Articles

Code-switching is the practice of moving back and forth between two languages in spoken or written form of communication. In this paper, we address the problem of word-level language identification of code-switched sentences. Here, we primarily consider Hindi-English (Hinglish) code-switching, which is a popular phenomenon among urban Indian youth, though …...

This paper presents a method to predict re- trieval terms from relevant/surrounding words or descriptive texts in Japanese by using deep belief networks (DBN), one of two typical types of deep learning. To determine the effec- tiveness of using DBN for this task, we tested it along with baseline methods …...

Many knowledge repositories nowadays con- tain billions of triplets, i.e. (head-entity, re- lationship, tail-entity), as relation instances. These triplets form a directed graph with enti- ties as nodes and relationships as edges. How- ever, this kind of symbolic and discrete stor- age structure makes it difficult for us to exploit …...

Many news papers publish articles for chil- dren. Journalists use their experience and in- tuition to write these. They might not aware of readability of articles they write. There is no evaluation tool or method available to deter- mine how appropriate these articles are for the target readers. In this …...

Memorizing the whole set of graphemes is generally accepted as the first step of learning a phonogramic language. However, it is demanding for L2 learners to familiarize the whole inventory of graphemes in advance if the language has a relatively large inventory size. We propose that learning a subset of …...

A Winograd schema is a pair of twin sentences containing a referential ambiguity that is easy for a human to resolve but difficult for a com- puter. This paper explores the characteristics of world knowledge necessary for resolving such a schema. We observe that people tend to avoid ambiguous antecedents …...

The study of second language speech perception usually put L1-L2 phonological mapping as the rule of thumb in predicting learning outcome, and seldom included more fine-grained aspects such as frequency. This study examines how frequency of sounds in L1 may influence L2 segmental production and perception, with examples from English …...

State of the art statistical machine translation systems are typically trained by symmetrizing word alignments in two translation directions. We introduce a new method that improves word alignment results, based on self learn- ing using the initial symmetrized word align- ments results. The method involves align- ing words and symmetrizing …...

This work is concerned with incrementally training statistical machine translation (SMT) models when new data becomes available. That, in contrast to re-training new models based on the entire accumulated data. In- cremental training provides a way to per- form faster, more frequent model updates, en- abling keeping the SMT system …...

Statistical machine translation (SMT) suffers from the accuracy problem that the translation pairs and their feature scores in the transla- tion model can be inaccurate. The accuracy problem is caused by the quality of the unsu- pervised methods used for translation model learning. Previous studies propose estimating comparable features for …...

. We address the problem of transferring semantic annotations, more specifically predicate labellings, from one language to another using parallel corpora. Previous work has transferred these annotations directly at the token level, leading to low recall. We present a global approach to annotation transfer that aggregates information across the whole …...

By analyzing corpus data, we have shown that the tendencies of restricting perfective past marking to Accomplishments and Achievements and imperfective marking to Statives and Activities as described by the Aspect Hypothesis (Shirai, 1991; Andersen & Shirai, 1996), undesirable in the acquisition of various languages, are desirable in the acquisition …...

. Information Extraction systems must cope with two problems : they heavily depend on the considered domain but the cost of development for a domain-specific system is important. We propose a new solution for role la- beling in the event-extraction task that relies on using unsupervised word representations (word embeddings) …...

The current paper proposes a novel approach to Spanish zero pronoun resolution in the context of Spanish to Korean Machine Translation (MT). Spanish is one of the well-known 'pro-drop' languages so that especially a subject pronoun is often omitted, if it can be inferred from the linguistic as well as …...

Crowdsourced data annotation is noisier than annotation from trained workers. Previ- ous work has shown that redundant annota- tions can eliminate the agreement gap be- tween crowdsource workers and trained work- ers. Redundant annotation is usually non- problematic because individual crowdsource judgments are inconsequentially cheap in a class-balanced dataset. However, …...

Annotating a corpus with error information is a challenging task. This paper describes the design, evaluation and refinement of an an- notation scheme for Spanish article errors in learner data, so that future work on corpus an- notation and automatic article error detection can progress. To evaluate reliability, 300 noun …...

Several studies contend that the main motivation for scrambling is heaviness. In particular, Yamashita (2002) maintains that scrambling has nothing to do with givenness and that heaviness is the primary factor for scrambling. However, her conclusions count on only 19 examples and she does not distinguish VP-internal scrambling from VP-external …...

It has been widely acknowledged that the choice of Japanese demonstratives (the distal a-series, the medial so-series, and the prox- imal ko-series) in their anaphoric use is reg- ulated by the rules concerned with the inter- locutors’ knowledge of the referent. In cross- linguistic discussions of anaphoric demon- stratives, on …...

Sentiment analysis has become an important classification task because a large amount of user-generated content is published over the Internet. Sentiment lexicons have been used successfully to classify the sentiment of user review datasets. More recently, microblog- ging services such as Twitter have become a popular data source in the …...

The popularity of the user generated content, such as Twitter, has made it a rich source for the sentiment analysis and opinion min- ing tasks. This paper presents our study in automatically building a training corpus for the sentiment analysis on Indonesian tweets. We start with a set of seed …...