Language Processing Applications Research Articles

Lightweight Morphological Analysis Model for Smart Home Applications Based on Natural Language Interfaces

With the rapid evolution of the smart home environment, the demand for natural language processing (NLP) applications on information appliances is increasing. However, it is not easy to embed NLP-based applications in information appliances because most information appliances have hardware constraints such as small memory, limited battery capacity, and restricted processing power. In this paper, we propose a lightweight morphological analysis model, which provides the first step module of NLP for many languages. To overcome hardware constraints, the proposed model modifies a well-known left-longest-match-preference (LLMP) model and simplifies a conventional hidden Markov model (HMM). In the experiments, the proposed model exhibited good performance (a response time of 0.0195 sec per sentence, a memory usage of 1.85 MB, a precision of 92%, and a recall rate of 90%) in terms of the various evaluation measures. On the basis of these experiments, we conclude that the proposed model is suitable for natural language interfaces of information appliances with many hardware limitations because it requires less memory and consumes less battery power.

Populating a multilingual ontology of proper names from open sources

Even if proper names play a central role in natural language processing (NLP) applications they are still under-represented in lexicons, annotated corpora, and other resources dedicated to text processing. One of the main challenges is both the prevalence and the dynamicity of proper names. At the same time, large and regularly-updated knowledge sources containing partially-structured data, such as Wikipedia or GeoNames, are publicly available and contain large numbers of proper names. We present a method for a semi-automatic enrichment of Prolexbase, an existing multilingual ontology of proper names dedicated to natural language processing, with data extracted from these open sources in three languages: Polish, English and French. Fine-grained data extraction and integration procedures allow the user to enrich previous contents of Prolexbase with new incoming data. All data are manually validated and available under an open licence.

Natural language inference for arabic using extended tree edit distance with subtrees

Many natural language processing (NLP) applications require the computation of similarities between pairs of syntactic or semantic trees. Many researchers have used tree edit distance for this task...

Unknown Word Detection via Syntax Analyzer

A knowledge resource is the central repository of data for all Natural Language Processing (NLP) applications and development of NLP applications mostly depend on coverage of knowledge resources. The multipurpose Myanmar Language Lexico-conceptual Knowledge Resource (ML2KR) and Myanmar function tagged corpus were developed as initial resources by using semiautomatic approach. ML2KR consists of Myanmar WordNet, Myanmar English bilingual computational lexicon and morphological processor. Myanmar language is morphologically rich and agglutinative language. Therefore, it is usually required to segment Myanmar texts prior to further processing. Segmentation has two main problems, word ambiguity that more than one meaning and unknown word occurrence that a word does not have in the lexicon. In this paper, we address on the unknown word occurrence issue. To detect the new unrestricted character patterns of words, character based parsing syntax analyzer is built by using Context Free Grammar (CFG). Firstly, unknown words are considered as a Name by Name Entity Recognition with forward and backward rule based approach. If the name does not agree with syntax analyzer, all possible unknown words are verified to update the lexicon and Myanmar WordNet. DOI: http://dx.doi.org/10.11591/ij-ai.v2i3.1802 Full Text: PDF

Named Entity Recognition in English Using Hidden Markov Model

Named Entity Recognition (NER) is one of the tasks of processing documents to detect and categorize proper nouns, which is known to be a useful component in various Natural Language Processing (NLP) applications, that enable the extraction of important information from texts very easily. In the following paper we have discussed about NER, various approaches of NER and the results obtained by performing NER in English using Hidden Markov Model(HMM).

