Recently,many local review websites such as Yelp are emerging, which have greatly facilitated people�s daily life such as cuisine hunting.\r\nHowever they failed to meet travelers� demands because travelers are more concerned about a city�s local specialties instead of\r\nthe city�s high ranked restaurants. To solve this problem, this paper presents a local specialty mining algorithm, which utilizes both\r\nthe structured data from local review websites and the unstructured user-generated content (UGC) from community Q&A websites,\r\nand travelogues. The proposed algorithm extracts dish names from local review data to build a document for each city, and\r\napplies tfidf weighting algorithm on these documents to rank dishes. Dish-city correlations are calculated from unstructured UGC,\r\nand combined with the tfidf ranking score to discover local specialties. Finally, duplicates in the local specialty mining results are\r\nmerged. A recommendation service is built to present local specialties to travelers, along with specialties� associated restaurants,\r\nQ&A threads, and travelogues. Experiments on a large data set show that the proposed algorithm can achieve a good performance,\r\nand compared to using local review data alone, leveraging unstructured UGC can boost the mining performance a lot, especially\r\nin large cities.
Loading....