Morphological Segmentation for Keyword Spotting

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 880–885, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Morphological Segmentation for Keyword Spotting Karthik Narasimhan1, Damianos Karakos2, Richard Schwartz2, Stavros Tsakalidis2, Regina Barzilay1 1Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology 2Raytheon BBN Technologies {karthikn, regina}@csail.mit.edu {dkarakos, schwartz, stavros}@bbn.com Abstract We explore the impact of morpholog- ical segmentation on keyword spotting (KWS). Despite potential benefits, state- of-the-art KWS systems do not use mor- phological information. In this paper, we augment a state-of-the-art KWS sys- tem with sub-word units derived from su- pervised and unsupervised morphological segmentations, and compare with phonetic and syllabic segmentations. Our exper- iments demonstrate that morphemes im- prove overall performance of KWS sys- tems. Syllabic units, however, rival the performance of morphological units when used in KWS. By combining morphologi- cal, phonetic and syllabic segmentations, we demonstrate substantial performance gains. 1 Introduction Morphological analysis plays an increasingly im- portant role in many language processing appli- cations. Recent research has demonstrated that adding information about word structure increases the quality of translation systems and alleviates sparsity in language modeling (Chahuneau et al., 2013b; Habash, 2008; Kirchhoff et al., 2006; Stal- lard et al., 2012). In this paper, we study the impact of morpho- logical analysis on the keyword spotting (KWS) task. The aim of KWS is to find instances of a given keyword in a corpus of speech data. The task is particularly challenging for morphologi- cally rich languages as many target keywords are unseen in the training data. For instance, in the Turkish dataset (Babel, 2013) we use, from the 2013 IARPA Babel evaluations, 36.06% of the test words are unseen in the training data. However, 81.44% of these unseen words have a morpholog- ical variant in the training data. Similar patterns are observed in other languages used in the Babel evaluations. This observation strongly supports the use of morphological analysis to handle out- of-vocabulary (OOV) words in KWS systems. Despite this potential promise, state-of-the-art KWS systems do not commonly use morphologi- cal information. This surprising fact can be due to multiple reasons, ranging from the accuracy of ex- isting morphological analyzers to the challenge of integrating morphological information into exist- ing KWS architectures. While using morphemes is likely to increase coverage, it makes recogni- tion harder due to the inherent ambiguity in the recognition of smaller units. Moreover, it is not clear a priori that morphemes, which are based on the semantics of written language, are appropriate segmentation units for a speech-based application. We investigate the above hypotheses in the context of a state-of-the-art KWS architec- ture (Karakos et al., 2013). We augment word lattices with smaller units obtained via segmenta- tion of words, and use these modified lattices for keyword spotting. We consider multiple segmen- tation algorithms, ranging from near-perfect su- pervised segmentations to random segmentations, along with unsupervised segmentations and purely phonetic and syllabic segmentations. Our exper- iments show how sub-word units can be used …

Content truncated to 500 words