Literature Mining Solutions for Life Science Research

Research and development in the area of biomedical literature analysis aims at providing life science researchers with effective means to access and exploit knowledge contained in scientific publications. Virtually all journal publications and many conference proceedings are nowadays readily available in an electronic form—for instance, as abstracts through the MEDLINE citation index or as full-text articles through PubMed Central. Nevertheless, keeping up to date with and searching for recent findings in a research domain remains a tedious task hampered by inefficient and ineffective means for access and exploitation. Biomedical text analysis aims to improve access to unstructured knowledge by alleviating searches, providing auto generated summaries of documents and topics, linking and integrating publications with structured resources, visualizing content for better understanding, and guiding researchers to novel hypotheses and into knowledge discovery. Focused research over recent years has improved fundamental solutions for biomedical text mining, such as document retrieval, named entity recognition, normalization and grounding, and extraction of relationships, with levels of accuracy that reach human annotators when considering inter annotator agreement. Consequently, more and more integrative analysis tools were put forward by the text mining community targeting a broad audience of end users: generic and task-specific search engines for life science researchers, interfaces for networks synthesis based on textual evidences, or more specialized tools searching for transcription factors, or primer sequences. This special issue of Advances in Bioinformatics presents overviews and examples of end-user-oriented biomedical text mining tools for bioinformaticians, molecular biologists, biochemists, clinicians, pharmacologists, and other researchers in life sciences. We start with A. Manconi et al. survey on “Literature retrieval and mining in bioinformatics: state of the art and challenges.” The authors introduce the major concepts that life science researchers should be familiar with getting the best out of existing text mining solutions, and survey key tools and research. In a dedicated second part of their survey, the authors address the major challenges both life science researchers and solution developers are facing at this point. The reader will find plenty of references to existing search tools, resources, and research papers. A. E. Thessen et al. focus on a particular domain, presenting an overview of “Applications of natural language processing in biodiversity science.” The authors review the application of natural processing and machine learning for biological information extraction regarding cellular processes, taxonomic names, and morphological characters. You will find detailed examples, a summary of all steps involved in information extraction, and lots of references to existing tools and resources. S. T. Ahmed et al. introduce their semantic faceted search engine, BioEve, in “A novel framework to facilitate interactive literature search.” They couple an automated extraction system with a cognitive search and navigation service, to alleviate the process of searching and browsing huge amounts of literature such as provided/delivered by MEDLINE. BioEve enables interactive query refinement and suggests concepts and entities (like genes, drugs, and diseases) to quickly filter and modify search directions, thereby achieving semantic enrichment that improves insight gains while searching literature. S. V. Landeghem et al. present their EVEX resource in “Exploring biomolecular literature with EVEX: connecting genes through events, homology, and indirect associations.” The authors extracted more than 20 million biomolecular events involving genes and proteins, such as phosphorylation and gene regulation, from MEDLINE. The online tool generates a summary on the searched gene denoting all regulated genes, binding partners, subcellular locations, and other related data linked to the searched gene. We conclude this special issue with the paper by A. Divoli et al., discussing whether “Do Peers see more in a paper than its authors.” In a meta-analysis using automatic text analysis, they address questions such as how informative an abstract is compared to the full text; and how peers and authors might view the major contributions of a paper differently. Their analyses are comparing the information content of an abstract, as written by the paper's authors, to sentences that mention the paper as a reference, written by peers. Using this strategy, A. Divoli et al. found, for example, that citing sentences contain 20% additional concepts (likely important contributions) that were not mentioned in the abstract of the paper referred to, but maybe should have been to help attract even more peers.

A Highest Sense Count Based Method for Disambiguation of Web Queries for Hindi Language Web Information Retrieval

The ambiguity in word senses has been recognized as a major challenge for the information retrieval systems. Hindi language web information retrieval, like other languages, faces the problem of sense ambiguity. The sense ambiguity problem deteriorates the performance of every natural language processing (NLP) application. The performance of Hindi language web information retrieval is also affected by it. In this paper, the author formalized an approach for the disambiguation of the senses to improve the performance of Hindi web information retrieval. Our system works in such a way that ambiguity detection has been performed before disambiguation of web queries. Test samples of 100 queries have been selected. When these queries were subjected to ambiguity detection, we found that 43% of them have been detected unambiguous. After ambiguity detection, the disambiguation approach is followed which is based on HSC (Highest Sense Count). Query disambiguation approach further follows query expansion. The expanded query generates the new result set which results into high precision and high similarity score. The 57 expanded queries are tested against 1000 test document instances. The overall improvement is 45% in the average precision, 23% in interpolated average precision and a significant improvement in the average similarity score of the new generated result set. The overall accuracy of our approach has been 61.4% and it improves the performance of the system by 45%.

Discriminate Chinese Word Segmenter with Global and Context Features

Chinese Word segmenter is the basis for all subsequent applications of natural language processing. The Corpus-based statistic method has become the predominant method. However, the training corpora are not enough especially in certain areas. Therefore, we introduce some global features and context features in order to get almost the same performance only with much smaller scale corpus. The experiments results show that our approach significantly outperforms the original feature sets in the same training data. Meanwhile, the time-consuming of model training is also reduced. In addition, these features do not depend on classifiers, so our method can easily be changed to other models.

Research on Chinese Named Entity Recognition Based on Ontology

As a critical role in many Natural Language Processing (NLP) applications, such as Information Extraction, Machine Translation etc, Chinese Named Entity Recognition (NER) remains a challenging task because of its characteristics. This paper proposes a method of Chinese NER, which combining Conditional Random Fields (CRFs) model with domain ontology as a semantic feature besides word and part of speech features. Experiments were made to compare the two kinds of feature templates, and the precision rate and recall rate of Chinese NER rose to 90.86% and 88.23%, which showed remarkable performance of the proposed approach. Combination of ontology and CRFs method increased effectively the precision and recall of Chinese NER.

