New Transcription System using Automatic Speech Recognition ( ASR

Technical Consultant of the House The Japanese Parliament (Diet) was founded in 1890. Since the very first session, verbatim records had been made by manual shorthand over a hundred years. However, early in this century, the government terminated recruiting stenographers, and investigated alternative methods including ASR technologies. The House of Representatives has chosen ASR for the new system. The system was deployed and tested in 2010, and it has been in official operation from April 2011. The new system handles all plenary sessions and committee meetings. Speech is captured by the stand microphones in meeting rooms. Separate channels are used for interpellators and ministers. The speaker-independent ASR system generates an initial draft, which is corrected by reporters. Roughly speaking, the system's recognition error rate is around 10%, and disfluencies and colloquial expressions to be corrected also account for 10%. Thus, reporters still play an important role. There are Japanese language-specific issues. First, we need to convert kana phonetic symbols to kanji or Chinese characters. This conversion often involves ambiguity because of many homonyms. Therefore, it is very hard to type in real time. Only limited stenographers using a special keyboard can perform. Moreover, there are differences between the spoken-style and the transcript style. So, we need to rephrase in many cases, but re-speaking or shadow speaking is not so simple. Requirements for the ASR system are as follows. The first is high accuracy; over 90% is preferred. This can be easily achieved in plenary sessions, but is difficult in committee meetings, which are interactive, spontaneous, and often excited. The second requirement is fast turnaround. In the House, each reporter is assigned every 5-minute segment of a meeting session. ASR should be performed almost in real-time, so reporters can start working promptly even during the session. The third issue is compliance to the orthodox transcript guideline of the House. The electric dictionary of 60K lexical entries used in the system was proofed. In summary, the compliance issue is solved by hard work, fast turnaround is feasible by current computers, and high accuracy is technically most challenging.