Girish Dalvi

Design and evaluation of Devanagari virtual keyboards for touch screen mobile phones

Lack of an easy and efficient text input mechanism in Indic scripts has been a barrier to large-scale adoption of ICTs in India. We present findings from a usability evaluation of three keyboard designs for Indic scripts for touch screen phones. The design of one of the keyboards is based on the frequency of characters, while the designs of the other two are based on the logical structure of the script. We evaluated the keyboards with participants with low-level of education through a first-time usability test and a longitudinal usability test. One of the logically structured keyboards started out with significantly higher success rate, typing speed, and lesser errors than the other two. The longitudinal test involving text input of 500 words did not conclusively prove that either design was better. Our study establishes benchmarks for text input speeds, errors and ratings for the initial learning phase for text input in Marathi among less educated users.

A Protocol to Evaluate Virtual Keyboards for Indian Languages

We were asked to evaluate input mechanisms for touch-screen devices with the objective of standardising one of them for each of 14 major languages of India. For this purpose, we propose a protocol that consists of a 45-minute long training session, a 20-word first-time usability test, and a longitudinal test consisting of about 30 sessions, each of which required the user to type about 10 phrases 4 to 6 words long (a total of 300 phrases). The evaluation should be done with school children from standards 4th to 7th. The course of the evaluation may last 2-4 weeks for each user. To help follow the protocol over this long period and to collate the data, we offer a tool. Currently, we provide test corpora for Assamese, Bengali, Gujarati, Hindi, Kannada, Marathi, Odia, Tamil, Telugu, and Urdu. We have ensured that each corpus represents a mix of informal communication between people, popular phrases from films, songs, poetry and public discourse, and formal texts from school books and literature. We have tagged each phrase according to typing difficulty, phrase length, and memorability and age appropriateness. We evaluated the protocol through pilot tests with 206 users in Marathi, Gujarati, Hindi, Bengali, Odia, Assamese and Tamil. In this paper, we present the original protocol, the detailed findings from the Marathi pilots, and the proposed modifications to the protocol.

Corpus of Marathi Word Frequencies from Touch-Screen Devices Using Swarachakra Android Keyboard

We describe and publish online a corpus containing word frequencies of Marathi texts that were actually typed by 27,474 users using the Android version of the Swarachakra Marathi keyboard on their mobile devices between August 2013 and September 2014. The corpus has 1,484,059 total words and 184,257 unique words. The paper also provides a preliminary analysis of the word frequencies and some comparisons with two existing corpora. It also provides a qualitative review of the nature of errors that users have made while typing and some idiosyncrasies that they have exhibited. We hope and expect that this corpus will be useful for future researchers, particularly those involved in word completion and auto-correction of user errors.

Shift+Tap or Tap+LongPress?: The Upper Bound of Typing Speed on InScript

This paper presents the results of a within-subject longitudinal evaluation on Inscript keyboard, which is the national standard layout for Indian scripts. We studied the practical upper bound speed and accuracy as well as the effect of practice. Through longitudinal transcription task of 400 repeated attempts, we observed typing speeds for highly experienced users consistently peak close to 120 cpm i.e. 2.5 times that of fastest speeds reported in literature. Our analysis compared the lower bound times for Tap, Tap+LongPress and Shift+Tap, the three text input mechanisms in this keyboard. Among the two alternative methods, our findings established Tap+LongPress method to be faster than Shift+Tap method and almost equally accurate. Also, we derived a model which explains the influence of corrected errors and number of practice attempts on the typing speed.

Development of an Intuitive User-Centric Font Selection Menu

The font selection menu in most application software’s is arranged alphabetically; in recent years one can also see the split menu approach being used. An alphabetical arrangement presupposes that the users are aware of the font characteristics and usage scenario through its name. Unless the font name specifies it; the scheme does not in any which way indicate the morphological features or the usage scenarios of a given font. In order to address these issues, a set of systematic experiments was conducted to capture the typeface classification strategies of users with a modified card sorting technique. The collected data was subjected to hierarchical clustering algorithms to come up with a collective user-centric classification system for Devanagari fonts. The classes created by this method appears to address the problems raised by alphabetical orderings by allowing the groupings to be more intuitive (and realistic) while retaining statistical validity over large user sets.

Human-Computer Interaction - INTERACT 2017

Recently, combining Information and Communication Technologies (ICT) with Technical Vocational Education and Training (TVET) for a low literate population is gaining interest, as this can lead to more effective socio-economic development. This strategy can more easily provide employment and bring community wide change because of the improved quality and relevance of education materialQuery. Although TVET providers are present throughout India that uses some ICT, challenges remain for prospective students including illiteracy, language, resource limits and gender boundaries. Providing TVET that is accessible to low-literate people in rural village communities requires a shift in the design of ICT so that it is universally useable, even for communities like tribal India that has a largely oral culture. In this article, we detail the design and development of an ICT driven TVET model for a mostly illiterate audience in rural India and measure its efficacy. Through our ethnographic and usability study with 60 low-literate oral and novice village users, we present the issues faced and the solutions we incorporated into our new model. The results show that users performed better in the vocational course units with the solutions incorporated.