Text generation may require the pluralization of nouns, such as in context-sensitive user interfaces and in natural language generation more broadly. While this has been solved for the widely-used languages, this is still a challenge for the languages in the Bantu language family. Pluralization results obtained for isiZulu and Runyankore showed there were similarities in approach, including the need to combine syntax with semantics, despite belonging to different language zones. This suggests that bootstrapping and generalizability might be feasible. We investigated this systematically for seven languages across three different Guthrie language zones. The first outcome is that Meinhof’s 1948 specification of the noun classes are indeed inadequate for computational purposes for all examined languages, due to non-determinism in prefixes, and we thus redefined the characteristic noun class tables of 29 noun classes into 53. The second main result is that the generic pluralizer achieved over 93% accuracy in coverage testing and over 94% on a random sample. This is comparable to the language-specific isiZulu and Runyankore pluralizers.

Our research aims to support community health workers (CHWs) in low-resource settings by providing them with personalized information regarding their work. This information is delivered through a combination of voice- and web-based feedback that is derived from data already collected by CHWs. We describe the in situ participatory design approach used to create usable and appropriate feedback for low-literate CHWs and present usage data from a 12-month study with 71 CHWs in India. We show how the system supported and motivated CHWs, and how they used both the web- and voice-based systems, and each of the visualizations, for different reasons. We also show that the comparative feedback provided by the system introduced elements of competition that discouraged some CHWs while motivating others. Taken together, our findings suggest that providing personalized voice- and web-based feedback could be an effective way to support and motivate CHWs in low-resource settings.

There is already strong evidence that mobile videos are a good vehicle for public health information dissemination, but there remain open questions around sustainability, appropriate target users, consumption patterns, content, and usage models. We analyse log and interview data of 42 community health workers (who were first time smartphone users) from a longitudinal 17-month deployment to better understand how the utility of mobile videos played out over time in rural Lesotho. During the study period, videos were viewed at an average of 170 times per month, for a total of 2898 views. Through this data we draw these primary findings: a) pausing is not contextually necessary, b) age is not a barrier to usage, c) the primary predictor of popularity of a given video is topical relevance and national campaigns, d) there is no apparent relationship between video length, popularity and completion rates, and e) new videos have only a short-lived novelty effect. Furthermore, we affirm that regular engagement with CHWs has an impact on continued usage, in addition to being important for reducing attrition due to technical issues.

Natural Language Generation (NLG) can be used to generate personalized health information, which is especially useful when provided in one's own language.However, the NLG technique widely used in different domains and languages-templates-was shown to be inapplicable to Bantu languages, due to their characteristic agglutinative structure.We present here our use of the grammar engine NLG technique to generate text in Runyankore, a Bantu language indigenous to Uganda.Our grammar engine adds to previous work in this field with new rules for cardinality constraints, prepositions in roles, the passive, and phonological conditioning.We evaluated the generated text with linguists and non-linguists, who regarded most text as grammatically correct and understandable; and over 60% of them regarded all the text generated by our system to have been authored by a human being.

There are many domain-specific and language-specific NLG systems, which are possibly adaptable across related domains and languages.The languages in the Bantu language family have their own set of features distinct from other major groups, which therefore severely limits the options to bootstrap an NLG system from existing ones.We present here our first proof-of-concept application for knowledge-to-text NLG as a plugin to the Protégé 5.x ontology development system, tailored to Runyankore, a Bantu language indigenous to Uganda.It comprises a basic annotation model for linguistic information such as noun class, an implementation of existing verbalisation rules and a CFG for verbs, and a basic interface for data entry.

