Natural language is hierarchically structured: smaller units (e.g., phrases) are nested within larger units (e.g., clauses). When a larger constituent ends, all of the smaller constituents that are nested within it must also be closed. While the standard LSTM architecture allows different neurons to track information at different time scales, it does not have an explicit bias towards modeling a hierarchy of constituents. This paper proposes to add such an inductive bias by ordering the neurons; a vector of master input and forget gates ensures that when a given neuron is updated, all the neurons that follow it in the ordering are also updated. Our novel recurrent architecture, ordered neurons LSTM (ON-LSTM), achieves good performance on four different tasks: language modeling, unsupervised parsing, targeted syntactic evaluation, and logical inference 1 . * Equal contribution. {yi-kang.shen,jing.shan.shawn.tan}@umontreal.ca.

No abstract

The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.

We report several studies developing a parent-rated measure of emotional distress for children in Singapore, with the key objectives being to derive a very brief valid measure of global distress. The refined item set comprised behaviourally expressed broad manifestations of emotional distress. Three developmental studies were undertaken, with the first two involving parental ratings on the measure for validation against clinician-rated distress levels, while also testing two rating options for the measure. We established clear comparative advantages to the rating anchors used in the Revised Rutter Scales. High inter-rater agreement was established across parental ratings, with the latter finding supporting objectives for the measure. Paternal scores correlated more strongly than maternal scores with clinician-generated distress scores. Additional properties of the measure were tested in a large community sample of nearly 2,000 Singapore schoolchildren in their last 2 years of primary school, allowing prevalence estimates and mean scores to be derived for each item. Here, girls and boys received identical total scores, scores were also independent of the number of children in the family and of ordinal position, and mothers returned higher scores than fathers.

Naltrexone has been demonstrated in western studies to be a useful pharmacological adjunct within treatment programmes for alcoholic patients. We report the first study of its efficacy and usefulness in an Asian region. This project was designed to allow naltrexone˜s performance to be assessed under routine clinical conditions but with patients selected on the basis of their being likely to comply. Following in-patient detoxification, 53 male alcohol-dependent patients admitted to the Alcohol Treatment Centre at Woodbridge Hospital, Singapore, were enrolled in a 12-week, placebocontrolled trial of naltrexone hydrochloride (50 mg/day). Subjects were randomized on a 2:1 basis, with 35 receiving naltrexone and 18 receiving placebo. Analyses identified that a higher percentage of naltrexone patients completed the study (40% vs. 22%). In the study non-completers, the dropout rate due to drinking relapse was also lower in the naltrexone group (9% vs. 43%). Of the 39 patients for whom drinking status over the trial could be ascertained, fewer naltrexone-treated patients drank (33% vs. 53%). Alcohol craving scores also showed a selective and distinct reduction in the naltrexone-treated group. Results suggest that naltrexone may be an effective and safe aid to treatment of alcohol dependent patients in Asian patients, for whom the aims are to reduce alcohol craving and drinking reinstatement, but where compliance is likely to be low. [Lee A, Tan S, Lim D, Winslow RM, Wong KE, Allen J, Hall W, Parker G. Naltrexone in the treatment of male alcoholics-an effectiveness study in Singapore. Drug Alcohol Rev 2001;20:193 199]