CYUT at IJCNLP-2017 Task 3: System Report for Review Opinion Diversification
Review Opinion Diversification (RevOpiD) 2017 is a shared task which is held in International Joint Conference on Natural Language Processing (IJCNLP). The shared task aims at selecting top-k reviews, as a summary, from a set of re-views. There are three subtasks in RevOpiD: helpfulness ranking, rep-resentativeness ranking, and ex-haustive coverage ranking. This year, our team submitted runs by three models. We focus on ranking reviews based on the helpfulness of the reviews. In the first two models, we use linear regression with two different loss functions. First one is least squares, and second one is cross entropy. The third run is a random baseline. For both k=5 and k=10, our second model gets the best scores in the official evaluation metrics.
Code (0)
등록된 구현이 없습니다.
Tasks
regressionMethods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
CYUT-III System at Chinese Grammatical Error Diagnosis Task
This paper describe the CYUT-III system on grammar error detection in the 2016 NLP-TEA Chinese Grammar Error Detection shared task CGED. In this task a system has to detect four types of errors, in-cluding redundant word…
CYUT Team Chinese Grammatical Error Diagnosis System Report in NLPTEA-2020 CGED Shared Task
This paper reports our Chinese Grammatical Error Diagnosis system in the NLPTEA-2020 CGED shared task. In 2020, we sent two runs with two approaches. The first one is a combination of conditional random fields (CRF) and …
Deep LearningCYUT-III Team Chinese Grammatical Error Diagnosis System Report in NLPTEA-2018 CGED Shared Task
This paper reports how we build a Chinese Grammatical Error Diagnosis system in the NLPTEA-2018 CGED shared task. In 2018, we sent three runs with three different approaches. The first one is a pattern-based approach by …
Language ModelingLanguage ModellingIJCNLP-2017 Task 3: Review Opinion Diversification (RevOpiD-2017)
Unlike Entity Disambiguation in web search results, Opinion Disambiguation is a relatively unexplored topic. RevOpiD shared task at IJCNLP-2107 aimed to attract attention towards this research problem. In this paper, we …
Document RankingDocument SummarizationEntity DisambiguationInformation Retrieval+1JUNLP at IJCNLP-2017 Task 3: A Rank Prediction Model for Review Opinion Diversification
IJCNLP-17 Review Opinion Diversification (RevOpiD-2017) task has been designed for ranking the top-k reviews of a product from a set of reviews, which assists in identifying a summarized output to express the opinion of …