Personal speech calendar with timing keywords aware and schedule time prediction functions

Speech calendar functionality to provide users with voice recording for schedule database and to retrieve it by speech query inputs. For example, users can record a database sentence: "Go to attend the TENCON conference in Fukuoka, Japan, on November 24" as a personal schedule, and use a nature sentence input, such as "Any schedule on November 24?" to retrieve the database. To implement this spoken document retrieval (SDR) system on embedded systems, we have previously made the column-based row-based (CBRB) partial matching algorithm and the whole-matching-plane-based (WMPB) accumulation algorithm. Without using a speech recognizer, the CBRB and WMPB computes feature distances between database sentences and query sentence. If common words appear, such as "November" and "24", the similarity score would be high. However; if a query sentence: "Any trip in next week?" the system will not be able to use the CBRB or WMPB algorithm for SDR, because there is no matched word (common word) in both sentences. In addition; if an itinerary is recorded as "Go to TENCON conference on next Monday" such a speech sentence will face difficulties for the future SDR. The timing keywords "next Monday" which is not a datable schedule leads to a query input: "Any trip on November 24?" will not be able to find the itinerary. Moreover, the system cannot determine whether this kind of indeterminate schedule should be automatically deleted or not if users do not delete this kind of schedule on their own. Another problem is that if the keyword "Monday" appears in the query, this schedule will be continuously retrieved every time in the future. To solve the above problems; this paper presents a personal speech calendar with timing keywords aware and schedule time prediction functions. After a schedule to be recorded, system takes down the recording-time and determines the timing keywords, then makes a timing-tag which indicates the expected timing information including the fields of year, month, date and time of the schedule. The timing-tag can provide explicit schedule timing for the future comparison with the speech query inputs.