Webinar presenter Eric LaScola answered a number of your questions after his presentation, “How Speech Recognition Technology is Transforming Police Reporting.” Here are a few of his responses.
Audience Question: What level of precision does Dragon have for dictating in other languages specifically in Spanish?
Eric LaScola: Dragon, today, is actually in English only. We do accommodate, as I mentioned, a wide number of accents but in terms of the language itself, we’re currently not in Spanish or other languages. Although, that’s something that we are taking a look at as we expand beyond the United States and Canada.
Audience Question: Is Dragon Law Enforcement compatible with CAD and RMS systems and if so, which ones?
Eric LaScola: We’re compatible actually with virtually all RMS and CAD systems. The thing to remember about Dragon is that we kind of operate, you’ll have a toolbox that gets launched when you launch Dragon and operates over Windows, if you will. So, anywhere where you can place a cursor and type with a keyboard, you can use Dragon. Now the only difference is there are certain applications like Microsoft Word and Google Chrome, where you can actually do more from an editing standpoint but pretty much anywhere where you can insert your cursor and type with a keyboard, you can use Dragon, and obviously that includes all of the RMS systems.
Audience Question: With all of the free offerings in speech recognition today, how does Nuance differentiate itself from other offerings such as speech recognition built into Windows?
Eric LaScola: That’s actually a question we get a lot of. I’d say the primary difference is the amount of research that goes into when we approach a new market like the Law Enforcement market. So, it’s the customization that goes on with the product and things like, say, we’re going to incorporate profanity, we’re going to put in car and trucks, makes, and models. The other thing that is also key is, unlike a lot of the free offerings in the market today, you don’t need an internet connection to use Dragon, so it can operate right on the MDT itself and all of that processing happens on the device, none of it is sent out over the cloud. So, that’s another key difference. Then, just from accuracy in a robustness standpoint, and the number of features built into Dragon, what I’ve mentioned at the top, we’re still very much, we began in this over twenty years ago, so we really have leadership experience, if you will in developing these solutions.
Audience Question: Does your research indicate if the technology as more popular with newer recruits, veterans, or both?
A: Both, actually and for different reasons. If you look at some of the veterans in the force, many of those candidly are not great typists. So the idea that you can now speak and use dictation is a great help to them so they don’t have to type. From a new recruit standpoint and those that are just coming out of the academy, it really comes down to this attendant expectation that they have, growing up with technology and I’ve mentioned the use of consumer tech in Alexa and Siri, etc. So they’re very accustomed to these and they frankly expecting in the workplace, so it’s very popular with both.
Audience Question: Can we use Dragon to transcribe body-worn camera recordings?
Eric LaScola: The short answer is no and the reason is that you’re dealing with multiple speakers and you’re dealing with an environment where the background is very unpredictable in the level and quality of what you’re capturing from audio will vary tremendously. So, what we found happens is, you can often capture a pretty well the officer’s voice but all of the ambient background noise, particularly when you have multiple speakers. We get asked also questions by detectives that are interested interview room settings and they say, “Geez, we ‘d love to be able to record three or four individuals and have Dragon split out the audio for each person speaking and provide a transcript. The truth is nobody does that well today. Now, that being said, the technology is advancing very rapidly and could we foresee a day when we’re able to do that? Yes, but today, it’s just not something that frankly, anybody can do very well.
Audience Question: Can you talk about some of the sizes of police organizations that you are currently working with and provide some information about the volume of reporting that they are doing?
Eric LaScola: It has really run the gamut. We do a lot of work with smaller police departments that may be resource constrained, and so they’re looking at Dragon as a way to help alleviate some of that burden. And we’ve also, we’ve worked with large agencies. In fact, if you look at the case study that we have as a download for everybody, a large mid-western state police department is using Dragon every day as part of their reporting. In using it, not just to create incident reports but also to use the command and control functions to open up all of their applications at the start of the day that’s created macros etc. So, I’d encourage you to look at that and there’s an agency with over two hundred and fifty personnel, so it really has run the gamut. The more, the greater the number of officers, frankly, the greater the efficiency gauged. If you think about saving one officer per day, twenty, thirty minutes, obviously that scale is over the amount of officers that you have in the department. So the larger the organization, the larger your potential savings.
Audience Question: We got a couple of questions related to software pricing and things like that. I’m guessing probably the pricing is going to be different from organization to organization. So, could you just kind of talk about how pricing works for Dragon?
Eric LaScola: It is in fact based on a volume-licensing agreement. We’d be happy to talk of specifics with anybody that’s interested with our sales department. But it’s based on tiers and the number of officers. So, the way it’s priced is by voice profile, so every speaker that you would have. Typically that would be, of course, every officer using the product, that constitutes a single license. So, we price it by voice profile by an officer and it’s based on tiers given the number of officers in the department. And again, we’re happy to speak specifics with anybody, to get in touch with our sale personnel.
Audience Question: How does the text correction process work when using the product?
Eric LaScola: So, what you can do, there are multiple commands where you can say select a beginning of a word or beginning of a sentence through the end of the sentence or you can say, select a misspelled word, you can say “select that word” and basically it will come up with the corrections menu. That way you’ll have the option. Dragon will anticipate potential corrections, so it will give you the option, say, 1 through 4, if it’s a spelling correction, and you can just say “select 1.” You can also then, go in and say, “You know, Dragon didn’t understand this particular word, you can say “select that word” and then just repeat another word or the same word again and Dragon will just overwrite it. And we’d be very happy of course, to show anyone who’s interested, live demonstration of the product for purposes of time, we didn’t do that today, but very happy to arrange that.
Audience Question: What is the training period for voice profiles and is there a backend and frontend speech option?
Eric LaScola: The training process is essentially non-existent today. Many folks remember back from the early 90s that consumer product where you had to read large passages of text. With all of the advancements today and the availability of voice files, basically today, you don’t need to do any training out of the box. I’m not clear on the second part of the question.
Audience Question: They might be talking about whether it can be used either on or offline. The specific wording is, “is there a backend and frontend speech option?”
Eric LaScola: So, if I understand this question correctly, frontend speech option, I would interpret it as kind of live dictation, if you will, so if you’re speaking, the Dragon will basically convert your speech to text right away, so yes that would be a frontend speech. From a backend perspective, I think just referring to the idea that can you create a voice file and then on the backend, feed that voice file into Dragon when you have downtime and have it transcribed in a later time, then yes, you can do that as well.
Audience Question: How long does it typically take to implement the system in an agency?
Eric LaScola: It clearly depends on the size of the agency. What I’ll tell you is we’ve gotten pretty good at it over the years. The process typically involves doing a small and a brief pilot with a dedicated group of users. We then go to a formalized training process and introducing the team to Dragon and typically look to have a group of champions, if you will, within the organization. And then, we work with the chief-of-police, his appointed his or her appointed staff to expand that and roll that out. One of the things that’s most important to us is ensuring customers’ success with the program because, frankly, we want to go back to that customer and ask to use them as a reference for all the efficiency that they gain. So, the short answer to the question is, the pilot might take a couple of weeks and then after that, from a roll-out perspective, maybe thirty to sixty days, depending on the size of the organization.
Audience Question: Our last question for today’s asking whether or not there are UK, Canadian, US spelling contexts available.
Eric LaScola: Not today. But again, as I’d mentioned with the question on whether it’s available in Spanish, we’re looking at that particularly as we had interest. In fact, most recently in the last IACP meeting, we were approached by several people, law enforcement professionals from the UK, so definitely on the radar screen.
Click Here to Watch a Recording of “How Speech Recognition Technology is Transforming Police Reporting.”