Using Zero-Resource Spoken Term Discovery for Ranked Retrieval

Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL, pages 588–597, Denver, Colorado, May 31 – June 5, 2015. c⃝2015 Association for Computational Linguistics Using Zero-Resource Spoken Term Discovery for Ranked Retrieval Jerome White New York University Abu Dhabi, UAE jerome.white@nyu.edu Douglas W. Oard University of Maryland College Park, MD USA oard@umd.edu Jiaul Paik University of Maryland College Park, MD USA jiaul@umd.edu Rashmi Sankepally University of Maryland College Park, MD USA rashmi@umd.edu Aren Jansen John Hopkins HLTCOE Baltimore, MD USA aren@jhu.edu Abstract Research on ranked retrieval of spoken con- tent has assumed the existence of some auto- mated (word or phonetic) transcription. Re- cently, however, methods have been demon- strated for matching spoken terms to spoken content without the need for language-tuned transcription. This paper describes the first application of such techniques to ranked re- trieval, evaluated using a newly created test collection. Both the queries and the collection to be searched are based on Gujarati produced naturally by native speakers; relevance assess- ment was performed by other native speak- ers of Gujarati. Ranked retrieval is based on fast acoustic matching that identifies a deeply nested set of matching speech regions, cou- pled with ways of combining evidence from those matching regions. Results indicate that the resulting ranked lists may be useful for some practical similarity-based ranking tasks. 1 Introduction Despite new methods of interaction, speech contin- ues to be a dominant modality for information ex- change, particularly among the half of the world’s almost five billion mobile phone users who currently lack text-based Internet access. Recording speech poses no particular problems, but retrieval of spoken content using spoken queries is presently available only for the approximately two dozen languages in which there is an established path to market; En- glish, German, or Chinese, for example. However, many of the mobile-only users who could benefit most from such systems speak only one of the sev- eral hundred other languages that each have at least a million speakers;1 Balochi, Mossi or Quechua, for example. Addressing this challenge in a scalable manner requires an integration of speech processing and information retrieval techniques that can be ef- fectively and affordably extended to a large number of languages. To this end, the experiments in this paper were conducted in a conventional ranked retrieval frame- work consisting of spoken queries, spoken “doc- uments” (responses, hereafter), graded relevance judgments, and standard evaluation measures. As with other information retrieval tasks, there is an el- ement of uncertainty in our best representations of what was said. Our focus on speech processing tech- niques that are language-agnostic creates the poten- tial for explosive growth in the uncertainty that our search techniques must accommodate. The design and evaluation of such techniques is therefore the central focus of the work explored in this paper. Our results are both heartening and disconcerting. On the positive side, useful responses can often be found. As one measure of success, we show that a Mean Reciprocal Rank near 0.5 …

Content truncated to 500 words