Hernani Costa

Universidad de Málaga, Faculty of Philosophy and Letters, Early Stage Researcher

Followers

121

Following

106

Co-authors

Public Views

InterestsView All (16)

Uploads

Books by Hernani Costa

Proceedings of the EXPERT Scientific and Technological Workshop

by Hernani Costa, Gloria Corpas Pastor, and Miriam Seghiri

EXPERT (EXPloiting Empirical appRoaches to Translation): http://expert-itn.eu

Download

Automatic Extraction and Validation of Lexical Ontologies from text: Creating Lexical Ontologies from text

Nowadays, semantic information plays an important role in NLP, more specifically describing and r... more Nowadays, semantic information plays an important role in NLP, more specifically describing and representing the meanings of the words. In the last two decades, there have been efforts to create a large database that represents lexical knowledge. However, in most of the cases, this resources are created manually. For instance Princeton WordNet is considered the standard model of a lexical ontology for the English language. Besides that, also for Portuguese there have been some attempts to create a broad-coverage ontology, also created manually and not publicly available. Still, they are not public available for download, and also all of them were manually created. Despite being less prone to errors, the problem is that the manual creation of these resources takes a lot of time consuming and requires a team, and researchers specialised in the area. Having this in mind, this book describes how to create a system capable of automatically acquire semantic knowledge from any kind of Portuguese text. In addition, it is analysed the benefits from applying similarity distributional metrics based on the occurrence of words in documents to our system outputs.

Papers by Hernani Costa

Compiling Specialised Comparable Corpora. Should we always thrust (Semi-)automatic Compilation Tools?

Decisions at the outset of compiling a comparable corpus are of crucial importance for how the co... more Decisions at the outset of compiling a comparable corpus are of crucial importance for how the corpus is to be built and analysed later on.
Several variables and external criteria are usually followed when building a corpus but little is been said about textual distributional similarity in this context and the quality that it brings to research. In an attempt to fulfil this gap, this paper aims at presenting a simple but efficient methodology capable of measuring a corpus internal degree of relatedness. To do so, this methodology takes advantage of both available natural language processing technology and statistical methods in a successful attempt to access the relatedness degree between documents. Our findings prove that using a list of common entities and a set of distributional similarity measures is enough not only to describe and assess the degree of relatedness between the documents in a comparable corpus, but also to rank them according to their degree of relatedness within the corpus.

Download

Measuring the Relatedness between Documents in Comparable Corpora

by Hernani Costa and Gloria Corpas Pastor

This paper aims at investigating the use of textual distributional similarity measures in the con... more This paper aims at investigating the use of textual distributional similarity measures in the context of comparable corpora. We address the issue of measuring the relatedness between documents by extracting, measuring and ranking their common content. For this purpose, we designed and applied a methodology that exploits available natural language processing technology with statistical methods. Our findings showed that using a list of common entities and a simple, yet robust set of distributional similarity measures was enough to describe and assess the degree of relatedness between the documents. Moreover, our method has demonstrated high performance in the task of filtering out documents with a low level of relatedness. By a way of example, one of the measures got 100%, 100%, 95% and 90% precision when injected 5%, 10%, 15% and 20% of noise, respectively.

Download

Comparing Approaches to the Identification of Similar Languages

Joint Workshop on Language Technology for Closely Related Languages, Varieties and Dialects (LT4VarDial'15). 2nd Discriminating between Similar Languages Shared Task (DSL'15), Sep 9, 2015

This paper describes the submission made by the MMS team to the Discriminating between Similar La... more This paper describes the submission made by the MMS team to the Discriminating between Similar Languages (DSL) shared task 2015. We participated in the closed submission track using only the dataset provided by the shared task organisers which contained short texts from 13 similar languages and language varieties. We submitted three runs using different systems and compare their performance. As a result, our best system achieved 95.24% accuracy for test set A (containing original texts) and 92.78% accuracy for test set B (containing texts without named entities).

