Natural Language Processing Aspects Of The Device example essay topic

Natural Language Processing in Theoretical Application Abstract: In this paper, I will be discussing the creation and implementation of a device that will utilize the concepts of natural language processing and apply it to everyday activities. The device will be a carry-along unit that can be adapted to several devices a person would use everyday, like the car, items in the kitchen, and your computer. This device will be portable, compact, durable and adaptable. The device will not just adapt to any device however, the capability to interface with the device will have to be built into the objects that the device will interact with. I will discuss about this in length during the paper, as this will present the largest difficulty in making the device an actuality and usable. Another aspect I will also talk about is training.

No natural language processing system is perfect, actually none are. The user must be trained with the system and adapted to the system to really make it work. This will require technicians to not only train the user to use the device and adapt it to the user's products but also to provide constant technical support. In this design a preset lexicon must be set to make this design plausible.

I will relate how big this lexicon can be and how flexible future designs of this lexicon will appear. I will also provide logical breakdowns of commands being used, with copious amounts of examples for several environments. I will also discuss where troubles could arise in the actual implementation of the product and what semantic flaws could occur with frequency. Finally, I will discuss the feasibility of this design and whether the market and the technology are ready for this kind of attempt at natural language processing software integration.

Introduction "The Jacques BoxTM will knock your socks off! This revolutionary product will change the way humans use their everyday products forever. Want to change the radio station in your car without taking your eyes off the road? Wrist deep in cake batter and you need to preheat the oven? Tired of using the excuse that you " ve "only got two hands"? Well not anymore, with The Jacques BoxTM, you can now activate anything from your car, your computer or even your whole kitchen with just the sound of your voice".

Sounds good coming across the radio, or being offered for three easy payments of 149.99, but the reality of this is much further then the present. Natural Language Processing systems, while existent are heavily flawed and often proved useless. For example, the speech recognition process my phone uses is nice but ultimately useless. When it asks me what contact I am trying to call and I say the name, it has about a 20% success rate. That's no good.

So for this idea to work serious continued effort must be applied to the Natural Language Processing aspects of the device. The basic design of the device will be to take keywords and modifies on those keywords to adjust the properties of the item the device is currently interacting with. Setting it up that way can keep our programming simple and reduce the amount of errors. Locations I will be taking a look at the three places described above: the car, the kitchen, and the home computer. When a user enters the car he will place the device in its port just like he might place something in the cigarette lighter. The device will control several things in the car.

First is the temperature. Using simple commands the temperature can be lowered or raised. Second is the radio and all the inherent functions within a radio: CD player, tuner, volume and even satellite. Third is the Cruise Control.

This includes speeding up slowing down and coasting. Lastly might be any on board navigation systems. There are many things in the kitchen that have the possibility of some form of automated control. Any controls associated with the microwave, its timing and power level would be controlled. The oven would also be controlled by voice, adjust settings like temperature, self-clean mode and preheat. Using other technologies in unison you could query your pantry and refrigerator for their contents.

Tandem with your computer you could even order more of what you need over the internet from your local grocer. The last possible location for the device's use is the home computer. Using the device you could open programs, send mail or even dictate a paper (I wish I had this product right now!) The computer would also be the place for tweaking the product and managing its impact and effects. Installation and Management The tough part about the concept of this model is the fact that all the items that it would be used with must already be tailored to the technology.

The easiest way to adapt all of these would be to use Bluetooth technology. Most companies are probably not eager to sink capital into installing Bluetooth technology or any other kind of technology into an oven without first seeing some kind of need in the market. But let's ignore that for now. The technology that will be installed into the different items to be controlled will not have to be complex. It just needs to be able to receive commands that would take over its manual operation. The real commands will be in the device and your home computer.

For your car and your computer you would only need some kind of dock to provide interface for the device. The kitchen would need a different kind of interface that can control all the desired appliances to be controlled for one central place. There are two ways this could be done. First we could have the unit itself communicate directly with each item. This would impose some restrictions on the device. We would have to sacrifice the small size for that kind of direct control.

This is undesirable. The second option is better. There will be a dock for the device that would be permanent and the device would relate to each appliance though it. The drawback is the cost of that extra component but that cost can be passed to the con sumer and we can still keep the device to the smaller, portable size we are aiming for.

