Search

Citation recommendation is a task that aims to automatically select suitable references for a working manuscript. This task has become increasingly urgent as the typical pools of candidates continue to grow, in the order of tens or hundreds of thousands or more. While several approaches to citation recommendation have been proposed in the literature, they generally seem to lack principled mechanisms to ensure diversity and other global properties among the recommended citations. For this reason, in this paper we propose a novel citation recommendation approach that leverages a submodular scoring function and a deep document representation to achieve an effective trade-off between relevance to the query and diversity of the references. To optimally train the scoring function and the deep representation, we propose a novel training objective based on a structural/multiclass hinge loss and incremental recommendations. The experimental results over three popular citation datasets have showed that the proposed approach has led to remarkable accuracy improvements, with an increase of up to 1.91 pp of MRR and 3.29 pp of F1@100 score with respect to a state-of-the-art citation recommendation system.

The exponential growth of social media has created informative content, which leads to the emersion of social listening tools, in order to serve the need of companies and organizations. These tools are developed mainly in English language and for those which deal with Vietnamese, they lack the analysis capabilities to capture insightful information. In our research, we propose a framework to automatically extract and analyze information from social media, including online news and social networks. Our framework consists of multiple modules: data crawler, analysis module (topic analysis, sentiment analysis, event detection), visualization module. The results for analysis modules are promising: F1-score for topic analysis is 53%, average accuracy for sentiment analysis (using Maximum Entropy with three classes) is 62% and F1-score for event detection is 88.36%.

Automatically building a large bilingual corpus that contains millions of words is always a challenging task. In particular in case of low-resource languages, it is difficult to find an existing parallel corpus which is large enough for building a real statistical machine translation. However, comparable non-parallel corpora are richly available in the Internet environment, such as in Wikipedia, and from which we can extract valuable parallel texts. This work presents a framework for effectively extracting parallel sentences from that resource, which results in significantly improving the performance of statistical machine translation systems. Our framework is a bootstrapping-based method that is strengthened by using a new measurement for estimating the similarity between two bilingual sentences. We conduct experiment for the language pair of English and Vietnamese and obtain promising results on both constructing parallel corpora and improving the accuracy of machine translation from English to Vietnamese.

We can observe the rapid development of the spoken human-machine interface, thanks to the big progress in automatic recognition and text to speech technologies. Especially, the mobile virtual assistants are becoming more popular than ever. Recognizing user's intent in interaction is one of the biggest problems while the system is designed and is attracting the attention of researchers. Identifying utterance's speech act in an automated manner is important to reveal user's intention (such as informing, asking, requesting and expressing emotion), which can provide useful indicators to improve the performance of human-machine interaction. Automatic speech act classification has been studied in many languages such as English, Chinese, Slovakian, Arabic, but not yet in Vietnamese. In this paper, we present a speech act scheme suitable for Vietnamese utterances in mobile voice inter- action. Also, we built a speech act classifier that achieved 84.36% accuracy. This classifier has been applied in our mobile virtual assistant for Vietnamese (VAV). The experiment result has demonstrated the ability of this classifier such as robust, compact and can work well on mobiles.

Most recent Vietnamese QA systems have not considered so far in using the data crawled from the community web services as a useful resource. In this paper, we take into accounts the community-based resource to build a Vietnamese question answering system named VnCQAs. Our system comprises of three modules for building the database of question-answer pairs, analyzing questions and choosing the best answer respectively. Experimental results show that our system achieves promising performances.

This paper describes the approach for analyzing questions in our community-based Vietnamese question answering system (VnCQAs), in which we focus on two subtasks: question classification and keyword identification. The question classification employs the machine learning approaches with a feature which represents a measure of similarity between two questions, while the keyword identification uses the dependency-tree-based features. Experimental results are promising, in which the question classification obtains the accuracy of 95.7% and the keyword identification gains the accuracy of 85.8%. Furthermore, these two subtasks help to improve the accuracy for finding the similar questions in our VnCQAs by 6.75%.