Download

Assessing Comparable Corpora through Distributional Similarity Measures

Describing, comparing and evaluating corpora are key issues in corpus-based translation and corpu... more Describing, comparing and evaluating corpora are key issues in corpus-based translation and corpus linguistics for which there is still a notable lack of standards. Bearing this in mind, this paper aims at investigating the use of textual distributional similarity measures in the context of comparable corpora. More precisely, we address the issue of measuring the relatedness between documents by extracting and measuring their common content. For this purpose, we designed and applied a methodology that exploits available natural language processing technology with statistical methods. Our findings showed that using a list of common entities and a simple, yet robust and high performance set of distributional similarity measures was enough to describe and assess the degree of relatedness between the documents in a comparable corpus.

Download

MiniExperts: An SVM Approach for Measuring Semantic Textual Similarity

by Hernani Costa, Constantin Orasan, and Gloria Corpas Pastor

9th Int. Workshop on Semantic Evaluation (SemEval'15), Jun 4, 2015

This paper describes the system submitted by the University of Wolverhampton and the University o... more This paper describes the system submitted by the University of Wolverhampton and the University of Malaga for SemEval-2015 Task 2: Semantic Textual Similarity. The system uses a Supported Vector Machine approach based on a number of linguistically motivated features. Our system performed satisfactorily for English and obtained a mean 0.7216 Pearson correlation. However, it performed less adequately for Spanish, obtaining only a mean 0.5158.

Download

iCorpora: Compiling, Managing and Exploring Multilingual Data

by Miriam Seghiri and Hernani Costa

In the last decade, there has been a growing interest in bilingual and multilingual corpora. Part... more In the last decade, there has been a growing interest in bilingual and multilingual corpora. Particularly, in translation their benefits have been demonstrated by several authors (cf. Bowker and Pearson, 2002; Bowker, 2002; Zanettin et al., 2003; Corpas Pastor and Seghiri, 2009). Their objectivity, reusability, multiplicity and applicability of uses, easy handling and quick access to large volume of data are just an example of their advantages. Thus, it is not surprising that the use of corpora has been considered an essential resource in several research domains such as translation, language learning, stylistics, sociolinguistics, terminology, language teaching, automatic and assisted translation, amongst others. Nevertheless, the lack of sufficient/up-to-date parallel corpora and linguistic resources for narrow domains and poorly-resourced languages is currently one of the major obstacles to further advancement on these areas. One potential solution to the insufficient parallel translation data is the exploitation of non-parallel bilingual and multilingual text resources, also known as comparable corpora (i.e. corpora that include similar types of original texts in one or more language using the same design criteria (cf. EAGLES, 1996; Corpas Pastor, 2001:158)). Even though comparable corpora can compensate for the shortage of linguistic resources and ultimately improve automated translations quality for under-resourced languages and narrow domains for example, the problem of data collection presupposes a significant technical challenge. The solution proposed in iCorpora project and presented in this article is to exploit the fact that comparable corpora are much more widely available than parallel translation data. This ongoing project aims to increase the flexibility and robustness of the compilation, management and exploration of both comparable and parallel corpora by creating a new web-based application from scratch.

Download

iCompileCorpora: A Web-based Application to Semi-automatically Compile Multilingual Comparable Corpora

by Gloria Corpas Pastor, Hernani Costa, and Miriam Seghiri

This article presents an ongoing project that aims to design and develop a robust and agile web-b... more This article presents an ongoing project that aims to design and develop a robust and agile web-based application capable of semi-automatically compiling monolingual and multilingual comparable corpora, which we named iCompileCorpora. The dimensions that comprise iCompileCorpora can be represented in a layered model comprising a manual, a semi-automatic and a Cross-Language Information Retrieval (CLIR) layer. This design option will not only permit to increase the flexibility of the compilation process, but also to hierarchically extend the manual layer features to the semi-automatic web-based layer and then to the semi-automatic CLIR layer. The manual layer presents the option of compiling monolingual or multilingual corpora. It will allow the manual upload of documents from a local or remote directory onto the platform. The second layer will permit the exploitation of either monolingual or multilingual corpora mined from the Internet. As nowadays there is an increasing demand for systems that can somehow cross the language boundaries by retrieving information of various languages with just one query, the third layer aims to answer this demand by taking advantage of CLIR techniques to find relevant information written in a language different from the one semi-automatically retrieved by the methodology used in the previous layer.

