About two weeks ago, I posted about the MGM DVD Class Action Settlement. As in my last post, this one will be half about MGM DVDs and half about speech applications. And if you are looking for a really meaningful discussion of the technical details behind the settlement, you should stop reading my blog immediately and go to The Digital Bits.
Yesterday, I received the Proof of Claim. It’s pretty simple to fill out. After providing some basic contact info, you enter the number of DVDs you are returning and you check the box next to the titles. You can return only one copy of each title. The Proof of Claim comes with a prepaid shipping label.
You then choose whether to receive $7.10 for each DVD or an equal number of replacement DVDs from a list of about 330 DVDs. No mixie-matchie. It’s either all cash or all new MGM DVDs.
While I have four DVDs from the trade-in list that were purchased during the relevant time period, I really didn’t see anything from the Exhibit “A” list that I wanted more than what I already have. Sure, both “Breakin 2: Electric Boogaloo” and “Candyman 2: Farewell to the Flesh” were major hits at Sundance and Cannes. Also, Lambada and Showgirls have a lot going for them, but, dammit, I do like my copy of “24 Hour Party People”. After all, I did buy these DVDs for a reason.
Actually, the DVDs on the exchange list aren’t that bad. Choices include:
- 12 Angry Men
- Elmer Gantry
- Henry V [gets my vote for best film with English as a foreign language]
- Mississippi Burning
- How to Stuff a Wild Bikini
Er, uh, maybe I should have stopped at five. There are a few other good movies on the list (though not many good ones from after 1995), but it’s mostly slim pickins (though I don’t think any Slim Pickens).
And now about that speech app. Like I posted before, the speech app did not try to automatically capture my address. Lo and behold, and I have to say that I told you or somebody else so, the “make a recording and have someone transcribe it later” trick didn’t work so well. While I obviously did receive the letter, the address wasn’t correct.
The street I live on has two f’s. The address on the envelope replaced the f’s with s’s. There is no Tissin road in Oakland. Any decent address verification algorithm should have picked that up, especially since they captured the zip code properly. While you might not think that an automated speech application would be very good at capturing addresses, they actually can be extremely good at it.
For the applications we develop at work, we use a USPS database for all addresses in the US. Our speech apps are very good at taking an ambiguous utterance from a caller and very quickly examining a large number of possible matches. If a person were performing this work, it would be such a mind numbing task that errors would very likely creep in. But, speech apps don’t get tired or bored. Since the app already has the text of what was recognized, validating that address against streets in the zip code is very easy. And, if it turns out not to be valid or if it gets multiple matches, the app can easily and quickly reprompt the caller for more info. Yes, I, for one, welcome our new speech application overlords.