Experiments in spoken document retrieval using phoneme n-grams

Managing misspelled queries in IR applications

“…Therefore, the system will be better prepared for working in noisy environments, since it is able to cope not only with spelling errors, but also with out-of-vocabulary words and spelling, morphological or even historical variants Lee and Ahn, 1996;Mustafa and Al-Radaideh, 2004), in contrast with classical conflation techniques based on stemming, lemmatization or morphological analysis, which are negatively affected by these phenomena. This feature is extremely valuable, not only for regular text retrieval tasks, but also for specialized tasks such as spoken document retrieval (SDR) (Ng et al, 2000), or cross-lingual information retrieval (CLIR) over closely-related languages using no translation, but only cognate matching 4 (McNamee and Mayfield, 2004a). The third major factor for the success of n-grams in IR applications comes from their inherent language-independent nature.…”

Automatic Summarization

“…Further, [195] shows that ignoring word boundaries when extracting phone-based features does not affect retrieval performance significantly.…”

“…In [195], different methods for extracting overlapping phonesequence indexing features for SCR are explored in detail. This article arrives at the general conclusion that phone-based retrieval is not as effective as word-based retrieval, but there are certain situations where it is appropriate.…”

“…Further, if speech recognition must be performed on a platform with limited capacity (i.e., a hand-held device), then a small language model, such as a phoneme bigram model, makes the ASR system lightweight and compact. The authors of [195] find that in terms of phone-sequence-based indexing features, a combination of phone 3-grams and 4-grams proved most effective. This result confirms the findings of [304] that phone-based features derived from word-level transcripts are able to help compensate for word-level error.…”

Combining Word and Phonetic-Code Representations for Spoken Document Retrieval

“…The first one considers the transcription of speech utterances into phoneme or syllable sequences instead of word sequences by using a phoneme/syllable recognizer [8,9,10]. On the other hand, the second method proposes making use of more than the top-1 transcription hypothesis.…”