Machine translation using corpus-based acquisition of transfer rules

Recent advances in mobile devices and their sensing capabilities have enabled the collection of rich contextual information and mobile device usage records through the device logs. These context-rich logs open a venue for mining the personal preferences of mobile users under varying contexts and thus enabling the development of personalized context-aware recommendation and other related services, such as mobile online advertising. In this article, we illustrate how to extract personal context-aware preferences from the context-rich device logs, or context logs for short, and exploit these identified preferences for building personalized context-aware recommender systems. A critical challenge along this line is that the context log of each individual user may not contain sufficient data for mining his or her context-aware preferences. Therefore, we propose to first learn common context-aware preferences from the context logs of many users. Then, the preference of each user can be represented as a distribution of these common context-aware preferences. Specifically, we develop two approaches for mining common context-aware preferences based on two different assumptions, namely, context-independent and context-dependent assumptions, which can fit into different application scenarios. Finally, extensive experiments on a real-world dataset show that both approaches are effective and outperform baselines with respect to mining personal context-aware preferences for mobile users.

Typical tag recommendation systems for photos shared on social networks such as Flickr, use visual content analysis, collaborative filtering or personalization strategies to produce annotations. However, the dependence on manual intervention and the knowledge of sufficient personal preferences coupled with the folksonomic issues limit the scope of these strategies. In this paper, we present a fully automatic and folksonomically scalable tag recommendation model that can recommend tags for a user's photos without an explicit knowledge of the user's personal tagging preferences. The model is learned using the collective tagging behavior of other users in the user's local interaction network, which we believe approximates the user's preferences, at least partially. The tag recommendation model generates content-based annotations and then uses a Naive Bayes formulation to translate these annotations to a set of folksonomic tags selected from the tags used by the users in the local interaction network. Quantitative and qualitative comparisons with 890 Flickr networks show that this approach is highly useful for tag recommendation in the presence of insufficient information of a user's own preferences.

Parallel corpora are essential for training statistical machine translation models. Since parallel sentence-aligned corpora are usually noisy due to inexact automatic methods when generated from parallel or comparable documents, we need to clean parallel corpora. In this paper, new features are introduced to assess the correctness of a sentence pair. Also, the impact of new features in combination with state-of-the-art features introduced in the literature is systematically evaluated. Statistical methods have been used for feature extraction and therefore this approach is independent to language. In order to better understand the problem characteristics, four supervised classification algorithms are used to classify sentence pairs as noise or parallel. Evaluating the models by taking accuracy and f-measure into account shows that using the system for cleaning a noisy parallel Farsi-English corpus, the maximum entropy model performs better than the main filtering techniques used in this paper and shows a significant improvement over two other systems.

The aim of personalizing web information systems is to tailor content (media) to the user's personal preferences, goals and context, in turn increasing the reusability of that content. However, most developers are increasingly seeking to apply 'Web as a Platform' based approaches where web-based content is integrated with web services to provide the platform and environment features that were traditionally offered by desktop environments and applications. Therefore, next generation personalized web information systems need to be capable of dynamically personalizing, adapting and integrating web content and web services to create customized experiences for individual users. The goal of the research presented in this paper is to provide innovative techniques and technologies to achieve this. As part of this research, a novel architecture for the generation of Personalized Learning Activities, which highlights both the need for, and the power of combining and personalizing content and services has been developed. The system has been evaluated in the context of a Higher Education use case based on the user's perception of three key elements of the Personalized Learning Activity: the personalization, the usability and the educational benefits.

In many of today's software systems, most notably online social networks, users can share personal information. Behind the simple action of sharing is a more complicated thought process regarding privacy: which content to share, with whom to share, and why to share. For a user, it's time-consuming and error-prone to check individual personal content for privacy violations. Hence, it would be ideal if a personal assistant can learn its users' privacy preferences and subsequently help users' decision-making by signaling potentially private content. A personalized privacy assistant can help its user make privacy decisions taking into account the ambiguity and uncertainty of privacy predictions as well as its user's personal preferences. Moreover, an explanation of why an image is considered public or private can aid the user in understanding the assistant's decisions.

In order to identify a user's personal preference in navigating the Web or to recommend collected Web information, it is very useful to analyze the user's Web-browsing behavior. However, it is difficult to determine which Web-browsing behaviors are influential on predicting a user's interest because each individual has his/her own habit and personal manner in surfing the Web and locating documents of interest. In this study, we propose an adaptive personalized recommender system based on a preference-thesaurus constructed by learning a user's Web-browsing behavior. The major components of the proposed recommender system are the Web-browsing behavior monitor, preference-thesaurus constructor, relevant document recommender, and user feedback learner. The adaptive nature of the proposed system allows the personalization of recommendation and the identification of the ever-changing influential Web-browsing behaviors. A proof-of-concept system is implemented and experiments are performed to verify the system's capability to personalize recommendation and to learn through user's feedbacks.

