Call for Participation

NTCIR-11 Spoken Query and Spoken Document Retrieval Task (SpokenQuery&Doc)

http://www.nlp.cs.tut.ac.jp/ntcir11

(Although the official deadline of the NTCIR-11 task registration is 20th

January, the organizers will accept the registration for SpokenQuery&Doc until

the end of March, 2014.)

INTRODUCTION

The NTCIR-11 SpokenQuery&Doc task will evaluate information retrieval systems

that make use of speech technologies for query input and document retrieval,

i.e. speech-driven information retrieval and spoken document retrieval.

Spoken document retrieval (SDR) in the SpokenQuery&Doc task builds on the

previous NTCIR-9 SpokenDoc and NTCIR-10 SpokenDoc-2 tasks, and will evaluate two

SDR tasks: spoken term detection (STD) and spoken content retrieval (SCR).

Common search topics will tbe used for STD and SCR which will enable component

and whole system evaluations of STD and SCR.

The emergence of mobile computing devices means that it is increasingly

desirable to interactive with computing applications via speech input. The

SpokenQuery&Doc provides the first benchmark evaluation using spontaneously

spoken queries instead of typed text queries. Here, a spontaneously spoken query

means that the query is not carefully arranged before speaking, and is spoken in

a natural spontaneous style, which tends to be longer than a typed text query.

Note that this spontaneousness contrasts with spoken queries in the form of

spoken isolated keywords which are carefully selected in advance, and represent

very different situations in terms of speech processing and composition. One of

the advantages of such spontaneously spoken queries as input to retrieval

systems is that this enables users to easily submit long queries which give

systems rich clues for retrieval, although their spontaneous nature means that

they are harder to recognise reliably.

TASK OVERVIEW

The target data for the SpokenQuery&Doc task is recordings of the first to

seventh annual Spoken Document Processing Workshop. For this speech data, manual

and automatic transcriptions (with several ASR conditions) will be provided to

task participants. These enable researchers interested in SDR, but without

access to their own ASR system to participate in the tasks.

The main task of SpokenQuery&Doc is searching spoken documents for contents

described in response to spontaneously spoken queries (spoken-query-driven

spoken content retrieval: SQ-SCR). Partial sub-tasks of the main task will also

be conducted. The sub-tasks include a spoken term detection task for the spoken

queries (SQ-STD), and a SCR task from the search results of SQ-STD (STD-SCR).

For these tasks, manual and automatic transcriptions of the spoken queries are

also to be provided. These enable participants from the previous SpokenDoc tasks

to participate in the tasks using the text queries. For the SQ-SCR and STD-SCR

tasks, a target search unit is either a speech segment that is spoken within a

presentation slide (slide retrieval task) or a boundary-free speech segment

(passage retrieval task).

FOR MORE INFORMATION

Please visit: http://www.nlp.cs.tut.ac.jp/ntcir11

TASK REGISTRATION

To register for the SpokenQuery&Doc please visit the main NTCIR-11 website at:

http://research.nii.ac.jp/ntcir/ntcir-11/

Registration deadline: 20th January 2014

(Although the official deadline of the NTCIR-11 task registration is 20th

January, the organizers will accept the registration for SpokenQuery&Doc until

the end of March, 2014.)

ORGANIZERS

Tomoyosi Akiba (Toyohashi University of Technology, Japan)

Hiromitsu Nishizaki (University of Yamanashi, Japan)

Hiroaki Nanjo (Ryukoku University, Japan)

Gareth Jones (Dublin City University, Ireland)

If you have any questions, please send e-mails to the task organizers mailing

list: ntcadm-spokenqueryanddoc@nlp.cs.tut.ac.jp

======================================================================