Francisco Oliveira

Combining Domain Adaptation Approaches for Medical Text Translation

This paper explores a number of simple and effective techniques to adapt statistical machine translation (SMT) systems in the medical domain. Comparative experiments are conducted on large corpora for six language pairs. We not only compare each adapted system with the baseline, but also combine them to further improve the domain-specific systems. Finally, we attend the WMT2014 medical summary sentence translation constrained task and our systems achieve the best BLEU scores for Czech-English, EnglishGerman, French-English language pairs and the second best BLEU scores for reminding pairs.

Systematic processing of long sentences in rule based portuguese-chinese machine translation

The translation quality and parsing efficiency are often disappointed when Rule based Machine Translation systems deal with long sentences. Due to the complicated syntactic structure of the language, many ambiguous parse trees can be generated during the translation process, and it is not easy to select the most suitable parse tree for generating the correct translation. This paper presents an approach to parse and translate long sentences efficiently in application to Rule based Portuguese-Chinese Machine Translation. A systematic approach to break down the length of the sentences based on patterns, clauses, conjunctions, and punctuation is considered to improve the performance of the parsing analysis. On the other hand, Constraint Synchronous Grammar is used to model both source and target languages simultaneously at the parsing stage to further reduce ambiguities and the parsing efficiency.

A relationship: word alignment, phrase table, and translation quality.

In the last years, researchers conducted several studies to evaluate the machine translation quality based on the relationship between word alignments and phrase table. However, existing methods usually employ ad-hoc heuristics without theoretical support. So far, there is no discussion from the aspect of providing a formula to describe the relationship among word alignments, phrase table, and machine translation performance. In this paper, on one hand, we focus on formulating such a relationship for estimating the size of extracted phrase pairs given one or more word alignment points. On the other hand, a corpus-motivated pruning technique is proposed to prune the default large phrase table. Experiment proves that the deduced formula is feasible, which not only can be used to predict the size of the phrase table, but also can be a valuable reference for investigating the relationship between the translation performance and phrase tables based on different links of word alignment. The corpus-motivated pruning results show that nearly 98% of phrases can be reduced without any significant loss in translation quality.

Query Translation for Cross-Language Information Retrieval by Parsing Constraint Synchronous Grammar

With the availability of large amounts of multilingual documents, cross-language information retrieval (CLIR) has become an active research area in recent years. However, researchers often face with the problem of inherent ambiguities involved in natural languages. Moreover, this task is even more challenging for processing the Chinese language because word boundaries are not defined in the sentence. This paper presents a Chinese-Portuguese query translation for CLIR based on a machine translation (MT) system that parses constraint synchronous grammar (CSG). Unlike traditional transfer-based MT architectures, this model only requires a set of CSG rules for modeling syntactic structures of two languages simultaneously to perform the translation. Moreover, CSG can be used to remove different levels of disambiguation as the parsing processes in order to generate a translation with quality.

Overcoming Data Sparseness Problem in Statistical Corpus Based Sense Disambiguation

The problem of data sparseness is considered as a common problem for Statistical Corpus based Sense Disambiguation approaches [1]. Usually large amounts of data are required in the corpus to guarantee that all senses of an ambiguous word are presented. However, this is not easily achieved, especially for words that do not occur frequently in the training corpus. On the other hand, for languages that do not have large amount of digitized resources, the sparseness problem is even worse.

Comparison of Rumen Fluid pH by Continuous Telemetry System and Bench pH Meter in Sheep with Different Ranges of Ruminal pH

We aimed to compare the measurements of sheep ruminal pH using a continuous telemetry system or a bench pH meter using sheep with different degrees of ruminal pH. Ruminal lactic acidosis was induced in nine adult crossbred Santa Ines sheep by the administration of 15u2009g of sucrose per kg/BW. Samples of rumen fluid were collected at the baseline, before the induction of acidosis (T 0) and at six, 12, 18, 24, 48, and 72 hours after the induction for pH measurement using a bench pH meter. During this 72-hour period, all animals had electrodes for the continuous measurement of pH. The results were compared using the Bland-Altman analysis of agreement, Pearson coefficients of correlation and determination, and paired analysis of variance with Students t-test. The measurement methods presented a strong correlation (r = 0.94, P < 0.05) but the rumen pH that was measured continuously using a telemetry system resulted in lower values than the bench pH meter (overall mean of 5.38 and 5.48, resp., P = 0.0001). The telemetry system was able to detect smaller changes in rumen fluid pH and was more accurate in diagnosing both subacute ruminal lactic acidosis and acute ruminal lactic acidosis in sheep.

Clinical response and transfusion reactions of sheep subjected to single homologous blood transfusion.

