Selection of Lexical Units for Continuous Speech Recognition of Basque

Abstract

The selection of appropriate Lexical Units (LUs) is an important issue in the development of Continuous Speech Recognition (CSR) systems. Words have been used classically as the recognition unit in most of them. However, proposals of non-word units are beginning to arise. Basque is an agglutinative language with some structure inside words, for which non-word morpheme like units could be an appropriate choice. In this work a statistical analysis of units obtained after morphological segmentation has been carried out. This analysis shows a potential gain of confusion rates in CSR systems, due to the growth of the set of acoustically similar and short morphemes. Thus, several proposals of Lexical Units are analysed to deal with the problem. Measures of Phonetic Perplexity and Speech Recognition rates have been computed using different sets of units and, based on these measures, a set of alternative non-word units have been selected.

Language Resources for a Bilingual Automatic Index System of Broadcast News in Basque and Spanish

“…Table 1 shows the main features of the three textual samples relating to size, number of words and pseudo-morphemes and vocabulary size, both in words and pseudomorphemes for each database [6]. Figure 1 shows some of the interesting conclusions derived of this analysis.…”

“…This approach has been evaluated over three textual samples analysing both the coverage and the Out of Vocabulary rate, when we use words and pseudo-morphemes obtained by the automatic morphological segmentation tool AHOZATI [6]. Table 1 shows the main features of the three textual samples relating to size, number of words and pseudo-morphemes and vocabulary size, both in words and pseudomorphemes for each database [6].…”

“…The alternative transcriptions of the word are spotted in second place, and the morphological information is later added, and it includes morpho-syntactic information, lemma information [4] and its corresponding sub-lexical unit segmentation as explained in [6]. …”

Decision Tree-Based Context Dependent Sublexical Units for Continuous Speech Recognition of Basque

“…Then, the task underwent an automatic morphological segmentation and we created two sets of lexical units as alternative to the words. We considered these new lexical units because Basque is an aglutinative language [9]. Thus, MLA task reduces the vocabulary size to 35 pseudo-morphemes (PS-MORPHS).…”

“…Thus, MLA task reduces the vocabulary size to 35 pseudo-morphemes (PS-MORPHS). Finally, N-WORDS acoustically more robust units [9] were obtained resulting in, 40. The sentences of MLA task were divided into 14,500 sentences for training and 500 for test.…”