needed and the assessment methods used in each sector, revealing varying needs concerning explanation strategies based on their effectiveness and costs. Additionally, the paper suggests a new evaluation framework that distinguishes between universal criteria and specific industry needs, highlighting the rising significance of AI interpretability in essential fields.
Key facts
- The survey covers seven domains: medicine, finance, systematic reviews, customer relationship management, chatbots, social and behavioral science, and human resources.
- Models such as GPT-4o, Gemini, and BERT are mentioned as examples of NLP models used in critical sectors.
- The review compares explanation method families on scope, evidence of faithfulness, and computational cost.
- A two-tier evaluation protocol is proposed, separating shared technical metrics from domain-specific ones.
- The paper is available on arXiv with identifier 2502.00837.
- The announcement type is 'replace-cross'.
- The survey focuses on explainable NLP (XNLP) as actually deployed in practice.
- The need for transparency arises from the black-box nature of these models.
Entities
Institutions
- arXiv