In this paper, we describe the TATO system which participated in the SemEval-2015 Task 2a: “Semantic Textual Similarity (STS) for English”. Our system is trained on published datasets from the previous competitions. Based on some machine learning techniques, it combines multiple similarity measures of varying complexity ranging from simple lexical and syntactic similarity measures to complex semantic similarity ones to compute semantic textual similarity. Our final model consists of a simple linear combination of about 30 main features out of a numerous number of features experimented. The results are promising, with Pearson’s coefficients on each individual dataset ranging from 0.6796 to 0.8167 and an overall weighted mean score of 0.7422, well above the task baseline system.

In this paper, we propose a new approach to construct a system of transformation rules for the Part-of-Speech (POS) tagging task. Our approach is based on an incremental knowledge acquisition method where rules are stored in an exception structure and new rules are only added to correct the errors of existing rules; thus allowing systematic control of the interaction between the rules. Experimental results on 13 languages show that our approach is fast in terms of training time and tagging speed. Furthermore, our approach obtains very competitive accuracy in comparison to state-of-the-art POS and morphological taggers., Version 1: 13 pages. Version 2: Submitted to AI Communications - the European Journal on Artificial Intelligence. Version 3: Resubmitted after major revisions. Version 4: Resubmitted after minor revisions. Version 5: to appear in AI Communications (accepted for publication on 3/12/2015)

This paper describes our robust, easyto-use and language independent toolkit namely RDRPOSTagger which employs an error-driven approach to automatically construct a Single Classification Ripple Down Rules tree of transformation rules for POS tagging task. During the demonstration session, we will run the tagger on data sets in 15 different languages.

This paper presents a new conversion method to automatically transform a constituent-based Vietnamese Treebank into dependency trees. On a dependency Treebank created according to our new approach, we examine two state-of-the-art dependency parsers: the MSTParser and the MaltParser. Experiments show that the MSTParser outperforms the MaltParser. To the best of our knowledge, we report the highest performances published to date in the task of dependency parsing for Vietnamese. Particularly, on gold standard POS tags, we get an unlabeled attachment score of 79.08% and a labeled attachment score of 71.66%.

Previous researches mainly focus on the approaches which are essentially inspirited from the log-linear model background in machine learning or other adaptations. However, not a lot of studies deeply focus on improving word-alignment models to enhance the quality of phrase translation table. This research will follow on that approach. The experiments show that this scheme could also improve the quality of the word-alignment component better. Hence, the improvement impacts the quality of translation system in overall around 1% for the BLEU score metric.

In this paper, we present a Vietnamese Natural Language Inter face to a survey dat abase for individuals and businesses who want to know economic information from economic surveys. We carry out analysis of various Vietnamese question types and investigate the st ability of our approach using GATE framework and R language. Our system uses R language to specifically deal with statistical question types. Experiment a results are very promising where our system achieves an accuracy of 78.5% on economic survey dat a domain.

In statistical phrase-based machine translation, the step of phrase learning heavily relies on word alignments. This paper provides a systematic comparison of applying various statistical alignment models for statistical English-Vietnamese phrase-based machine translation. We will also invest a heuristic method for elevating the translation quality of using higher word-alignment models by improving the quality of lexical modelling. In detail, we will experimentally show that taking up the lexical translation seems to be an appropriate approach to force "higher" word-based translation models be able to efficiently "boost" their merits. We hope this work will be a reliable comparison benchmark for other studies on using and improving the statistical alignment models for English-Vietnamese machine translation systems.

Advertising has appeared in almost all areas of life. The large number of advertisements, especially in real estate domain, has raised a need for an effective way to search and find useful information for users. In this paper, we propose a rule-based approach to build an Information Extraction system for Vietnamese online real estate advertisements. Experimental results shows that our approach is very promising with an overall F-measure of 91% on a collection of data collected from popular Vietnamese real estate websites.

