Ruli Manurung

Using genetic algorithms to create meaningful poetic text

This article presents a series of experiments in automatically generating poetic texts. We confined our attention to the generation of texts which are syntactically well-formed, meet certain pre-specified patterns of metre and broadly convey some given meaning. Such aspects can be formally defined, thus avoiding the complications of imagery and interpretation that are central to assessing more free forms of verse. Our implemented system, McGONAGALL, applies the genetic algorithm to construct such texts. It uses a sophisticated linguistic formalism to represent its genomic information, from which can be computed the phenotypic information of both semantic representations and patterns of stress. The conducted experiments broadly indicated that relatively meaningful text could be produced if the constraints on metre were relaxed, and precise metric text was possible with loose semantic constraints, but it was difficult to produce text which was both semantically coherent and of high quality metrically.

THE CONSTRUCTION OF A PUN GENERATOR FOR LANGUAGE SKILLS DEVELOPMENT

Since the early 1990s, there have been a number of small-scale computer programs that automatically constructed simple verbal jokes (puns), but none of these were fully developed systems that could be used for a practical application. We describe the building and testing of the STANDUP program – a large-scale, robust, interactive, user-friendly pun-generator (inspired by Binsteds JAPE program), which is aimed at allowing children, particularly those with communication disabilities, to develop their linguistic skills. The STANDUP system was designed in consultation with potential users and suitable experts, was rigorously engineered using public-domain linguistic data, and has a special purpose, child-friendly, graphical user interface. The software was tested successfully with real users (children with complex communication needs).

Evaluating the STANDUP Pun Generating Software with Children with Cerebral Palsy

The interactive STANDUP software was developed to provide children who use augmentative and alternative communication (AAC) with a “language playground.” The software provides appropriate functionality for users with physical, speech, and language impairments to generate and tell novel punning riddles at different levels of complexity. STANDUP was evaluated with nine children with cerebral palsy during an eight-week study. Results show that the participants were able to generate and tell novel jokes with minimal or no support. The use of STANDUP impacted favorably on general AAC use. The study results also suggested that STANDUP could potentially have a positive effect on social and pragmatic skills. Further research to investigate the impact of STANDUP on communication skills is proposed. Suggestions for future software development include providing users with opportunities to complete jokes and to integrate online dictionaries when new vocabulary is encountered.

Integrated visual vocabulary in latent Dirichlet allocation–based scene classification for IKONOS image

Abstract Scene classification based on latent Dirichlet allocation (LDA) is a more general modeling method known as a bag of visual words, in which the construction of a visual vocabulary is a crucial quantization process to ensure success of the classification. A framework is developed using the following new aspects: Gaussian mixture clustering for the quantization process, the use of an integrated visual vocabulary (IVV), which is built as the union of all centroids obtained from the separate quantization process of each class, and the usage of some features, including edge orientation histogram, CIELab color moments, and gray-level co-occurrence matrix (GLCM). The experiments are conducted on IKONOS images with six semantic classes (tree, grassland, residential, commercial/industrial, road, and water). The results show that the use of an IVV increases the overall accuracy (OA) by 11 to 12% and 6% when it is implemented on the selected and all features, respectively. The selected features of CIELab color moments and GLCM provide a better OA than the implementation over CIELab color moment or GLCM as individuals. The latter increases the OA by only ∼ 2 to 3%. Moreover, the results show that the OA of LDA outperforms the OA of C4.5 and naive Bayes tree by ∼ 20 % .

Adding phonetic similarity data to a lexical database

As part of a project to construct an interactive program which would encourage children to play with language by building jokes, we developed a lexical database, starting from WordNet. To the existing information about part of speech, synonymy, hyponymy, etc., we have added phonetic representations and phonetic similarity ratings for pairs of words/phrases.

Discrete Firefly Algorithm for Traveling Salesman Problem: A New Movement Scheme

The “firefly algorithm” (FA) is a nature-inspired technique originally designed for solving continuous optimization problems. There are several existing approaches that apply FA also as a basis for solving discrete optimization problems, in particular the “traveling salesman problem” (TSP). In this chapter, we present a new movement scheme called edge-based movement, an operation which guarantees that a candidate solution more closely resembles another one. This leads to a more FA-like behavior of the algorithm. We investigate the performance of the ‘evolutionary discrete firefly algorithm” when using this new edge-based movement and compare it against previous methods. Computer simulations show that the new movement scheme produces slightly better accuracy with much faster average time. The average speedup factor is 14.06 times.

Designing an Indonesian part of speech tagset and manually tagged Indonesian corpus

We describe our work on designing a linguistically principled part of speech (POS) tagset for the Indonesian language. The process involves a detailed study and analysis of existing tagsets and the manual tagging of an Indonesian corpus. The results of this work are an Indonesian POS tagset consisting of 23 tags and an Indonesian corpus of over 250.000 lexical tokens that have been manually tagged using this tagset.

Automatic Wayang Ontology Construction using Relation Extraction from Free Text

This paper reports on our work to automatically construct and populate an ontology of wayang (Indonesian shadow puppet) mythology from free text using relation extraction and relation clustering. A reference ontology is used to evaluate the generated ontology. The reference ontology contains concepts and properties within the wayang character domain. We examined the influence of corpus data variations, threshold value variations in the relation clustering process, and the usage of entity pairs or entity pair types during the feature extraction stages. The constructed ontology is examined using three evaluation methods, i.e. cluster purity (CP), instance knowledge (IK), and relation concept (RC). Based on the evaluation results, the proposed method generates the best ontology when using a consolidated corpus, the threshold value in relation clustering is 1, and entity pairs are used during feature extraction.

Building an Indonesian rule-based part-of-speech tagger

This paper describes work on a part-of-speech tagger for the Indonesian language by employing a rule-based approach. The system tokenizes documents while also considering multi-word expressions and recognizes named entities. It then applies tags to every token, starting from closed-class words to open-class words and disambiguates the tags based on a set of manually defined rules. The system currently obtains an accuracy of 79% on a manually tagged corpus of roughly 250.000 tokens.

Building an Indonesian named entity recognizer using Wikipedia and DBPedia

This paper describes the development of an Indonesian NER system using online data such as Wikipedia 1 and DBPedia 2. The system is based on the Stanford NER system [8] and utilizes training documents constructed automatically from Wikipedia. Each entity, i.e. word or phrase that has a hyperlink, in the Wikipedia documents are tagged according to information that is obtained from DBPedia. In this very first version, we are only interested in three entities, namely: Person, Place, and Organization. The system is evaluated using cross fold validation and also evaluated using a gold standard that was manually annotated. Using cross validation evaluation, our Indonesian NER managed to obtain precision and recall values above 90%, whereas the evaluation using gold standard shows that the Indonesian NER achieves high precision but very low recall.