Speech-Driven Text Retrieval: Using Target IR Collections for Statistical Language Model Adaptation in Speech Recognition

Fujii, Atsushi; Itou, Katunobu; Ishikawa, Tetsuya

doi:10.1007/3-540-45637-6_9

Cited by 17 publications

(18 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The study underpinned the previous conclusion by Barnett et al Moreover, it also concluded that both standard relevance feedback and pseudo relevance feedback enable to improve the effectiveness of SQP, in particular for short queries. Fujii and his colleagues showed that using a language model generated from the target collection can significantly improve both the recognition and retrieval accuracy (Fujii, Itou, & Ishikawa, 2002). However, these studies focused solely on investigating the effects of speech recognition accuracy on IR methods based on non-spontaneous (i.e.…”

Section: Related Workmentioning

confidence: 99%

“…In fact, SQP is more complicated. It involves the integration of an ASR system and an IR system, and is not "simply connected by way of an input/output protocol" to an IR system (Fujii et al, 2002). This view was taken by (Chien, Wang, Bai, & Li, 2000) who built an efficient spoken-access approach for both Chinese text and Mandarin speech information retrieval, enabling users to submit spoken queries in an interactive way.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Written versus spoken queries: A qualitative and quantitative comparative analysis

Crestani

2006

J. Am. Soc. Inf. Sci.

View full text Add to dashboard Cite

This paper reports on an experimental study on the differences between spoken and written queries. A set of written and spontaneous spoken queries are generated by users from written topics. These two sets of queries are compared in qualitative terms and in terms of their retrieval effectiveness. Written and spoken queries are compared in terms of length, duration, and part of speech. In addition, assuming perfect transcription of the spoken queries, written and spoken queries are compared in terms of their aptitude to describe relevant documents. The retrieval effectiveness of spoken and written queries are compared using three different IR models. The results show that using speech to formulate one's information need provides a way to express it more naturally and encourages the formulation of longer queries. Despite that, 2 longer spoken queries do not seem to significantly improve retrieval effectiveness compared with written queries.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Written versus spoken queries: A qualitative and quantitative comparative analysis

Crestani

2006

J. Am. Soc. Inf. Sci.

View full text Add to dashboard Cite

show abstract

“…According to Fujii and his colleagues [6], these methods can be classified into two fundamental categories: spoken document retrieval (SDR) and spoken query retrieval (SQR). In SDR, written queries are used to search speech archives for relevant speech information, while SQR uses spoken queries to retrieve relevant textual information.…”

Section: Introductionmentioning

confidence: 99%

“…For example, earlier research from Barnett et al [1] showed that longer queries are more robust in terms of tolerating errors than shorter queries. More recently, Fujii and his colleagues [6] showed that using a language model generated from the target collection can significantly improve both the recognition and retrieval accuracy. However, these studies focused solely on investigating the effects of speech recognition accuracy on IR methods based on non-spontaneous and long queries and did not take into account the major properties of IR during the searching process, such as the effects of different query interfaces on the performance of IR systems.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Speech user interfaces for information retrieval

Gilbert¹,

Zhong²

2003

Proceedings of the Twelfth International Conference on Information and Knowledge Management - CIKM '03

View full text Add to dashboard Cite

The research proposed here concentrates on the problem of designing and developing a spoken query retrieval (SQR) system to access large document databases via voice. The main challenge is to identify and address issues related to designing an effective and efficient speech user interface (SUI), especially if the aim is to facilitate spoken queries of large document databases. Furthermore, the task of presenting large query result sets aurally should be performed such that the user's short term memory is not overloaded. In this paper, a framework allowing information retrieval to large document databases via voice is presented and findings from a research study using the framework will be discussed as well.

show abstract

Language model adaptation for fixed phrases by amplifying partial n‐gram sequences

Akiba

Itou

Fuji

2007

Systems & Computers in Japan

Self Cite

View full text Add to dashboard Cite

SUMMARYWe propose a method for creating an N-gram language model for use in a speech-operated question-answering system. We note that input questions to such a system frequently consist of an initial section, relating to the query topic, and a formulaic sentence final expression that is used in questions (a fixed phrase). While we are able to model the initial sections adequately using the target query newspaper corpus, we are not able to model the fixed phrases adequately with this data source. In this paper we frame the problem as one of adapting a language model created using a generic corpus to fixed phrases and propose a language model adaptation method that makes use only of a list of fixed phrases created by hand, rather than attempting the more difficult task of collecting an adaptation corpus. In the proposed method we determine which sections in the generic corpus correspond to N-gram sequences on the list of fixed phrases, and perform language model adaptation by amplifying the probabilities of those N-grams; this is equivalent to performing maximum a posteriori (MAP) estimation treating these partial N-gram sequences from the generic corpus itself as posterior information. We perform recognition experiments with spoken questions consisting of input to a question-answering system and confirm the effectiveness of the proposed method.

show abstract

Speech-Driven Text Retrieval: Using Target IR Collections for Statistical Language Model Adaptation in Speech Recognition

Cited by 17 publications

References 18 publications

Written versus spoken queries: A qualitative and quantitative comparative analysis

Written versus spoken queries: A qualitative and quantitative comparative analysis

Speech user interfaces for information retrieval

Language model adaptation for fixed phrases by amplifying partial n‐gram sequences

Contact Info

Product

Resources

About