Research‎ > ‎

iREBOOK: Intelligent social REtrieval for BOOKs

Social Book Search and Recommendation

   Social book search is a suggestion (search and recommendation) system to support users in searching and navigating professional metadata and user-generated content from social media.

   Effective book search has been discussed for decades and is still future-proof in areas as diverse as computer science, informatics, e-commerce and even culture and arts. A variety of social information contents (e.g, ratings, tags and reviews) emerge with the huge number of books on the Web, but how they are utilized for searching and finding books is seldom investigated. Here we develop an Integrated Search And Recommendation Technology (IsART), which breaks new ground by providing a generic framework for searching books with rich social information. ISART comprises a search engine to rank books with book contents and professional metadata, a Generalized Content-based Filtering model to thereafter rerank books with user-generated social contents, and a learning-to-rank technique to finally combine a wide range of diverse reranking results. Experiments show that this technology permits embedding social information to promote book search effectiveness, and ISART, by making use of it, has the best performance on CLEF/INEX Social Book Search Evaluation datasets of all 4 years (from 2011 to 2014), compared with some other state-ofthe-art methods.  Moreover, our proposed system won INEX 2014 Social Book Search Suggestion Task and all our 6 runs ranked top 7 in the competition (In the following figure, the runs with the prefix "USTB-" are our technologies).



Biomedical Semantic Information Retrieval and QA

(1) A Generic Retrieval System for Biomedical Literatures

   In document retrieval, we build a generic retrieval model based on the sequential dependence model, Word Embedding and Ranking Model. In addition, from the view of the special significance of titles (Title Significance Validation), we re-rank the top-K results by counting the meaningful nouns in the titles. The top-K documents are split into sentences and indexed for snippets retrieval. The similar models of document retrieval are applied for this part. To extract the biomedical concepts and corresponding RDF triplets, we use concept recognition tools MetaMap and Banner. Statistics indicate that our systems outperform other results. Moreover, Our technology won several batches of 2015 BioASQ Challenge Task 3b: Biomedical Semantic QA.

(2) Mining Semantic Relationships between Snippets for Biomedical Question Answering

   Biomedical question answering is a hot and challenging topic in AI and NLP as it helps to analyze multiple, large and fast-growing biomedical knowledge sources. Most of researchers try to construct a knowledge base to address the problem but suffer from the lack of specialty and workload. Here, we present an approach taken to mine the semantic relationships between pairs of snippets in order to find the most relevant snippets with the
questions from massive literatures. Firstly, we utilize the whole corpus to reorganize the sentences from literatures into candidate snippets. Afterwards, word embedding and an improved SDM model are relatively used to reconstruct the questions in order to emphasize the effect of synonyms and word sequences. Finally, a probability model and content-based filtering are combined into a hybrid model to measure the relevance between questions and corresponding candidate snippets. A higher relevance indicates a greater possibility to be a good answer. In addition, a two-stage biomedical question
answering system is designed based on the proposed approach and evaluated on three-year BioASQ challenges datasets. Statistics show that our system outperforms than others extensively.



[1] Xu-Cheng Yin, Bo-Wen Zhang, Xiao-Ping Cui, Jiao Qu, Bin Geng, Fang Zhou, Li Song, and Hong-Wei Hao, "IsART: A Generic Framework for Searching Books with Social Informaiton," PLoS ONE, vol. 11, no. 2, pp. e0148479, 2016. (Our proposed technology won the top place of CLEFT/INEX 2014/2015 Social Book Search Suggestion Task.) <Paper Link>

[2] Zhijuan Zhang, Tiantian Liu, Bo-Wen Zhang, Yan Li, Chun Hua Zhao, Shao-Hui Feng, Xu-Cheng Yin, and Fang Zhou, A generic retrieval system for biomedical literatures: USTB at BioASQ2015 Question Answering Task, CLEF (Working Notes), 2015. (Our technology won several batches of 2015 BioASQ Challenge Task 3b: Biomedical Semantic QA.)


[3] Chunhua Zhao, Fang Zhou, Bo-Wen Zhang, Xu-Cheng Yin, Ming Hao, Zhijuan Zhang, and Tiantian Liu, USTB at Social Book Search 2015 Suggestion Task: Metadata Expansion and Reranking, CLEF (Working Notes), 2015.

[4] Bo-Wen Zhang, Xu-Cheng Yin, Xiao-Ping Cui, Jiao Qu, Bin Geng, Fang Zhou, Li Song, and Hong-Wei Hao, “Social book search reranking with generalized content-based filtering”, ACM International Conference on Information and Knowledge Management (CIKM'14), accepted, 2014. (Our proposed technology (USTB) won the 1st place of INEX 2014 Social Book Search Suggestion Task.)

[5] Bo-Wen Zhang, Xu-Cheng Yin, Xiao-Ping Cui, Jiao Qu, Bin Geng, Fang Zhou, Hong-Wei Hao, "USTB at INEX2014: Social Book Search Track," CLEF (Working Notes), 2014.

Funding and Sponsoring

|Powered By Google Sites