Under word-based alignment, frequent words with consistent translations can be aligned at a high rate of precision. However, the words that are less frequent or exhibit diverse translations in training corpora generally do not have statistically significant evidences for confident alignments [7]. In this work, we will focus on proposing a bootstrapping algorithm to capture those less frequent or exhibit diverse alignments. Interestingly, we avoid making any explicit assumption concerning with the pair of languages used. As the result, we take the experimental evaluations on two phrase-based translation systems: the English-Vietnamese and English-French translation systems. Experiments point out a significant "boosting" capacity for the quality in overall for both these tasks.

One of the biggest problems with rule based systems is how to avoid the conflict between rules when a new rule is added. Ripple Down Rules (RDR) is considered a good systematic approach to address this for classification problems. In this paper, we present a system using RDR to build the set of rules for Vietnamese Named Entity Recognition which is important for many natural language processing tasks. Experimental results on comparing the proposed approach with a standard method where rules are added in an ad-hoc manner prove to be very promising.

Ontologies have served as a knowledge representation about the whole world or some part of it. Building ontologies is a challenging and active research area. Manually constructed Ontologies often have higher quality than the ones created by automatic or semi-automatic approaches but they tend to be more applicable to small domains. Automatic approaches are considered more suitable for building large scale Ontologies where time and efforts of human experts become a bottleneck. For both paradigms, approaches to building Ontologies from Vietnamese texts are still very limited. In this paper, we propose a system that automatically builds Ontology from Vietnamese texts using cascades of annotation-based grammars. Obtained experimental results on a university organizational structure domain are very promising.

This paper presents a new approach to learn a rule based system for the task of part of speech tagging. Our approach is based on an incremental knowledge acquisition methodology where rules are stored in an exception-structure and new rules are only added to correct errors of existing rules; thus allowing systematic control of interaction between rules. Experimental results of our approach on English show that we achieve in the best accuracy published to date: 97.095% on the Penn Treebank corpus. We also obtain the best performance for Vietnamese VietTreeBank corpus.

Sentiment analysis is one of the most important tasks in Natural Language Processing. Research in sentiment analysis for Vietnamese is relatively new and most of current work only focus in document level. In this paper, we address this problem at the sentence level and build a rule-based system using the Gate framework. Experimental results on a corpus of computer products reviews are very promising. To the best of our knowledge, this is the first work that analyzes sentiment at sentence level in Vietnamese.

Word segmentation is one of the most important tasks in NLP. This task, within Vietnamese language and its own features, faces some challenges, especially in words boundary determination. To tackle the task of Vietnamese word segmentation, in this paper, we propose the WS4VN system that uses a new approach based on Maximum matching algorithm combining with stochastic models using part-of-speech information. The approach can resolve word ambiguity and choose the best segmentation for each input sentence. Our system gives a promising result with an F-measure of 97%, higher than the results of existing publicly available Vietnamese word segmentation systems.

In this paper, we present our method of using Information Extraction techniques to tackle the task of automatically translating English weather bulletins to Vietnamese. It is simple yet effective in satisfying the constraints of low processing power and storage space for the deployment on an embedded system. Experimental results are very promising with the F-measure going up to 96% for extracting relevant information from the weather bulletins.

In this chapter we discuss ways of assisting experts to develop complex knowledge bases for a variety of natural language processing tasks. The proposed techniques are embedded into an existing knowledge acquisition framework, KAFTIE, specifically designed for building knowledge bases for natural language processing. Our intelligent agent, the rule suggestion module within KAFTIE, assists the expert by suggesting new rules in order to address incorrect behavior of the current knowledge base. The suggested rules are based on previously entered rules which were “hand-crafted” by the expert. Initial experiments with the new rule suggestion module are very encouraging as they resulted in a more compact knowledge base of comparable quality to a fully hand-crafted knowledge base. At the same time the development time for the more compact knowledge base was considerably reduced.

