Document Ranking Methods For Supporting Implicit Temporal Queries In Information Retrieval

The popularity of the Internet contributes to the explosive growth of information resources.It increases the difficulty of finding useful information although it provides users with more opportunities.Therefore,how to use a search engine to find the documents that meet users' requirements has become an important challenge.With the increasing number of web pages and queries that contain temporal information,quite a few scholars and researchers begin to investigate Temporal Information Retrieval(TIR),studying how to use effective time information processing methods to extract temporal information,analyze query temporal intention and establish time-related ranking models to improve the quality of retrieval.There are two kinds of temporal queries in temporal information retrieval.One kind of queries contains explicit time expression,which is called explicit temporal query;while the other does not provide a clear time criterion,but the relevant results tend to occur in a particular time interval,called implicit temporal query.Previous study finds that more than 7% of the queries in the Internet contain implicit temporal intention,and about 1.5% of the queries contain explicit temporal expression.It can be found that implicit temporal queries occupy a larger proportion,and there are more research work need to be done.In this dissertation,we mainly study how to analyze the temporal intention of implicit temporal queries to optimize retrieval performance,the main work is as follows:(1)For implicit temporal queries,we propose an approach that considers either DBpedia or top-k documents in the resultant list to analyze their time intention: if the query is about a major event or celebrities in history,then the temporal intent of the query can be obtained via querying DBpedia;otherwise,we use the query to make a search in the search system and look at the top-k documents returned,the time expression frequently appeared in those top k results is regarded as a time point of interest to the query.(2)On the basis of the language modeling,we propose a document ranking method for implicit temporal queries,the temporal score of each document is the probability of generating the query while considering the temporal uncertainty,and then re-ranking documents according to the linear combination of temporal relevance score and content relevance score.(3)In order to evaluate the performance of the methods proposed in this paper,we use the data set in Temporal Information Access(Temporalia)task of the NTCIR-11 conference for experiment.Firstly,we compare our method of analyzing query temporal intention with several existing methods.Experiment results show that the analysis of query temporal intention before calculating relevance score is meaningful,and our method of combining DBpedia and top-k documents in the resultant list does well in analyzing query temporal intention.After the query temporal intention is obtained,we compare the ranking method proposed in this paper with several existing ranking models.Results show that most of the metric values in the ranking models which consider temporal relevance are higher than the initial ranking result that only consider content relevance.This can be seen as considering temporal relevance in the retrieval model is helpful for improving the quality of retrieval.Compared with other ranking methods,our language model-based ranking method is better than all of them.