Download

A comparative User Evaluation of Terminology Management Tools for Interpreters

by Hernani Costa, Gloria Corpas Pastor, and Isabel Durán-Muñoz

When facing new fields, interpreters need to perform extensive searches for specialised knowledge... more When facing new fields, interpreters need to perform extensive searches for specialised knowledge and terminology. They require this information prior to an interpretation and have it accessible during the interpreting service. Fortunately, there are currently several terminology management tools capable of assisting interpreters before and during an interpretation service. Although these tools appear to be quite similar, they provide different kind of features and as a result they exhibit different degrees of usefulness. This paper aims at describing current terminology management tools with a view to establishing a set of features to assess the extent to which terminology tools meet the specific needs of the interpreters. Subsequently, a comparative analysis is performed to evaluate these tools based on the list of features previously identified.

Download

Overcoming Information Overload with Artificial Selective Agents

by Hernani Costa and Luis Macedo

We describe an approach, based on artificial forms of selective attention, for overcoming the pro... more We describe an approach, based on artificial forms of selective attention, for overcoming the problem of information and interruption overload of intelligent agents. Inspired on natural selective attention studies, we propose a computational model of selective attention that relies on the assumption that uncertain, surprising and motive congruent/incongruent information demands attention from an intelligent agent.

Download

Overcoming Information Overload with Artificial Selective Agents: an Application to Travel Information Domain

by Hernani Costa and Luis Macedo

We describe an application of Macedo's computational model of selective attention for overcoming ... more We describe an application of Macedo's computational model of selective attention for overcoming the problem of information and interruption overload of intelligent agents in travel information systems. This computational model has been integrated into the architecture of a BDI artificial agent so that this can autonomously select relevant, interesting travel information of the (external or internal) environment while ignoring other less relevant information. The advantage is that the agent can communicate only that interesting, selective information to its processing resources (focus of the senses, decision-making, etc.) or to its human owner's processing resources so that these resources can be allocated more effectively. We illustrate and provide experimental results of this role of the artificial, selective attention mechanism in the travel domain.

Download

Technology-assisted Interpreting

by Hernani Costa, Gloria Corpas Pastor, and Isabel Durán-Muñoz

Unlike translators, for whom a myriad of computer-assisted tools are available, interpreters have... more Unlike translators, for whom a myriad of computer-assisted tools are available, interpreters have not benefited from the same level of automation or innovation. Their work relies by and large on traditional or manual methods. The solutions tailored to the interpreters' needs are few and still far behind. Fortunately, there is a growing interest in developing tools addressed at interpreters as end users, although the number of these technology tools is still very low and they are not intended to cover all interpreters' needs.

Download

A Selective Attention-based, Multi-Agent, Travel Information System

by Hernani Costa and Luis Macedo

We describe a selective attention-based, multi-agent, travel information system. This system invo... more We describe a selective attention-based, multi-agent, travel information system. This system involves a master agent and personal agents. The master agent collects selectively information from several information sources and sends it to the personal agents so that they can selectively deliver information to the several mobile devices owned by humans. The main agents’ models are described namely those of beliefs, desires, feelings, and, with special emphasis, the computational model of selective attention. Then, we describe an experiment to evaluate the performance and the potential benefits of these personal selective attention agents for filtering out unnecessary traffic information for their human owners.

Download

Recommending POIs based on the User’s Context and Intentions

by Hernani Costa and Bárbara Furtado

Highlights on Practical Applications of Agents and Multi-Agent Systems, May 2013

This paper describes a Recommender System that implements a Multiagent System for making personal... more This paper describes a Recommender System that implements a Multiagent System for making personalised context and intention-aware recommendations of Points of Interest (POIs). A two-parted agent architecture was used, with an agent responsible for gathering POIs from a location-based service, and a set of Personal Assistant Agents (PAAs) collecting information about the context and intentions of its respective user. In each PAA were embedded four Machine Learning algorithms, with the purpose of ascertaining how well-suited these classifiers are for filtering irrelevant POIs, in a completely automatic fashion. Supervised, incremental learning occurs when the feedback on the true relevance of each recommendation is given by the user to his PAA. To evaluate the recommendations' accuracy, we performed an experiment considering three types of users, using different contexts and intentions. As a result, all the PAA had high accuracy, revealing in specific situations F1 scores higher than 87%.

