ISCA & IEEE Workshop on Spontaneous Speech Processing and Recognition
April 13-16, 2003
This paper proposes an interactive approach to spoken interactive open-domain question answering (ODQA) systems. The goal of ODQA systems is to extract an exact answer to userís question from unstructured information sources such as large text corpora. When the reliabilities for answer hypotheses obtained by an ODQA system are low, systems need more information to effectively distinguish the exact answer required by users. In our spoken interactive ODQA system, SPIQA, spoken questions are recognized by an automatic speech recognition (ASR) system and disambiguous queries (DQs) are automatically generated to disambiguate transcribed questions. To derive appropriate DQs, ambiguous information is detected based on recognition reliability, dependency structures between phrases in the usersí questions, and features of word occurrence in the retrieved corpus. We confirmed the appropriateness of the derived DQs by comparing them with manually prepared ones. We also reconstructed the questions manually using additional information that was required by the DQs. We then tested the effect of the additional information on the performance of our ODQA system.
Bibliographic reference. Hori, Chiori / Hori, Takaaki / Isozaki, Hideki / Maeda, Eisaku / Katagiri, Shigeru / Furui, Sadaoki (2003): "Study on spoken interactive open domain question answering", in SSPR-2003, paper MAP19.