Create a database for storing data about utterances with the parameters below.
speech_id: primary key of the table
lang: language of the segment (often the page). Type [varbinary(35)] suggested per Page table, it can probably be a lot smaller
seg_hash: Unique hash of the segment/sentence. To be investigated, adding on length to end of hash might make sense
voice: The selected voice. Could also make use of the NameTableStore to make the type an int instead
date_stored (could also be expiry_date): indication of when to expire the entry. Having an expiration ensures audio related to sentences that have since been edited are phased out and that lexicon improvements are phased in. Exact logic for this to be investigated.
For adding database tables for extensions see Manual:Developing extensions.