Improving the Arabic Pronunciation Dictionary for Phone and Word Recognition with Linguistically-Based Pronunciation Rules

Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the ACL, pages 397–405, Boulder, Colorado, June 2009. c⃝2009 Association for Computational Linguistics Improving the Arabic Pronunciation Dictionary for Phone and Word Recognition with Linguistically-Based Pronunciation Rules Fadi Biadsy∗ and Nizar Habash† and Julia Hirschberg∗ ∗Department of Computer Science, Columbia University, New York, USA {fadi,julia}@cs.columbia.edu †Center for Computational Learning Systems, Columbia University, New York, USA habash@ccls.columbia.edu Abstract In this paper, we show that linguistically mo- tivated pronunciation rules can improve phone and word recognition results for Modern Stan- dard Arabic (MSA). Using these rules and the MADA morphological analysis and dis- ambiguation tool, multiple pronunciations per word are automatically generated to build two pronunciation dictionaries; one for training and another for decoding. We demonstrate that the use of these rules can significantly improve both MSA phone recognition and MSA word recognition accuracies over a base- line system using pronunciation rules typi- cally employed in previous work on MSA Au- tomatic Speech Recognition (ASR). We ob- tain a significant improvement in absolute ac- curacy in phone recognition of 3.77%–7.29% and a significant improvement of 4.1% in ab- solute accuracy in ASR. 1 Introduction The correspondence between orthography and pro- nunciation in Modern Standard Arabic (MSA) falls somewhere between that of languages such as Span- ish and Finnish, which have an almost one-to-one mapping between letters and sounds, and languages such as English and French, which exhibit a more complex letter-to-sound mapping (El-Imam, 2004). The more complex this mapping is, the more diffi- cult the language is for Automatic Speech Recogni- tion (ASR). An essential component of an ASR system is its pronunciation dictionary (lexicon), which maps the orthographic representation of words to their pho- netic or phonemic pronunciation variants. For lan- guages with complex letter-to-sound mappings, such dictionaries are typically written by hand. However, for morphologically rich languages, such as MSA,1 pronunciation dictionaries are difficult to create by hand, because of the large number of word forms, each of which has a large number of possible pro- nunciations. Fortunately, the relationship between orthography and pronunciation is relatively regu- lar and well understood for MSA. Moreover, re- cent automatic techniques for morphological anal- ysis and disambiguation (MADA) can also be useful in automating part of the dictionary creation process (Habash and Rambow, 2005; Habash and Rambow, 2007) Nonetheless, most documented Arabic ASR systems appear to handle only a subset of Arabic phonetic phenomena; very few use morphological disambiguation tools. In Section 2, we briefly describe related work, in- cluding the baseline system we use. In Section 3, we outline the linguistic phenomena we believe are crit- ical to improving MSA pronunciation dictionaries. In Section 4, we describe the pronunciation rules we have developed based upon these linguistic phenom- ena. In Section 5, we describe how these rules are used, together with MADA, to build our pronuncia- tion dictionaries for training and decoding automat- ically. In Section 6, we present results of our eval- uations of our phone- and word-recognition …

Content truncated to 500 words