Studies in relation to blood conservation and responses to transfusion are scarce for ruminants. We evaluated the clinical manifestations of sheep that received a single homologous transfusion of whole blood, focusing on transfusion reactions. Eighteen adult sheep were subjected to a single phlebotomy to withdraw 40% of the total blood volume, which was placed into CPDA-1 bags and then divided into G0, animals that received fresh blood, and G15 and G35, animals that received blood stored for 15 or 35 days, respectively. Clinical observations were recorded throughout the transfusion, whereas heart rate, respiratory rate, and rectal temperature were assessed at the following times: 24 hours after phlebotomy and before transfusion; 30 minutes, six, twelve, 24, 48, 72, and 96 hours and eight and 16 days after transfusion. All groups presented transfusion reactions, among which hyperthermia was the most frequent (50% of animals). Tachycardia occurred most frequently in the G35 animals (50% of them). During transfusion G35 animals presented more clinical manifestation (P < 0.05). Transfusion of fresh or stored total blood improved the blood volume, but transfusion reactions occurred, demonstrating that a single transfusion of fresh or stored blood can cause inflammatory and febrile nonhemolytic transfusion reactions in sheep.

Unsupervised word sense disambiguation and rules extraction using non-aligned bilingual corpus

This paper presents a statistical word sense disambiguation with application in Portuguese-Chinese machine translation systems. Due to the limited availability of Portuguese-Chinese resources in the form of digital corpora and annotated Treebank, an unsupervised learning and a non-aligned bilingual corpus are applied. The proposed method first identifies words related to each of the ambiguous words based on their surrounding words and relative distance. A mathematical model is then applied in the identification of the most suitable sense of an ambiguous word in terms of the related words. All the senses discovered are converted into a set of rules and stored in the sense knowledge base for later use in disambiguation and translation process. Preliminary experiment results show an improvement of 6% in assigning correctly the corresponding translation over the baseline method.

Prevalence of antibodies against Toxoplasma gondii and Neospora spp. in equids of Western Pará, Brazil

The present study aimed to determine the prevalence of antibodies against Toxoplasma gondii and Neospora spp. in equids raised for distinct purposes in the western state of Pará, Brazil, and to identify potential risk factors associated with parasite infection. A cross-sectional study was conducted with cluster sampling in 101 farms from 18 municipalities (farm horses). In visited municipalities, samples from sport and cart horses were included. Serum was obtained for detecting antibodies against T. gondii and Neospora spp. using indirect fluorescent antibody test, with a cut-off point of 1:64 and 1:50, respectively. Association analysis, using the chi-square test, was performed to evaluate possible risk factors related to the prevalence of antibodies. A total of 1298 equids were sampled (947 farm, 240 sport, and 111 cart horses), including 1244 horses, 2 donkeys, and 52 mules. The number of equids sampled per property ranged from 1 to 49. Of the total examined, 134 (10.3%) had antibodies against T. gondii, and of the 18 municipalities visited, 11 (61%) had at least one positive animal. Seventy-three of the 1298 equids (5.62%) from 44% of the municipalities tested positive for Neospora spp. The type of the animal (farm, sport, or carthorse) was not associated with the prevalence of antibodies against T. gondii; however, antibodies against Neospora spp. was more prevalent in carthorses (Pu2009=u20090.001) (13.5%) than in farm (5.0%) and sport (4.6%) horses. Breed of the horse and the presence of cats were risk factors for the prevalence of antibodies against T. gondii, and presence of dogs was a risk factor for the prevalence of antibodies against Neospora spp. This is the first report of seropositive equids for T. gondii and Neospora spp. in the state of Pará contributing to the knowledge on epidemiology of the two parasites.

Annotation System for the Construction of Synchronous Grammar Tree Alignment Relationships

The construction of syntactic tree structures is vital to different Natural Language Processing applications. In the meanwhile, the use of monolingual or bilingual structures directly affects the quality of Machine Translation systems. However, manually annotated syntactic tree structures are not only considered as a time consuming task but also a very expensive work, and automatic construction approaches cannot always guarantee the quality of the syntactic trees. In this paper, a system for annotating and constructing synchronous grammar tree structures in a semi-automatic way is proposed. The system is built in the Web environment with a graphical display for users to review and modify alignment relationships between nodes of the tree and strings or trees of the other language. The core part relies on the parsing of Constraint Synchronous Grammar, and consists of several modules in establishing alignments at different levels between the languages, including relationships between syntactic tree and strings of the other language, and bilingual tree alignments. Moreover, it provides import functions in obtaining monolingual skeletal bracketing syntactic tree and Translation Corresponding Tree structures for the creation of synchronous rules in order to have wider applicability.