Brenda Reyes Ayala

Metadata records machine translation combining multi-engine outputs with limited parallel data

One way to facilitate Multilingual Information Access (MLIA) for digital libraries is to generate multilingual metadata records by applying Machine Translation (MT) techniques. Current online MT services are available and affordable, but are not always effective for creating multilingual metadata records. In this study, we implemented 3 different MT strategies and evaluated their performance when translating English metadata records to Chinese and Spanish. These strategies included combining MT results from 3 online MT systems (Google, Bing, and Yahoo!) with and without additional linguistic resources, such as manually‐generated parallel corpora, and metadata records in the two target languages obtained from international partners. The open‐source statistical MT platform Moses was applied to design and implement the three translation strategies. Human evaluation of the MT results using adequacy and fluency demonstrated that two of the strategies produced higher quality translations than individual online MT systems for both languages. Especially, adding small, manually‐generated parallel corpora of metadata records significantly improved translation performance. Our study suggested an effective and efficient MT approach for providing multilingual services for digital collections.

A machine learning approach to evaluating translation quality

We explored supervised machine learning (ML) techniques to understand and predict the adequacy and fluency of English-Spanish machine translation. Five experiments were conducted using three classifiers in Weka, an open-source ML tool. We found that the highest performance was achieved by applying a dimensionality reduction approach to the classification task, which included collapsing a numeric scale of quality to two categories: high quality and low quality. Our results showed that the Support Vector Machine classifier performed the best at predicting the adequacy (65.65%) and fluency (65.77%) of the translations. More research is needed to explore the methodologies of applying ML to translation evaluation.

Building a multilingual test collection for metadata records

This paper describes the principles and processes of building a test collection that enables multilingual information retrieval for digital metadata records. The collection includes a multilingual collection of 1,005,752 metadata records, their Chinese and Spanish machine translation results, 45 topics generated through crowdsourcing, and their relevant judgments.

Towards building a collection of web archiving research articles

Presentation for the 2014 International Internet Preservation Consortium (IIPC) General Assembly. This presentation discusses building a collection of web archiving research articles.

Analysis of URL References in ETDs: A Case Study at the University of North Texas

Article discussing a case study at the University of North Texas (UNT) on an analysis of URL references in electronic theses and dissertations (ETDs).