http://air.sciedupress.com Artificial Intelligence Research 2016, Vol. 5, No. 2 ORIGINAL RESEARCH Cross-language phoneme mapping for phonetic search keyword spotting using multiple source languages Ella Tetariy 1 , Yossi Bar-Yosef 2 , Michal Gishri *1 , Ruthi Alon-Lavi 1,2 , Vered Aharonson 1 , Irit Opher 1,2 , Ami Moyal 1 1 Afeka Academic College of Engineering, Afeka Center for Language Processing, Tel Aviv, Israel 2 NICE Systems Ltd., Ra’anana, Israel Received: November 22, 2015 Accepted: January 12, 2016 Online Published: February 3, 2016 DOI: 10.5430/air.v5n2p24 URL: http://dx.doi.org/10.5430/air.v5n2p24 ABSTRACT Performing Phonetic Search Keyword Spotting (PS KWS) in new languages when language resources are scarce is an interesting and challenging task. In a previous paper we reported a methodology that enabled PS KWS under these conditions utilizing cross-language phoneme mappings from another sufficiently resourced and well-trained source language. We performed phoneme recognition in the new target language with the acoustic model of the source language. The keyword search was performed over a phoneme lattice of the target language phonemes following a mapping from one language to the other. In the present work we extend this method and its capabilities by mapping two source language phoneme sets into one target language set and performing a combined lattice search. Testing the technique on English and Arabic as source languages yielded a 50% Detection Rate (DR) and a False Alarm Rate (FAR - measured in number of false alarms per hour per keyword) of 2 when Spanish was the target language, a DR of 36% and FAR of 4 when Dari was the target language and a DR of 35% and FAR of 6 with Farsi as the target language. These results indicate that combining two source languages is better than using a single language since the acoustic space is better represented. Searching in a combined lattice while employing adequate phoneme transformations significantly improves performance. Such a system can be used as an initial version of a PS KWS system in a new language when sufficient language resources are not available. Key Words: Cross-language phoneme mapping, Keyword spotting, Spoken term detection, Phonetic search, Multi-lingual Keyword Spotting, Parallel lattice search 1. I NTRODUCTION There is a growing demand for Keyword Spotting (KWS) systems that enable specific words to be identified out of a stream of continuous speech. [1] This demand is mani- fested in international evaluation efforts that led to signif- icant advances in KWS research in recent years. [2, 3] The applications based on KWS are many and diverse: from call classification and speech database search for call centers and security-intelligence organizations, to multi-media search applications in the internet and enterprise markets. These ap- plications are relatively easy to develop for languages which are rich in Language Resources (LRs). However, when a new language is concerned, the process of collecting LRs, such as large speech and text databases for training acoustic and language models and a large vocabulary pronunciation lexicon, [4–7] is both long and costly. Our previous works [8, 9] reviewed existing technologies that utilize cross-language phoneme mapping to enable the use * Correspondence: Michal Gishri; Email: michalg@afeka.ac.il; Address: Afeka Academic College of Engineering, Afeka Center for Language Processing, 38 Mivtsa Kadesh St. Tel-Aviv 6998812, Israel. 24 ISSN 1927-6974 E-ISSN 1927-6982