Accredited Social Health Activists (ASHAs) have been shown to have a positive impact on health outcomes of the households they visit, particularly in maternal and neonatal health. As the first line of the public health system in many countries, they are a critical link to the broader public health infrastructure for community members. Yet they do this all with minimal training and limited support infrastructure. To a pregnant woman, an ASHA is a trusted ally in navigating the health system---information gathered is returned by appropriate advice and counseling. To the health system, the ASHA is a key channel of valuable householdlevel information for the public health system, yet she generally receives minimal guidance in return. In this paper we present ASTA---the ASHA Self-Tracking Application---a system that provides ASHAs with timely, on-demand information regarding their own performance compared to their peers. Using ASTA, ASHAs access comparative performance data through both a web-based and voice-based interface on demand. We evaluated ASTA through a 12-month deployment with 142 ASHAs in Uttar Pradesh, India, assessing the impact of providing feedback on ASHA performance. We find that ASHAs with access to the ASTA system made significantly more client visits, with average monthly visits 21.5% higher than ASHAs who had access to a control system. In addition, higher ASHA performance was correlated with increased usage of ASTA. However, the performance improvement was front-loaded, with the impact of the system decreasing toward the end of the study period. Taken together, our findings provide promising evidence that studying and incorporating tools like ASTA could be cost effective and impactful for ASHA programs.

The provision of personalized patient information has been encouraged as a means of complementing information provided during patient-doctor consultations, and linked to better health outcomes through patient compliance with prescribed treatments.The generation of such texts as a controlled fragment of Runyankore, a Bantu language indigenous to Uganda, requires the appropriate tense and aspect, as well as a method for verb conjugation.We present how an analysis of corpora of explanations of prescribed medications was used to identify the simple present tense and progressive aspect as appropriate for our selected domain.A CFG is defined to conjugate and generate the correct form of the verb.

A growing body of HCI4D research studies the use of SMS communication to deliver health and information services to underserved populations. This paper contributes a novel dimension to this field of study by examining if a hybrid computer-human SMS system can engage pregnant women in Kenya in health-related communication. Our approach leverages the different strengths of both the computer and the human. The computer automates the bulk-sending of personalized messages to patients, allowing the human to read patients' replies and respond to those in need of attention. Findings from a 12-month deployment with 100 women show that our approach is capable of engaging the majority of participants in health-related conversations. We show that receiving messages from the system triggers participant communication and the amount of communication increases as participants approach their expected due date. In addition, analysis of participants' messages shows that they often contain sensitive health information conveyed through a complex mixture of languages and 'txting' abbreviations, all of which highlight the benefits of including a human in the workflow. Our findings are relevant for HCI researchers and practitioners interested in understanding or engaging underserved populations.

Surveys conducted by human interviewers are one of the principal means of gathering data from all over the world, but the quality of this data can be threatened by interviewer fabrication. In this paper, we investigate a new approach to detecting interviewer fabrication automatically. We instrument electronic data collection software to record logs of low-level behavioral data and show that supervised classification, when applied to features extracted from these logs, can identify interviewer fabrication with an accuracy of up to 96%. We show that even when interviewers know that our approach is being used, have some knowledge of how it works, and are incentivized to avoid detection, it can still achieve an accuracy of 86%. We also demonstrate the robustness of our approach to a moderate amount of label noise and provide practical recommendations, based on empirical evidence, on how much data is needed for our approach to be effective.

Efficient health systems require reliable data. In developing countries the need for accurate data is particularly acute, as organizations are often forced to make decisions on a tight budget with limited capacity for data collection. In this note, we describe recent progress toward developing a set of algorithms that can help detect and classify anomalies in health worker data. Building on recent efforts to use unsupervised multinomial techniques for outlier detection, we outline the steps required to turn a set of statistical tests into a framework that can be implemented by health organizations, and calibrate these algorithms on a large dataset from a partner health organization. Here, we describe the core methods, present results from ongoing analyses, and outline our plan for future work, including plans to obtain labeled training data that will allow us to detect and classify different types of outlier in community health worker data.

