Search

We investigated how users evaluate passage-length answers for non-factoid questions. We conduct a study where answers were presented to users, sometimes shown with automatic word highlighting. Users were tasked with evaluating answer quality, correctness, completeness, and conciseness. Words in the answer were also annotated, both explicitly through user mark up and implicitly through user gaze data obtained from eye-tracking. Our results show that the correctness of an answer strongly depends on its completeness, conciseness is less important. Analysis of the annotated words showed correct and incorrect answers were assessed differently. Automatic highlighting helped users to evaluate answers quicker while maintaining accuracy, particularly when highlighting was similar to annotation. We fine-tuned a BERT model on a non-factoid QA task to examine if the model attends to words similar to those annotated. Similarity was found, consequently, we propose a method to exploit the BERT attention map to generate suggestions that simulate eye gaze during user evaluation.

Nowadays natural language user interfaces, such as chatbots and conversational agents, are very common. A desirable trait of such applications is a sense of humor. It is, therefore, important to be able to measure quality of humorous responses. However, humor evaluation is hard since humor is highly subjective. To address this problem, we conducted an online evaluation of 30 dialog jokes from different sources by almost 300 participants -- volunteers and Mechanical Turk workers. We collected joke ratings along with participants» age, gender, and language proficiency. Results show that demographics and joke topics can partly explain variation in humor judgments. We expect that these insights will aid humor evaluation and interpretation. The findings can also be of interest for humor generation methods in conversational systems.

The paper describes a work in progress on humorous response generation for short-text conversation using information retrieval approach. We gathered a large collection of funny tweets and implemented three baseline retrieval models: BM25, the query term reweighting model based on syntactic parsing and named entity recognition, and the doc2vec similarity model. We evaluated these models in two ways: in situ on a popular community question answering platform and in laboratory settings. The approach proved to be promising: even simple search techniques demonstrated satisfactory performance. The collection, test questions, evaluation protocol, and assessors’ judgments create a ground for future research towards more sophisticated models.