Skip to content

In order to easily visualize each system with its main purpose and possibilities, the notion of conceptual model is used.

It includes several key points that need to be answered:

The first one is requirement. It answers the question of what problem should the system solve. It encompasses the whole meaning of creating a system.
The second key point is the structure of the system. It defines the components and their relations within the system.
Another point is the behavior of the system. This needs to be clarified to clearly understand how the system solves the problems that is mentioned in the requirement point.
The fourth point is properties which includes functionalities of the system.
And lastly, operating characteristics, which encompasses inputs, outputs, and interactions with users.

Below is the depiction of conceptual model that deals with developing and supporting natural resource monitoring.

1

Speaker verification is a solution to a problem of authenticating a person by his/her voice. It verifies whether the person is the claimed one using the voice based on the biological fact that everyone has a unique voice. This solution is mostly used by call centers that needs user verification in their processes, such as banks.

To train a model that will verify speakers, a dataset with many audio speeches generated by different speakers will be needed so that the model will learn to differentiate them and compare with new voices. Many scientific papers and works have used the popular dataset named VoxCeleb [1]. This dataset has captured voices of many celebrities from interviews uploaded to the Youtube platform. Hence, it is a noisy dataset because interviews can have music as their background or they can be taken in streets where voices of people talking or sounds of cars can be recorded. Overall, the dataset consists of two versions where one is a smaller with 1251 unique speakers and over 100,000 utterances, and the other has 6,112 unique speakers with over 1 million utterances. The dataset labeled every speaker with unique id. Here is an example of one of the recordings:

Speaker with id0189 talking with music in the background.

Additionally, some datasets can be found in UCI Machine Learning Repository. One of them is ISOLET[2], which has 150 speakers with almost 8000 utterances. This dataset is also noisy, yet simple one, where speakers only name alphabet letters.

  1. https://www.robots.ox.ac.uk/~vgg/data/voxceleb/index.html
  2. https://archive.ics.uci.edu/ml/datasets/ISOLET

It is a very tough process to choose a topic to research that will be both valuable to the community and interesting to the author. One can advise to just start researching what you like and then see what will be the outcome - the famous principle of YOLO. Although one can choose to do researching this way, without a clear plan and map, it can be very hard to understand what is the scope and focus of the research, and in which direction they are going to. To give a little bit spoiler of the whole process of finding a research topic, the most important thing is having a definite goal and knowing that this goal is achievable.

To help a researcher with the choice of correct topic, there exist a technique that will tell in short time whether the research is worth of starting it. There are 8 questions to be answered that will help to define if the idea is doable, manageable and worthy.

The first question is "What are you trying to do?". This question needs to be answered without any jargons or scientific terms, and explain simply what is the thing that one tries to do or achieve. This could be any goal like making delicious dinner or buying the best computer.

The second question is "How is it done today?". Here it is implied that one should research how others have done it, in order to have an understanding about the topic. The question also covers limits, challenges or any difficulties of the current practice. For the topic of making delicious dinner, one challenge would be that the term "delicious" is a relative one, and the concept of a tasty meal changes based on the geography, culture and individual experiences.

The third question is stated as "What’s new in your approach?". One needs to think what difference they are making with the research. For the previous example of making delicious dinner, one can say that the meal will be prepared using only vegetables growing in a particular area. Moreover, a reason of starting the research could give a hint in answering this question since it could be specific case that will bring novelty. The author should also focus on the reason why he/she thinks that the new approach will be successful.

The fourth question is very simple: "Who cares?". To put it differently, who will benefit from this research and the author's results. For our example, it could be a family who will eat the dinner at the end of the day. It seems pretty simple and obvious question, however, thinking about an exact target group helps a lot in defining the scope and adjusting the topic to fit the need.

The fifth question is "What are the risks?". One should always take into account risks that would lead to the failure of the project. All the scenarios need to be considered because without considering them, the plan of conducting the research will be incomplete and prone to a failure. Like Dr. Strange in the movie of "The Avengers", the author should see all the possibilities of what could happen in the future. This will help one to prevent or mitigate the problems. For our example, it could be not holding enough knowledge about cooking temperatures of certain vegetables and making them either uncooked or overcooked. Also, there is always a slight chance of overcooking a meal (saying from my own experience).

The sixth question is "How much will it cost?". Here, the cost could be of any kind like financial or temporal. Knowing the cost beforehand could help to eliminate the ideas that will take more time or any other resources than available.

The seventh question is put as "What are the midterm and final exams to check for success?" by our professor. This simply says when are the milestones that are most important that need some checking or validation. In our example, midterm could be when a mother-in-law will just come to the kitchen to see how everything is going, and the final would be putting a meal on the table to be tasted by the whole family.

The last but not the least, question of "Why now?" should be answered. It is the one thing that has affected the author's decision the most to conduct the research now. Or the question can also be put that what has changed from the past to today that there is a possibility to do it now. In our example, it could be that the family has moved to that specific area where those vegetables are cultivated.

All in all, these questions are simple at the first sight, however, missing to answer one of these questions lead to some gap or hole within the topic which further cause getting out of scope or time, or maybe the research being completely valueless. Putting these questions in a check-list could really help to define the idea and extract all the "sauce" from it.

Good time of the day to the readers! My name is Natavan (surname is harder - Akhundova). I am a Master student at the dual degree program offered by George Washington University, U.S and ADA University, Azerbaijan. Concurrently, I work as a machine learning engineer at ATL Tech, Azerbaijan, though it has been harder to combine both study and work since I moved to Washington, DC.

The area of computer science that I am mostly focused is Natural Language Processing (NLP) and speech recognition. At the moment, I have published one poster about my research as a bachelor's student "Development of limited-vocabulary ASR for Azerbaijani", and two articles (also about speech recognition) inspired by our work at the company - "Automatic Speech Recognition in Taxi Call Service Systems" and "Development of Speech Recognition Systems in Emergency Call Centers".

In my leisure time (which is not much), I, together with other volunteers, translate a book about programming into the Azerbaijani language, in order to help beginners in our country to understand CS topics easily in their own language (the luxury that we did not have at the time we started). The translation should be finished till September, and the book will be published around that time (so excited!!!).

Algorithms Unlocked by Thomas H. Cormen - the book that we (volunteers) are translating.

Other than that, I devote myself to my family and try to spend time with them as much as possible since there is a recent addition of a new little cute member to our family who needs to be taken care of (those diapers are a very dangerous nuclear weapon!).

1

Welcome to your brand new blog at GW Blogs.

To get started, simply log in, edit or delete this post and check out all the other options available to you.

For assistance, visit our comprehensive support site, check out our Edublogs User Guide guide or stop by The Edublogs Forums to chat with other edubloggers.

You can also subscribe to our brilliant free publication, The Edublogger, which is jammed with helpful tips, ideas and more.