Search

Spoken Document Retrieval (SDR) and Spoken Term Detection (STD) have been two of the most intensively investigated topics in spoken document processing research according to the establishment of the SDR and STD test collections by the Text REtrieval Conference (TREC) and NIST. Because Japanese spoken document processing researchers also requires such test collections for SDR and STD, we have established a working group to develop these collections in Special Interest Group -Spoken Language Processing (SIG-SLP) of the Information Processing Society of Japan. The working group has constructed and made available a test collection for SDR, and is now constructing new test collections for STD that will be open to researchers. The present paper introduces the policies, outline, and schedule of the new test collections. Then, the new test collections are compared with the NIST STD test collections. Index Terms: spoken term detection, test collection

This paper describes the application of subword units in an effort of improving open-vocabulary spoken document retrieval performance in the case of highly corrupted recognition output. This paper presents the developed open-vocabulary spoken document retrieval system including the newly proposed subphonetic segment unit and combining multilayer subword units. Our experiments on Japanese spoken documents show that using the proposed subphonetic segment unit can improve retrieval performance, high precision and recall, and a combination of multilayer subword units is also effective.

We propose an approach for detecting topic boundaries and keywords in arbitrary speech, with neither recognition nor prosodic processing, aiming at guide access to the content of recorded raw speech. This approach is based on the general tendency that frequently repeated phrases/words in speech are characteristic of topics in discourse. So it uses pairs of phonetically similar segments (PPSSs) of speech to represent topics in speech. This approach has the advantage of being domain and language independent and robust against variations in the speaker and background noise, as it needs neither a language nor acoustic model in advance. Experiments using simulated dialogues confirmed the good performance of this approach. We also propose Incremental Reference Interval Free Continuous Dynamic Programming (IRIFCDP) as an algorithm for detecting PPSSs in speech for the above method. IRIFCDP can detect PPSSs efficiently in synchronization with the speech, so it is suitable for handling long speech samples.