Spoken Content Retrieval Using Distance Combination and Spoken Term Detection Using Hash Function for NTCIR10 SpokenDoc2 Task
Satoru Tsuge, Ken Ichikawa, Norihide Kitaoka, Kazuya Takeda, Kenji Kita · NTCIR · 2013
In this paper we describe a spoken content retrieval (SCR) and a spoken term detection (STD) which were used in the 2nd round of the IR (Information Retrieval) for Spoken Documents (SpokenDoc2) task. Our SCR method maps the target documents into multiple vector spaces, which include a word-based vector space for word-based speech recognition results and a syllable-based vector space for syllablebased speech recognition results. The syllable-based space is spanned by axes extracted using latent semantic indexing (LSI). We also apply query expansion and morpheme weighting to the word-based space. Finally, the distance between the query and the documents in each vector space are combined and ranked for retrieving the documents. On the other hand, our STD method extracts sub-sequences from the target documents and converts them into bit sequences using the hash function. The query is also converted into a bit sequence in the same way. Candidates are detected by calculating the hamming distance between the bit sequence of the query and that of the target documents. Then, our method calculates the distances between the query and the candidates using DP (Dynamic Programming) matching. To evaluate the proposed methods, we conducted spoken document retrieval experiments using the SpokenDoc task from the NTCIR-9 meeting. Using these experimental results to set our parameters, we submitted the results for the SpokenDoc2 task at NTCIR-10.