Simulated annealing based classifier ensemble techniques: Application to part of speech tagging

Part-of-Speech (PoS) tagging is an important pipelined module for almost all Natural Language Processing (NLP) application areas. In this paper we formulate PoS tagging within the frameworks of single and multi-objective optimization techniques. At the very first step we propose a classifier ensemble technique for PoS tagging using the concept of single objective optimization (SOO) that exploits the search capability of simulated annealing (SA). Thereafter we devise a method based on multiobjective optimization (MOO) to solve the same problem, and for this a recently developed multiobjective simulated annealing based technique, AMOSA, is used. The characteristic features of AMOSA are its concepts of the amount of domination and archive in simulated annealing, and situation specific acceptance probabilities. We use Conditional Random Field (CRF) and Support Vector Machine (SVM) as the underlying classification methods that make use of a diverse set of features, mostly based on local contexts and orthographic constructs. We evaluate our proposed approaches for two Indian languages, namely Bengali and Hindi. Evaluation results of the single objective version shows the overall accuracy of 88.92% for Bengali and 87.67% for Hindi. The MOO based ensemble yields the overall accuracies of 90.45% and 89.88% for Bengali and Hindi, respectively.

Natural language interaction with the web of data by mining its textual side

The Semantic Web is an extension of the classical web. The data and schemas it adds coexist with the documents that were already linked and available. This not only allows interoperability, reusability and potentially unforeseen applications of opened data, but it also creates a unique situation of availability on the web of huge collections of the same pieces of information available at the same time as text and as structured data. An interesting example is the couple Wikipedia-DBpedia: exploiting these interlinked structured and unstructured data sources in parallel can offer a great potential for both Natural Language Processing and Semantic Web applications. Starting from these observations, this paper addresses the problem of enhancing interactions between non-expert users and data available on the Web. In particular, we present QAKiS, a system for open domain Question Answering over linked data, that addresses the problem of question interpretation as a relation-based match, where fragments of the question are matched to binary relations of the triple store, using relational textual patterns automatically collected. In the current version, the relational patterns are automatically extracted from Wikipedia, while DBpedia is the data set to be queried using a natural language interface.

Computational Linguistics without Linguistics? View from Prague

We want to demonstrate on some selected linguistic issues that classical structural and functional linguistics even with its seemingly traditional approaches has something to offer to a formal description of language and its applications in natural language processing and to illustrate by a brief reference to Functional Generative Grammar (on the theoretical side of CL) and Prague Dependency Treebank (on the applicational side) a possible interaction between linguistics and CL.

Exploring patterns in dictionary definitions for synonym extraction

AbstractAutomatic determination of synonyms and/or semantically related words has various applications in Natural Language Processing. Two mainstream paradigms to date, lexicon-based and distributional approaches, both exhibit pros and cons with regard to coverage, complexity, and quality. In this paper, we propose three novel methods—two rule-based methods and one machine learning approach—to identify synonyms from definition texts in a machine-readable dictionary. Extracted synonyms are evaluated in two extrinsic experiments and one intrinsic experiment. Evaluation results show that our pattern-based approach achieves best performance in one of the experiments and satisfactory results in the other, comparable to corpus-based state-of-the-art results.

U-Compare: A modular NLP workflow construction and evaluation system

During the development of natural language processing (NLP) applications, developers are often required to repeatedly perform certain tasks. Among these tasks, workflow comparison and evaluation are two of the most crucial because they help to discover the nature of NLP problems, which is important from both scientific and engineering perspectives. Although these tasks can potentially be automated, developers tend to perform them manually, repeatedly writing similar pieces of code. We developed tools to largely automate these subtasks. Promoting component reuse is another way to further increase NLP development efficiency. Building on the interoperability enhancing Unstructured Information Management Architecture (UIMA) framework, we have collected a large library of interoperable resources, developed several workflow creation utilities, added a customizable comparison and evaluation system, and built visualization utilities. These tools are modularly designed to accommodate various use cases and potential reuse scenarios. By integrating all these features into our U-Compare system, we hope to increase NLP developer efficiency. Simple to use and directly runnable from a web browser, U-Compare has already found uses in a range of applications.