We present KAFTIE - an incremental knowledge acquisition framework which utilizes expert knowledge to build high quality knowledge base annotators. Using KAFTIE, a knowledge base was built based on a small data set that outperforms machine learning algorithms trained on a much bigger data set for the task of recognizing temporal relations. In particular, this can be incorporated to bootstrap the process of labeling data for domains where annotated data is not available.

The aim of our work is to develop a flexible and powerful Knowledge Acquisition framework that allows users to rapidly develop Natural Language Processing systems, including information extraction systems. Tasks on which we experimented with our framework are to identify concepts/terms of which positive or negative aspects are mentioned in scientific papers. The results so far are very promising as we managed to build systems with relative ease that achieve F-measures of around 84% on a corpus of scientific papers in the area of artificial intelligence.

The aim of our work is to develop a flexible and powerful Knowledge Acquisition framework that allows users to rapidly develop Natural Language Processing systems, including information extraction systems. In this paper we present our knowledge acquisition framework, KAFTIE, which strongly supports the rapid development of complex knowledge bases for information extraction. We specifically target scientific papers which involve rather complex sentence structures from which different types of information are automatically extracted. Tasks on which we experimented with our framework are to identify concepts/terms of which positive or negative aspects are mentioned in scientific papers. These tasks are challenging as they require the analysis of the relationship between the concept/term and its sentiment expression. Furthermore, the context of the expression needs to be inspected. The results so far are very promising as we managed to build systems with relative ease that achieve F-measures of around 84% on a corpus of scientific papers in the area of artificial intelligence.

The aim of our work is to provide support for reading (or skimming) scientific papers. In this paper we report on the task to identify concepts or terms with positive attributions in scientific papers. This task is challenging as it requires the analysis of the relationship between a concept or term and its sentiment expression. Furthermore, the context of the expression needs to be inspected. We propose an incremental knowledge acquisition framework to tackle these challenges. With our framework we could rapidly (within 2 days of an expert’s time) develop a prototype system to identify positive attributions in scientific papers. The resulting system achieves high precision (above 74%) and high recall rates (above 88%) in our initial experiments on corpora of scientific papers. It also drastically outperforms baseline machine learning algorithms trained on the same data.

Our research aims at interactive document viewers that can select and highlight relevant text passages on demand. Another related objective is the generation of topic-specific summaries of texts as opposed to general purpose summaries. This paper introduces our notions of discourse structure tree and level-of-detail tree. Both structures are used to represent relevant aspects of a text segment for the above mentioned purposes. Furthermore, we introduce a Knowledge Acquisition Framework for DIScourse processing (KAFDIS) that allows the incremental and efficient acquisition of knowledge for the reliable construction of the discourse structure graph and the level-of-detail tree based on cue phrases. An effective knowledge acquisition process is crucial to allow the economical development of systems that can handle a large variety of topics. Our knowledge acquisition approach ensures always a consistent knowledge base whose semantics are well controlled by the expert. It is an incremental approach that allows patching of the knowledge base as soon as the need arises without causing any inconsistencies. We also present promising experimental results with our approach.

This paper introduces a new method for the rapid development of complex rule bases involving cue phrases for the purpose of classifying text segments. The method is based on Ripple-Down Rules, a knowledge acquisition method that proved very successful in practice for building medical expert systems and does not require a knowledge engineer. We implemented our system KAFTAN and demonstrate the applicability of our method to the task of classifying scientific citations. Building cue phrase rules in KAFTAN is easy and efficient. We demonstrate the effectiveness of our approach by presenting experimental results where our resulting classifier clearly outperforms previously built classifiers in the recent literature.

The dominant approach for developing ontologies is currently based on description logics. We offer here a different approach, MOSS, based on semantic nets. The simple frame-based formalism that we are using allows specifying multilingual concepts easily and includes features like virtual classes and properties equivalent to rules in other approaches. Reasoning is done by traversing the graph of classes and instances in an efficient manner thanks to indexes. When restricted, the approach can be compiled into the OWL-DL formalism augmented with rules compatible with the JENA framework or using SPARQL queries. MOSS has been used in several applications including a multi-agent platform.