ITR Proposal: tPod1 - A Portable Translator


By Katie Lang



  1. Project Summary


Communication is an important part of every aspect of life. From the economy and international trade to everyday interactions with other people, communication is invaluable. Currently, technology has advanced communication to be simplistic and more convenient. Items such as cell phones, personal computers, and the Internet allow for quick communication of ideas between people. Even MP3 players, such as the iPod, have made listen to the message of musical artists faster, easier, and portable.

This proposal will take the commutation properties of MP3 players further than its current state. We propose to adapt an MP3 player2 to listen to the words a person is saying in one language and translate their ideas into the language of the user's choice. By adding the ability to listen with speech recognition software, the MP3 player will use machine translation software to deduce what was said and translate that into another language for the user to hear. For this project, the main focus is the communication aspect; however, the research that will be done to accomplish this goal will also be applicable to security, defense, and education.


  1. Background


Before the full description of the project, a general background of the major research areas will be provided. We begin with a short description of MP3 players. Also known as digital audio players (DAPs), MP3 players are an electronic device that can store and play audio files. Usually headphones or speakers are plugged into DAPs to listen to the audio files being played. For this project, we will take the current technology and add a microphone to record sounds, mainly a person's voice. This leads us to the topic of speech recognition.

Speech recognition software will convert the voice recording into a language that the translation software will be able to use. Currently, speech recognition can be found in multiple places. A primary example would be customer service via the telephone. Speech recognition is used to guide customers to commonly used sections of a company, depending on the user's needs and responses to questions. Our tPod will use the speech recognition software to identify the language being spoken and translate what is said into a language that can be used our translator. This then draws our attention to the topic of machine translation.

Machine translation software will take one language and parse the sentence before finding the most probable translation of that sentence in the language of the user's choice. Current research in this field has resulted in improvements to issues with translations of one language into another3. Thus, machine translation and speech recognition will be the research focus of this project in order for both to improve to the point were the accuracy is sufficient for commercial use4.


  1. Project Description


The goal of this project is to create software that will use an MP3's hardware to translate what is spoken into another language of the user's desire. In order to do this, machine translation and speech recognition software need to be altered and improved upon, at least to the point were the device would be commercial acceptable5 and the software would not cause a need to increase the physical size of the MP3.

The research will begin with an investigation on the best speech recognition because the ability to translate what is being said is imperative for the machine translation to work properly. This research will include the current issues with speech recognition and what solutions have been proposed. Then we will test and propose our solution based on this knowledge. Similarly, after the speech recognition software begins progressing, an inquiry on the best methods for machine translation will commence. The research for machine translation will include current issues and solutions which we will use to improve upon the area. Once these two subjects have improved enough for commercial use, they will be applied to the MP3 player and the resulting solutions will be published in an effort to further advance each area.

Some of the benefits of this device include the tPod's ability to be used as a learning tool. The user could speak one language into the microphone and listen to see if their statement was correct by translating their statement into a language they understand and comparing to what they had wanted to say. The reverse is true as well. The user could choose to listen in another language to practice listening skills. However, the main benefit would be to open communication between people without having either party learn a new language. This would be especially important to tourists in a country who need to communicate with local people when there is an emergency.

Furthermore, this project will support international collaboration in research. Different countries could contribute their versions of language parsers that are not available currently in the United States. Each collaborator would benefit from the experience of having his or her parser under scrutiny and from the communication with other countries about possible solutions to machine translation (and speech recognition) issues. Our tPod will only include major languages6, so collaboration from many of the countries in which these languages are used is greatly desired.


  1. Conclusion


The impact of this research will effect researchers and, hopefully, people world-wide. By including international cooperation, the results and learning experiences the researchers accumulate while participating will add unification to the field and diversity to methods and ways of thinking of researchers in that field. Additionally, the end product could open up multiple lines of communication between those who could not communicate clearly before because of a language barrier. This software could even be included in telephones or audio communication via the Internet, such as programs like Skype™. Also, as stated previously, our tPod can be used as a language learning device. Thus, the language barrier would no longer be an issue, the ability to learn any language would be available to the general public, and the areas of machine translation and speech recognition would be further advanced for the scientific community.



1This name will only be used if I won't be sued for using it.

2From here on referred to as tPod ('translator pod'), unless changed because of previous footnote.

3This information was gained from Dan Gildea's speech on machine translation given Monday, September 15, 2008 in the Computer Studies Building of the University of Rochester.

4In other words, the translation may not be perfect, but it falls within acceptable margins of error.

5Note: Our goal is not to produce a product, but to further the software it would support.

6Although we would like to include some dialects of the languages we would use, it is out of the scope of this project to include them. However, consideration to accents will be incorporated into the design of the speech recognition software.