Paul De Palma

Viewpoint: Why women avoid computer science

As it happens, the literature fairly bubbles over with speculation as to why there are so few young women in computer science courses. We hear about math anxiety, violent computer games, the scarcity of mentors, and a supposed female preference for “relational work”.7 Since no one really knows why women avoid computer science—or what to do about it—I feel justified in offering a guess of my own. Among the many reasons offered, math anxiety is the most obvious. It is also the least defensible. Commentators never seem to notice that women receive almost half of the undergraduate degrees in mathematics. In fact, they received nearly 40% of them in 1970, well before the women’s movement became a mass phenomenon.5 Not only do young women not avoid mathematics, they embrace it. What if the precision of mathematics, that “most masculine of subjects” in the words of one study,7 is exactly what has long invited women? The flip side is that the ill-defined nature of computing is what drives them away. Young men drawn to computer science, engineering, and physics like to tinker. They enjoy taking things apart and putting them back together. They like kits, gadgets, and screwdrivers. They were the boys who set up the audio-visual equipment in high school 30 years ago, and who now man—the choice of gender is deliberate—the school’s computer network. They are fascinated with anything that moves, especially if it has wheels or wings, and, crucially, is not alive.4 The men usually given credit for the microcomputer all started with screwdrivers and soldering irons. Bill Gates and Paul Allen built a Basic interpreter to run on their Altair 8800, a computer kit for hobbyists, in the mid-1970s. Steves Wozniak and Jobs, of Apple fame, built their first machine to dazzle pals in Silicon Valley’s Homebrew Computer Club around the same time. In fact, I claim that microcomputers are responsible for the steep rise in the number of women entering computer science following its introduction, as well as for the steep drop a few years later.6 In 1971, fewer than 2,400 students received degrees in computer science from a handful of academic departments. By 1986, that number had jumped to nearly 42,000, including almost 15,000 women. It is clear that the dramatic growth of computer science as an academic discipline is due to the microcomputer and, of course, to the extravagant promises that buzz around it. If the number of computer science degrees had continued to grow as it had from 1975 to 1985 (and if the population grew at its average annual rate over the same period), by next year everyone in the U.S. would be the proud holder of one. Lucky for us this didn’t

Cryptography and computer security for undergraduates

The panel discusses solutions to the problem of computer security education.

Trusses, NP-Completeness, and Genetic Algorithms

The optimization of large trusses often leads to a nearly optimal solution, rather than a truly optimal design. In fact, the problem space for truss optimization grows exponentially with the size of the truss. Using the method of problem reduction, this paper demonstrates that truss optimization is in the set of NP-complete problems. Hence, the only practical techniques for solving the truss problem are heuristic in nature. Genetic algorithms provide a viable solution for large trusses.

Speech recognition with syllables and concepts

The transformation of speech to words is not necessary to emulate human linguistic performance in some contexts. A large vocabulary continuous speech recognition (LVCSR) system can perform more accurately with a syllable-based, as opposed to a word-based, language model. This accuracy can be further enhanced by the addition of a concept model, where a concept is an equivalence class of words and phrases.

Syllabification by phone categorization

Syllables play an important role in speech synthesis, speech recognition, and spoken document retrieval. A novel, low cost, and language agnostic approach to dividing words into their corresponding syllables is presented. A hybrid genetic algorithm constructs a categorization of phones optimized for syllabification. This categorization is used on top of a hidden Markov model sequence classifier to find syllable boundaries. The technique shows promising preliminary results when trained and tested on English words.

Linguistic type-frequency in preschool boys and girls with and without hearing loss

Linguistic type-frequency, how many different lexical types are used, has been examined in usage-based models of child language acquisition. In general, it has been shown that exposure to greater type-frequencies increases children’s productive use of language and that language in turn bootstraps later development including language and literacy. It is not currently known if pediatric hearing loss impacts the type-frequency of those children’s early communicative productions. In this study, we used a public database available via HomeBank [http://homebank.talkbank.org] to examine the type-frequency in 53 cognitively intact children, 37 with mild to moderate hearing loss (HL) and 16 peers who were typically-developing (TD). For each child, we analyzed 15 minutes of high volubility from a representative daylong recording collected in a natural family setting via an audio recorder worn by the child. Results indicate a main effect of sex favoring girls, but no main effect of HL. There were, however, interacti...