Chinese new word identification: a latent discriminative model with global features

Chinese new words are particularly problematic in Chinese natural language processing. With the fast development of Internet and information explosion, it is impossible to get a complete system lexicon for applications in Chinese natural language processing, as new words out of dictionaries are always being created. The procedure of Hew words identification and POS tagging are usually separated and the features of lexical information cannot be fully used. A latent discriminative model, which combines the strengths of Latent Dynamic Conditional Random Field (LDCRF) and semi-CRF, is proposed to detect new words together with their POS synchronously regardless of the types of new words from Chinese text without being pre-segmented. Unlike semi-CRF, in proposed latent discriminative model, LDCRF is applied 10 generate candidate entities, which accelerates the training speed and decreases the computational cost. The complexity of proposed hidden semi-CRF could be further adjusted by tuning the number of hidden variables and the number of candidate entities from the Nbest outputs of LDCRF model. A new-word-generating framework is proposed for model training and testing, under which the definitions and distributions of new words conform to the ones in real text. The global feature called "Global Fragment Features" for new word identification is adopted. We tested our model on the corpus from SIGHAN-6. Experimental results show that the proposed method is capable of detecting even low frequency new words together with their POS tags with satisfactory results. The proposed model performs competitively with the state-of-the-art models.

The structure of the lexical component in a lexico-conceptual knowledge base

In the last years a language model, the Lexical Constructional model (mairal & Ruiz de mendoza, 2008, 2009; Ruiz de mendoza & mairal, 2008) has been developed to account for all facets of meaning, including pragmatic and discourse phenomena. Recent research in the LCm (Periñán & Arcas, 2004, 2005, 2007a, 2010a, 2010b) has built an online lexico-conceptual base, FunGramKB, for natural language processing (nLP) applications. FunGramKB consists of two levels of information (i.e. a lexical level and a conceptual level) which in turn comprise several modules. The present contribution focuses on the lexicon, one of the two modules in the lexical level. The aim of the paper is twofold: first, to analyse the FunGramKB lexical component; second, to illustrate the semantic representation in the lexicon of the concepts stored in the ontology through a contrastive analysis of the French and english lexical entries for speech act predicates.

Arabic Numerals Checker: Checking Agreement between Numerals and Counted Objects in the Arabic Language

Numerals are one of the major aspects to be considered in language processing. In this paper, we present an approach for developing a numeral checker application for the language, called Arabic Numerals Checker. Numerals Checker is intended to help users achieve competence in writing according to the numeral rules. It is supported by an explanation facility that provides the user with a clear explanation of the output. The use of the Numerals Checker, when integrated with other language processing applications, can enhance the quality of the text for users who read or write in the language. The Numerals Checker is implemented as a simple Finite State Transducer that maps numbers to the corresponding numerals in a way that respects the gender agreement feature. The proposed approach covers all the numbers in the range 1 to 999,999,999,999. Tests have been executed using random numbers in the range above. Experimental results have demonstrated that the proposed approach is promising in the sense that it provides efficient and error-free transformation.

Integration of Multiple Classifiers for Chinese Semantic Dependency Analysis

Semantic Dependency Analysis (SDA) has extensive applications in Natural Language Processing (NLP). In this paper, an integration of multiple classifiers is presented for SDA of Chinese. A Naive Bayesian Classifier, a Decision Tree and a Maximum Entropy classifier are used in a majority wins voting scheme. A portion of the Penn Chinese Treebank was manually annotated with semantic dependency structure. Then each of the three classifiers was trained on the same training data. All three of the classifiers were used to produce candidate relations for test data and the candidate relation that had the majority vote was chosen. The proposed approach achieved an accuracy of 86% in experimentation, which shows that the proposed approach is a promising one for semantic dependency analysis of Chinese.

The bootstrapping of the Yarowsky algorithm in real corpora

The Yarowsky bootstrapping algorithm resolves the homograph-level word sense disambiguation (WSD) problem, which is the sense granularity level required for real natural language processing (NLP) applications. At the same time it resolves the knowledge acquisition bottleneck problem affecting most WSD algorithms and can be easily applied to foreign language corpora. However, this paper shows that the Yarowsky algorithm is significantly less accurate when applied to domain fluctuating, real corpora. This paper also introduces a new bootstrapping methodology that performs much better when applied to these corpora. The accuracy achieved in non-domain fluctuating corpora is not reached due to inherent domain fluctuation ambiguities.

Scaling Conditional Random Fields by One-Against-the-Other Decomposition

As a powerful sequence labeling model, conditional random fields (CRFs) have had successful applications in many natural language processing (NLP) tasks. However, the high complexity of CRFs training only allows a very small tag (or label) set, because the training becomes intractable as the tag set enlarges. This paper proposes an improved decomposed training and joint decoding algorithm for CRF learning. Instead of training a single CRF model for all tags, it trains a binary sub-CRF independently for each tag. An optimal tag sequence is then produced by a joint decoding algorithm based on the probabilistic output of all sub-CRFs involved. To test its effectiveness, we apply this approach to tackling Chinese word segmentation (CWS) as a sequence labeling problem. Our evaluation shows that it can reduce the computational cost of this language processing task by 40-50% without any significant performance loss on various large-scale data sets.