RDI_Team at SemEval-2016 Task 3: RDI Unsupervised Framework for Text Ranking

Proceedings of SemEval-2016, pages 822–827, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics RDI_Team at SemEval-2016 Task 3: RDI Unsupervised Framework for Text Ranking Ahmed Magooda1,§, Amr M. Sayed2,§, Ashraf Y. Mahgoub1, §, Hany Ahmed1, §, Mohsen Rashwan1, Hazem Raafat3, Eslam Kamal4, Ahmad A. Al Sallab5 1 RDI, Cairo, Egypt. 2Computer Science Department, Faculty of Computers and Information, Cairo University, Egypt. 3Computer Science Department, Kuwait University, Kuwait 4Microsoft Research, Cairo, Egypt. 5Valeo Inter-branch Automotive Software, Cairo, Egypt. § These four authors contributed equally to this work. ahmed.ezzat.gawad@gmail.com, amr@fci-cu.edu.eg, ashraf.youssef.mahgoub@gmail.com, hanyahmed@aucegypt.edu, mrashwan@rdi-eg.com, hazem@cs.ku.edu.kw, eskam@microsoft.com, ahmad.el-sallab@valeo.com Abstract Ranking is an important task in the field of in- formation retrieval. Ranking may be used in different modules in natural language pro- cessing such as search engines. In this paper, we introduce a competitive ranking system which combines three different modules. The system participated in SemEval 2016 question ranking task for the Arabic language. The task is a ranking task that targets reordering results retrieved from search engine. Results reorder- ing is done based on relevancy between search result and the original query issued. The data provided in the competition is in the form of question (query) and 30 question answer pairs retrieved from search engine. For each ques- tion retrieved from the search engine the sys- tem generates a relevancy score that is to be used for ranking. The proposed system came in the third position in the Competition. Since the majority of modules are unsupervised the un- supervised naming was used. 1 Introduction This paper describes RDI System, the system par- ticipated in SemEval 2016 Community question ranking shared task for the Arabic language (Nakov et al., 2016). The task is a ranking task for medical community questions, whenever a new user wants to submit a new question, the system should retrieve similar questions that are previously answered. Questions retrieved by the search engine are then to be ordered by its relevancy to query question in or- der to reduce redundancy. The task proceeds as follows, for each query question 30 previously answered question answer pairs are retrieved through a search engine. The aim of the task is to reorder the retrieved 30 question an- swer pairs based on its relevancy to the query ques- tion. As query expansion and results re-ranking are two major paradigms for search engine enhance- ments, (Mahgoub et al., 2014) used Google syno- nyms and Arabic Wikipedia to expand search que- ries, on the other hand (Abouenour et al., 2010) pro- posed a system that uses Arabic WordNet to expand queries. In this paper we are going to propose a 822 ranking system which consists of three modules merged with each other to produce a relevancy score. The proposed system managed to achieve the third position in the SemEval 2016 Community question ranking shared task for the Arabic. The 3 modules presented in this system are: 1- TF-IDF based module 2- Language model (LM) based module 3- Wikipedia-based module The paper is organized as …

Content truncated to 500 words