| { | |
| "paper_id": "2021", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T10:54:26.367672Z" | |
| }, | |
| "title": "", | |
| "authors": [], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "", | |
| "pdf_parse": { | |
| "paper_id": "2021", | |
| "_pdf_hash": "", | |
| "abstract": [], | |
| "body_text": [ | |
| { | |
| "text": "Much progress has been made recently to improve conversation systems and chatbots using neural based deep models. While the architectures of conversation models evolve quickly, evaluation techniques over the years remain unchanged. As a matter of fact, evaluating and assessing conversation models has been a decades long challenge. One part of the difficulty comes from the fact that human dialogues exist in a variety of forms. Existing approaches that compare generated conversations from a neural model with ground truth can easily incur large biases due to the lack of diversity. Another part of the difficulty comes from the interactive nature of conversations, which often requires an agent (usually a real person), to conduct the assessments. Human evaluations, on the other hand, can introduce a large amount of variance and are often impractical on a large scale. A third dimension of evaluation has to do with the fairness and reliability of the models, which has become an increasingly important issue for commercial use of neural based systems. Finally, transferability, transparency and ethical issues in evaluations are among some of the other important topics to explore. ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": {}, | |
| "ref_entries": { | |
| "TABREF1": { | |
| "type_str": "table", | |
| "html": null, | |
| "num": null, | |
| "content": "<table/>", | |
| "text": "Counterfactual Matters: Intrinsic Probing For Dialogue State Tracking Yi Huang, Junlan Feng, Xiaoting Wu and Xiaoyu Du . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 GCDF1: A Goal-and Context-Driven F-Score for Evaluating User Models Alexandru Coca, Bo-Hsiang Tseng and Bill Byrne . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 A Comprehensive Assessment of Dialog Evaluation Metrics Yi-Ting Yeh, Maxine Eskenazi and Shikib Mehri . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15" | |
| } | |
| } | |
| } | |
| } |