For the management of the device and its effect, installers will have to set the defaults on each setting. When a user finds they don't like how their commands are affecting the actions of the items, they can access a program on the computer that will list each action and tweak it just a bit. The dilemma here is to make sure a user will not adjust a setting to the point where it breaks the connection between a command and its action. This could be unsafe and would mean a higher cost of customer support. To solve this we can make sure the commands cannot be disassociated from their actions. Training When installing the device technicians will have several key assignments to complete.

First they will have to adapt the device to the voice patterns that will be using the device as to eliminate many voice recognition problems. The technician will have to work out with them which words will be associated with which actions. This will serve to weed out any cultural, religious, or regional speech anomalies. Complications For each area the device will be used we must consider what factors could complicate use of the technology. In the car, the first thought must be safety. I am sure the lawmakers will have the same ideas.

Speech recognition technology always screws up, and that will be a recurring theme in complicating use. If the device will control the cruise control, steps must be taken to insure that it does not misinterpret the command as sudden braking or acceleration. We also must make sure the design of the device does not fool the user into thinking they cannot let their attention lapse. To prevent other unwanted adjustments the device will not even associate with other controls it was not designed for. This will be done by keeping it in something of a closed loop. Another thing to look for is the use of a stereo and how it can interfere with what the device recognizes.

This idea can also be applied to other people who may be in the car. This should be handled either in user training or in the direct discerning of the device on user voice authentication. The next environment we must consider possible complications in is the kitchen. People can be very particular about their food and ruining it would be a bad idea. With respect to the oven safety must also be covered. We must make sure the device communicates feedback with the person on the status of the oven.

We don't want the device to be responsible for the Christmas turkey being ruined. We don't want it ruined at all, so we can have the device constantly update the person on the status of whatever they are baking. As for the refrigerator we must make sure that any information that is relayed through the device is accurate. Milk spoils easily. The last environment, the computer has the least amount of possible complications.

Computers have been hand in hand with speech recognition for years and most of the complications have been solved already. One that we do have to think about as mentioned above is user error. This will be avoided by stringent controls on how much the user can change the settings of the device. Any major changes will have to go through a technician.

NLP Examples In this section I will provide three examples of possible statements, one for each environment. I will then deconstruct those statements to examine how the device will interpret and implement them. Every time the device is in environment it goes into a mode, recognizing only statements that are relevant to the location. Also the user must specify what part of the location they want to change. Radio: Turn the station to 93.3 FM. There are four keywords, besides the call for radio mode, the device will take notice of in this section: turn, station, 93.3 and FM.

The rest will be ignored. The device will have a knowledge base that it will use to create the commands to send to the item being controlled. Here is an example: RADIO Turn / Tune Volume Up Amount Down Amount Station Destination FM / AM Off Once Change Number Shuffle Song Track So you can see the logical path taken by the device to complete the command. IF a logical or designated end, the device will either prompt the user until it reaches that end, or it will go for a default or preset value. For example if a user asks to turn the radio's volume up but does not specify an amount, a modest amount will be self-assigned to the instruction carried out by the device. Here's another example: Oven: Preheat to 450 for 10 minutes.

Here we have the call for oven mode and four keywords: preheat, 450, 10 and minutes. Here is how the knowledge base would look: OVEN Preheat Temperature Duration Time Format Bake Temperature Duration Time Format Self-clean Off MICROWAVE Defrost Temperature Duration Time Format Cook Temperature Duration Time Format Power Level (default = 10) Preset Platter Popcorn Soup Stop Some values are required input. In the previous example, missing constraints were applied by the device itself. Here, the command requires user response, i.e. computer defaults would be inappropriate. Conclusion Natural Language Processing is a field with a lot of work ahead of it.

Many believe it will be the next "big" thing in computing, having the kind of influence that internet had. The best thing for the field to do is to get commercialized. A lot of companies exploring NLP are doing so on grant money and R & D capital. That can limit resources and frustrate busy researchers. Getting some kind of valid and useful NLP products out there can generate interest in the field and support their own research using the customers of their current product as the test population.

Commercialization can also show the government and governing bodies of computer standards to push for the common use of this technology just like they did with high-definition television. Natural Language Processing is a deep and complex field, but the end, success, is a far way off.