ICTNET at Context Suggestion Track TREC 2014

This year we did not use ClueWeb12 or ClueWeb12-B,while we solve this issue based on data we crawled from openweb.Firstly, we use external structured resource –Google Place API[1] to find all of the possible candidate places in the distance of 5 hours ' drive. Secondly, we use Yandex[2] to find a description of each place because we get URL of the corresponding place. Then, we classifythe descriptions of all places. Finally, we ranked the pages based on users ’ preferences. 1 Data Preparation Google Place API is used to generate candidate places for each query. We get pages from Google Place API and get descriptionsof them from Yandex. It is a challengeto get enough candidate places due to the visiting limitation of API. Another challenge is that Google Place API does not return enough candidate places for every specific geographical position. So the places we got from the API of current round are used as seeds to get more nearby places in the next round. After several rounds, we remove the duplicate places and get the candidate set. 2 Query Generation Queries are the basics of our scheme. We want to build good queries for the search engine to return us with good suggestions.Each query consists of base part and keyword part.The base part i