Download

Emotion-Based Recommender System for Overcoming the Problem of Information Overload

by Hernani Costa and Luis Macedo

Highlights on Practical Applications of Agents and Multi-Agent Systems, May 2013

"Nowadays, we are experiencing a huge growth in the available information, caused by the advent o... more "Nowadays, we are experiencing a huge growth in the available information, caused by the advent of communication technology, which humans cannot handle by themselves. Personal Assistant Agents can help humans to cope with the task of selecting the relevant information. In order to perform well, these agents should consider not only their preferences, but also their mental states (such as beliefs, intentions and emotions) when recommending information.
In this paper, we describe an ongoing Recommender System application, that implements a Multiagent System, with the purpose of gathering heterogeneous information from different sources and selectively deliver it based on: user's preferences; the community's trends; and on the emotions that it elicits in the user."

Download

Extracção de conhecimento léxico-semântico a partir de resumos da Wikipédia

INForum 2010, II Simpósio de Informática, Sep 2010

This paper presents a system for the automatic acquisition of semantic relations from Portuguese ... more This paper presents a system for the automatic acquisition of semantic relations from Portuguese text, which can be seen as core step in the automatic construction of lexico-semantic resources. The system was applied to Wikipedia, currently a huge and free source of knowledge. The obtained results are shown and their evaluation is discussed together with the current limitations and cues for further improvement.

Download

Context and Intention-Awareness in POIs Recommender Systems

by Hernani Costa, Bárbara Furtado, and Luis Macedo

6th ACM Conference on Recommender Systems (RecSys’12), Sep 2012

This paper describes an agent-based approach for making context and intention-aware recommendatio... more This paper describes an agent-based approach for making context and intention-aware recommendations of Points of Interest (POI). A two-parted agent architecture was used, with an agent responsible for gathering POIs from a location-based mobile application, and a set of Personal Assistant Agents (PAA), collecting information about the context and intentions of its respective user. Each PAA includes a probabilistic classifier for making recommendations given its information about the user's context and intentions. Supervised, incremental learning occurs when the feedback of the true relevance of each recommendation is given by the user to his PAA. To evaluate the system's recommendations, we performed an experiment based on the profile used in the training process, using different locations, contexts and goals.

Download

Folheador: browsing through Portuguese semantic relations

13th Conference of the European Chapter of the Association for computational Linguistics, Apr 2012

This paper presents Folheador, an online service for browsing through Portuguese semantic relatio... more This paper presents Folheador, an online service for browsing through Portuguese semantic relations, acquired from different sources. Besides facilitating the exploration of Portuguese lexical knowledge bases, Folheador is connected to services that access Portuguese corpora, which provide authentic examples of the semantic relations in context.

Download

Browsing Portuguese with Folheador

International Conference on Computational Processing of the Portuguese Language, Apr 2012

Lexical knowledge bases (LKBs) hold information on the words of a language and their interactions... more Lexical knowledge bases (LKBs) hold information on the words of a language and their interactions, according to their possible meanings. They are typically structured on word senses, which may be connected by semantic relations. Regarding the complexity of most LKBs, their data formats are generally not suited for being read by humans. User interfaces have thus been developed for providing easier ways of exploring the LKB and assessing its contents. WordNet Search, for Princeton WordNet, or MNEX, for MindNet, are examples of such interfaces. However, in addition to information on words and semantic relations, it is important that the interfaces provide usage examples where semantic relations hold, or at least where related words co-occur. Folheador is a browser for Portuguese LKBs. Besides an interface for navigating through semantic relations acquired from different sources, it is linked to services that provide access to Portuguese corpora, thus allowing the observation of related words co-occurring in authentic contexts of use.

Download

Hernani Costa

Uploads

Books by Hernani Costa

Papers by Hernani Costa

Log In