10 likes | 127 Views
Capstone Experience at UNH Manchester Student Guided Mentoring for an Undergraduate Research Group in Speech. Capstone Objectives. Challenges. Student Mentoring.
E N D
Capstone Experience at UNH Manchester Student Guided Mentoring for an Undergraduate Research Group in Speech Capstone Objectives Challenges Student Mentoring The Capstone course for Computer Information Systems (CIS) majors at the University of New Hampshire at Manchester (UNHM), that started in spring 2011, offers students a faculty guided project focused on research topics in speech [1]. The course exposes students to the rigors of speech processing, giving them the opportunity to engage in solving real world problems, gaining invaluable experience along the way. Students in the CIS major are non-traditional Students, some that matriculated from area community colleges. They tend to be a few years older than the average undergraduate student and many work up to 20 hours or more per week to support themselves and pay for their education. UNHM is a commuter college in an urban setting in Manchester, the largest city in New Hampshire. The second iteration of the course, in spring 2012, added student mentoring as a way to improve student learning and class flow. Several students from the spring 2011 course participated, either directly associated with the project or as external consultants. One student specifically enrolled in UNHM’s mentoring program, joining the class as an in-class tutoring guided by the Center of Academic Enrichment. These tutors are specifically trained to engage students and help improve both study and communication skills in a fast paced classroom environment. Students faced many challenges being immersed in an unfamiliar and difficult environment that is speech recognition. At the onset of the course students had to break into teams based on their interest and skill sets. There were both hardware and software needs that needed a collaborative effort to complete in a short amount of time. Once infrastructure was in place, students could then focus on speech specific tasks. Key points students had to address: o Organization o Planning o Time Management o Knowledge o How Speech Recognition works o Organizing large data sets o Collaboration o Communication o Accountability o Writing o Documentation o Personal work logs Student mentorship took two forms: an in-class tutor and external class consultants. The in-class tutor took project leadership responsibilities, organizing sub-groups and helping knowledge transfer from previous semester. This resulted in better communication and flow of information. With faculty leading project, emphasis for students was on passive knowledge transfer. With peers, students were more engaged in actively learning what needed to be done. Two external class consultants were used as resources by students outside class time. Both were involved in independent study courses with the faculty adviser on Capstone material Experience Students submitted self and group evaluations at end of the semester. Comparing comments from first iteration of course (2011) with the second (2012), students felt less apprehensive about the material. Students liked having the in-class tutor to help lead and organize actives. Overall project productivity improved: -> 2011 class finished 45% of proposed tasks -> 2012 class finished 75% of proposed tasks Set of tasks between two classes were similar with initial tasks in 2011 focused on installing system and speech software and in 2012 documenting and upgrading the same. This constituted 20% work for each. Remaining tasks were identical as work in 2011 needed to be repeated, both for learning and practical purposes. Research & Development Students were faced with real world research problems. Speech lends itself to both theoretical and applied research. Tasks students needed to tackle varied and included: o Theory o Improve Performance o Recognition accuracy o Performance speed o Generate new models o Combine different data sources o Applied o Solve problem using speech as solution o Collaborating with local industry Speech is also an Experimentation science and thus requires students to learn how to rigorously capture results and summarize outcomes. This is an invaluable learning exercise that helps improve student’s analytical and writing skills. Technology References Students worked with the state-of-the-art Carnegie Mellon University Sphinx Open Source Speech Recognition Toolkit [2], a set of associated tools and a comprehensive data set: o CMU Sphinx Speech System o Sphinx 1.0 Trainer o Sphinx 3.0 Decoder o CMU Language Modeling Toolkit o NIST SCLite Scoring Package o LDC Switchboard Telephone Corpus [1] Jonas, M., Capstone Experience – Lessons from an Undergraduate Research Group in Speech at UNH Manchester, SIGITE-2011, West Point, NY, 2011 [2] CMU Sphinx - Speech Recognition Toolkit, Carnegie Mellon University, School of Computer Science, Pittsburgh, PA, http://www.speech.cs.cmu.edu. Dr. Michael Jonas Assistant Professor of Computing Technology University of New Hampshire at Manchester