This paper describes a method for the automatic inference of structural transfer rules to be used in a shallow-transfer machine translation (MT) system from small parallel corpora. The structural transfer rules are based on alignment templates, like those used in statistical MT. Alignment templates are extracted from sentence-aligned parallel corpora and extended with a set of restrictions which are derived from the bilingual dictionary of the MT system and control their application as transfer rules. The experiments conducted using three different language pairs in the free/open-source MT platform Apertium show that translation quality is improved as compared to word-for-word translation (when no transfer rules are used), and that the resulting translation quality is close to that obtained using hand-coded transfer rules. The method we present is entirely unsupervised and benefits from information in the rest of modules of the MT system in which the inferred rules are applied.

Nowadays; Parallel corpus is one of the most important resources which can be employed in different researches such as machine translation, bilingual lexicography, and linguistics. This paper describes the process of building a large-scale (about 400, 000 sentence pairs) English-Persian parallel corpus called Tehran Parallel Corpus (TPC). The aim of study is to introduce the structure and explain the materials utilized for constructing TPC. In addition, some useful tools developed within the project have been introduced and three sorts of the statistical machine translation systems trained by TPC have been considered. In order to develop a high quality parallel corpus, unsure alignments recognized via a MaxEnt classifier have been eliminated from the corpus. As an intrinsic evaluation, 1,600 sentence pairs are elicited randomly and compared manually with a gold standard test set. As an extrinsic evaluation, three Phrase-based SMT systems, which is trained by TPC are incorporated. The results demonstrate the superiority of our translator systems over English to Persian Google translator system in term of BLEU and TER metrics.

This paper describes the system jointly developed by members of the Departament de Llenguatges i Sistemes Inform` atics at Universitat d’Alacant and the Prompsit Language Engineering company for the shared translation task of the 2014 Workshop on Statistical Machine Translation. We present a phrase-based statistical machine translation system whose phrase table is enriched with information obtained from dictionaries and shallowtransfer rules like those used in rule-based machine translation. The novelty of our approach lies in the fact that the transfer rules used were not written by humans, but automatically inferred from a parallel corpus.

This paper describes the constructing of a large-scale (above 500,000 pair sentences) Chinese-English parallel corpus. The current status of Chinese corpora is overviewed with the emphasis on parallel corpus. The XML coding principles for Chinese--English parallel corpus are discussed. The sentence alignment algorithm used in this project is described with a computer-aided checking processing. Finally, we show the design of the concordance of the parallel corpus and the prospect to further development.

Shared mobility can fundamentally change road transport and improve life quality. Car-sharing services have reached a high level of growth worldwide and members adopting them as an alternative transportation mode are constantly on the increase. The advent of highly automated vehicles could boost the adoption of car sharing as more inexpensive and convenient on-demand services could be achieved. In recent years, vehicle automation is facilitated and supported by the latest advances in information and communication technologies, forming the cornerstone of the new generation of In-Vehicle Intelligent Systems. The goal of this study is to introduce a novel in-vehicle cognitive framework that utilises (i) driver's/user's profile data and personal preferences, (ii) real-time information associated with the vehicle's driving environment, and (iii) previous knowledge and experience, in proposing adaptations of the vehicle's level of autonomy, in an automated manner, when a driver/user wishes to make a road journey within the automated vehicle sharing context. Knowledge is obtained through the exploitation of Bayesian networking principles. Indicative results on the performance of the proposed cognitive mechanism and the associated knowledge-based selection scheme are presented, mostly with regards to proactively identifying the optimum level of autonomy to be implemented.

In this paper, we discuss how a short-range wireless communication scheme called Mobile Encounter Network (MEN) can provide additional value for both the consumer and the service/product provider (the company). When used as an information search system on the market environment, a MEN-based system allows services and products being promoted at the location they are available. From the customer's point of view, MEN may provide a digitally augmented vision, an enhanced view to the current environment. Combined with a set of filtering and search rules this may provide a self-manageable context combining the user's own personal environment and preferences to the services and features available in the current surroundings. An imaginary case study of this kind of system in execution is provided. The conclusion is that such method may can generate additional revenue to the company and please the customers' buying process.

The present application translates German financial terms to the Greek language aiming to maximize the translation quality. The demand for better translation increases the need for more MT systems; journalists or professionals in the banking sector require accurate translations. A novel statistical method of MT is presented, where we use a probability distribution over sentence pairs from a parallel corpus (German-Greek). Both the MT system and the parallel corpus were built from scratch. The present application combines parameters in a new way and estimates the probabilities that extract the best translation.

In this paper, we introduce MyCampus, a Semantic Web environment for context-aware mobile services, which we are in the process of developing and validating on Carnegie Mellon University's campus. The environment revolves around a growing collection of customizable agents capable of (semi-) automatically discovering and accessing user personal resources (e.g. calendar, location tracking functionality, food preferences) and Web services as they assist their users in carrying out different tasks such as planning an evening out, organizing a study group or filtering incoming messages. The openness of the MyCampus architecture directly derives from a set of ontologies for describing personal resources, contextual attributes, user preferences and web services. Contextual information and other personal details about a user are accessed via a Semantic e-Wallet, subject to access privileges set by the user.

We present XSearcher, an associative memory based desktop search system, which exploits associations by creating semantic links of personal desktop resources from explicit and implicit user activities. With these links, associations among memory fragments can be built or rebuilt in a user's brain during a search. The personalized ranking scheme uses these links together with a user's personal preferences to rank results by both relevance and importance. XSearcher enhances traditional keyword based search systems since it is closer to the way that human associative memory works.