At NTCIR-2, RICOH submitted eight runs for the Japanese IR task. Of the eight runs, four runs use the title eld only and the other four use the description eld only. RICOH's system is built on our English text retrieval system and augmented to handle Japanese text. The system features (1) hybrid retrieval using a combination of n-gram indexing and wordbased document ranking; (2) word-based and ...

How to accurately interpret user click behaviour in search log is a key but challenging problem for search relevance. In this paper, we describe our solution to the relevance prediction challenge which achieves the first place among eligible teams. There are three stages in our solution: feature generation, feature augmentation and learning a ranking function. In the first stage, we extract fea...

The verification of final termination for counter systems is undecidable. For non flattable counter systems, the verification of this type of property is generally based on the exhibition of a ranking function. Proving the existence of a ranking function for general counter systems is also undecidable. We provide a framework in which the verification whether a given function is a ranking functi...

Word Ordering Errors (WOEs) are the most frequent type of grammatical errors at sentence level for non-native Chinese language learners. Learners taking Chinese as a foreign language often place character(s) in the wrong places in sentences, and that results in wrong word(s) or ungrammatical sentences. Besides, there are no clear word boundaries in Chinese sentences. That makes WOEs detection a...

In this paper, we report the experiments we conducted for our participation to the TREC 2012 Web Track. We experimented a brand new system that models the latent concepts underlying a query. We use Latent Dirichlet Allocation (LDA), a generative probabilistic topic model, to exhibit highly-specific query-related topics from pseudo-relevant feedback documents. We define these topics as the laten...

Method: Building a document ranking system involves two key decisions: choosing a retrieval model, and choosing a suitable index representation. The former determines the effectiveness of the system, the latter the efficiency; and each of them affects the other. The impact-based document ranking mechanism described by Anh and Moffat [2] was chosen for our system because of its balance between e...

Passage retrieval and pseudo relevance feedback/query expansion have been reported as two effective means for improving document retrieval in literature. Relevance models, while improving retrieval in most cases, hurts performance on some heterogeneous collections. Previous research has shown that combining passage-level evidence with pseudo relevance feedback brings added benefits. In this pap...

This work concerns learning probabilistic models for ranking data in a heteroge-neous population. The specific problem we study is learning the parameters of aMallows Mixture Model. Despite being widely studied, current heuristics for thisproblem do not have theoretical guarantees and can get stuck in bad local optima.We present the first polynomial time algorithm which provably...

We present a model for the inclusion of semantic role annotations in the framework of confidence estimation for machine translation. The model has several interesting properties, most notably: 1) it only requires a linguistic processor on the (generally well-formed) source side of the translation; 2) it does not directly rely on properties of the translation model (hence, it can be applied beyo...