Community health workers (CHWs) have been shown to be an effective and powerful intervention for improving community health. Routine visits, for example, can lower maternal and neonatal mortality rates. Despite these benefits, many challenges, including supervision and support, make CHW programs difficult to maintain. An increasing number of mHealth projects are providing CHWs with mobile phones to support their work, which opens up opportunities for real-time supervision of the program. Taking advantage of this potential, we evaluated the impact of SMS reminders to improve the promptness of routine CHW visits, first in a pilot study in Dodoma, Tanzania, followed by two larger studies with 87 CHWs in Dar es Salaam, Tanzania. The first Dar es Salaam study evaluated an escalating reminder system that sent SMS reminders directly to the CHW before notifying the CHW's supervisor after several overdue days. The reminders resulted in an 86% reduction in the average number of days a CHW's clients were overdue (9.7 to 1.4 days), with only a small number of cases ever escalating to the supervisor. However, when the step of escalating to the supervisor was removed in the second study, CHW performance significantly decreased.

Organizations in low income regions need tools for collecting and reporting on data. Existing paper-based approaches are often slow and incomplete when compared to data collection tools on mobile devices. In this demonstration, we present the authors Android-based system in addition to broader work by the OpenRosa Consortium - a community dedicated to building open source, standards-based tools for mobile data collection, aggregation, analysis, and reporting.

Every year almost 10 million children die before reaching the age of five despite the fact that two-thirds of these deaths could be prevented by effective low-cost interventions. To combat this, the World Health Organization (WHO) and UNICEF developed the Integrated Management of Childhood Illness (IMCI) treatment algorithms.

Operation scheduling (OS) is a fundamental problem in mapping an application to a computational device. It takes a behavioral application specification and produces a schedule to minimize either the completion time or the computing resources required to meet a given deadline. The OS problem is NP-hard; thus, effective heuristic methods are necessary to provide qualitative solutions. We present novel OS algorithms using the ant colony optimization approach for both timing-constrained scheduling (TCS) and resource-constrained scheduling (RCS) problems. The algorithms use a unique hybrid approach by combining the MAX-MIN ant system metaheuristic with traditional scheduling heuristics. We compiled a comprehensive testing benchmark set from real-world applications in order to verify the effectiveness and efficiency of our proposed algorithms. For TCS, our algorithm achieves better results compared with force-directed scheduling on almost all the testing cases with a maximum 19.5% reduction of the number of resources. For RCS, our algorithm outperforms a number of different list-scheduling heuristics with better stability and generates better results with up to 14.7% improvement. Our algorithms outperform the simulated annealing method for both scheduling problems in terms of quality, computing time, and stability

Design space exploration during high-level synthesis is often conducted through ad hoc probing of the solution space using some scheduling algorithm. This is not only time consuming but also very dependent on designer's experience. We propose a novel design exploration method that exploits the duality of time- and resource-constrained scheduling problems. Our exploration automatically constructs a time/area tradeoff curve in a fast, effective manner. It is a general approach and can be combined with any high-quality scheduling algorithm. In our work, we use the max-min ant colony optimization technique to solve both time- and resource-constrained scheduling problems. Our algorithm provides significant solution-quality savings (average 17.3% reduction of resource counts) with similar runtime compared to using force-directed scheduling exhaustively at every time step. It also scales well across a comprehensive benchmark suite constructed with classic and real-life samples.

Design space exploration during high level synthesis is often conducted through ad-hoc probing of the solution space using some scheduling algorithm. This is not only time consuming but also very dependent on designer's experience. We propose a novel design exploration method that exploits the duality between the time and resource constrained scheduling problems. Our exploration automatically constructs a high quality time/area tradeoff curve in a fast, effective manner. It uses the MAX-MIN ant colony optimization to solve both the time and resource constrained scheduling problems. We switch between the time and resource constrained algorithms to quickly traverse the design space. Compared to using force directed scheduling exhaustively at every time step, our algorithm provides a significant solution quality savings (average 17.3% reduction of resource counts) with similar run time on a comprehensive benchmark suite constructed with classic and real-life samples. Our algorithms scale well over different applications and problem sizes.