The influence of siblings on toddlers’ mean length of utterance

Linguistic complexity is an indicator of language development in young children. Complexity of a child’s linguistic productions have been shown to increase with development, but may be affected by factors such as disability or environmental variables. Here, we look into the role of family composition as a possible influence on a child’s developing ability to use increasingly complex language. In particular, we ask if a toddler’s mean length of utterance (MLU) is affected by the presence of siblings in the family and whether the sex of the child may play a role. MLU values were extracted from the public HomeBank database [http://homebank.talkbank.org] of transcribed natural child speech for both the target toddler and for siblings present in the recordings. Results indicate a main effect of increased MLU in children without siblings, but interaction effects suggest that differences may be driven by the boys without siblings alone. There was no correlation between the MLU of the target child and the MLU of ...

Using automatic speech processing to analyze fundamental frequency of child-directed speech stored in a very large audio corpus

Child-Directed-Speech (CDS) is associated with raised fundamental frequency (f0). In a previous paper we claimed that f0 could be extracted from 500 hours of audio recordings using soft computing techniques and that mothers, but not fathers, increase f0 in CDS. Using an audio corpus more than ten times larger, this paper reports that fathers do raise f0 but not as much as mothers. The principle finding is a proof of concept: 1) very large speech corpora, unavailable until recently, can be processed using soft computing techniques; 2) the use of very large corpora may force revisions of conclusions based on smaller datasets.

Fundamental frequency of speech directed to children who have hearing loss

Studies of child-directed speech (CDS) have shown that when talking to children, parents systematically use (among other strategies) increased fundamental frequency (F0). Lombard effects such as increased F0 have also been documented when addressing a listener who is hard-of-hearing (HH). Here, we examine F0 of mothers and fathers in families with HH versus typically developing (TD) children in CDS and adult-directed speech (ADS) contexts. Whole-day audio recordings were collected by a child-worn audio recorder and analyzed by automatic speech recognition (ASR) software to identify segments of vocal activity by children and their parents (LENA Research Foundation, Boulder, CO). Custom software extracted F0 values in all conditions. We found that (1) mothers are much more systematic in their use of CDS than fathers, (2) parents do not appear to be sensitive to the hearing status of their children, and (3) parents of HH children may have higher overall F0 irrespective of CDS or ADS. Results suggest that mot...

Optimizing Resources in Undergraduate Research

The Gonzaga University Center for Evolutionary Algorithms (GUCEA) is an interdisciplinary and collaborative research group which brings together faculty and undergraduate students from Civil Engineering, Computer Science, and Mathematics. GUCEA students develop genetic algorithm (GA) software which is applied to difficult problems in structural engineering and mathematics. Gonzaga University is primarily an undergraduate institution, so developing a research program involves challenges unique to smaller schools. This paper presents ways to overcome these challenges, discusses the benefits of creating a research program involving undergraduate students, and highlights GUCEA findings in optimal truss design and book embeddings of graphs. GONZAGA UNIVERSITY CENTER FOR EVOLUTIONARY ALGORITHMS (GUCEA) The Gonzaga University Center for Evolutionary Algorithms (GUCEA) aims at advancing the knowledge of genetic algorithms (GA) in optimization problems. Optimization deals with seeking the best possible solution to a design problem. GA reach this goal, mimicking the natural selection process [Haupt and Haupt, 1998]. The center was founded in the fall of 2005 as a result of a collaborative effort between Professor Paul De Palma in the Department of Computer Science, Dr. Sara Ganzerli in the Department of Civil Engineering, and Dr. Shannon Overbay in the Department of Mathematics. The development of GUCEA as a research center followed six years of collaboration between Professor Paul De Palma and Dr. Sara Ganzerli. Dr. Ganzerli came to Gonzaga in 1999 looking to expand her graduate research, which focused on structural optimization considering uncertainty by using GA. Professor De Palma identified an undergraduate student who could develop the software needed to perform the structural optimization. During the next year, Professor De Palma took an active role in this research effort, training new students in the area of GA. The newly-founded research team designed GA software to optimize structures. The resulting research activities generated peer-reviewed publications and invited presentations, and