Mohammed Hasanuzzaman

Propagation Strategies for Building Temporal Ontologies

In this paper, we propose to build temporal ontologies from WordNet. The underlying idea is that each synset is augmented with its temporal connotation. For that purpose, temporal classifiers are iteratively learned from an initial set of time-sensitive synsets and different propagation strategies to give rise to different TempoWordNets.

Feature selection for event extraction in biomedical text

In this paper we report our work on multiobjective optimization (MOO) based feature selection approach for event extraction in biomedical texts. Event extraction deals with the detection and classification of expressions that represent complex biological phenomenon involving genes and proteins. We perform feature selection within the framework of a robust machine learning algorithm, namely Conditional Random Field (CRF). We implement a set of diverse features that exploit lexical, shallow syntactic and contextual information. At first we develop a single objective optimization (SOO) based feature selection technique where we optimize F-measure function. Thereafter we develop two different models of MOO based feature selection by optimizing different pairs of objective functions, i.e. recall and precision; and feature count and F-measure. We carried out experiments on the benchmark setup of BioNLP-2013 shared task. We obtain the best performance with the overall average recall, precision and F-measure values of 57.04%, 75.08% and 64.77%, respectively. Evaluation shows that the classifier can achieve good performance level when trained with an effective feature set. We also observe that MOO can indeed performs better than the SOO based approach.

A Genetic Approach for Biomedical Named Entity Recognition

In this paper, we report a classifier ensemble technique using the search capability of genetic algorithm (GA) for Named Entity Recognition (NER) in biomedical domain. We use Maximum Entropy (ME) framework to build a number of classifiers depending upon the various representations of a set of features. The proposed technique is evaluated with the JNLPBA 2004 data sets that yield the overall recall, precision and F-measure values of 67.98\%, 71.68\% and 69.78\%, respectively.

Multiobjective Approach for Feature Selection in Maximum Entropy Based Named Entity Recognition

In this paper, we present the problem of appropriate feature selection for constructing a Maximum Entropy (ME) based Named Entity Recognition (NER) system under the multiobjective optimization (MOO) framework. Two conflicting objective functions are simultaneously optimized using the search capability of MOO. These objectives are (i). the dimensionality of features, which is tried to be minimized, and (ii). the corresponding F-measure value of the classifier, trained using the features present, is maximized. The features are encoded in the chromosomes. Thereafter, a multiobjective evolutionary algorithm in the steps of a popular MOO technique, NSGA-II, is developed to determine the appropriate feature subset. The proposed technique is evaluated to determine the suitable feature combinations for NER in a resource-constrained language, namely Bengali. Evaluation results yield the recall, precision and F-measure values of 72.45\%, 82.39\% and 77.11\%, respectively.

Temporal Orientation of Tweets for Predicting Income of Users.

Automatically estimating a user’s socioeconomic profile from their language use in social media can significantly help social science research and various downstream applications ranging from business to politics. The current paper presents the first study where user cognitive structure is used to build a predictive model of income. In particular, we first develop a classifier using a weakly supervised learning framework to automatically time-tag tweets as past, present, or future. We quantify a user’s overall temporal orientation based on their distribution of tweets, and use it to build a predictive model of income. Our analysis uncovers a correlation between future temporal orientation and income. Finally, we measure the predictive power of future temporal orientation on income by performing regression.

Local Event Discovery from Tweets Metadata

We present a two-step strategy that addresses fundamental deficiencies in social media-based event detection and achieves effective local event by taking advantage of geo-located data from Twitter. While previous work has mainly relied on an analysis of tweet text to identify local events, we show how to reliably detect events using meta-data analysis of geo-tagged tweets. The first step of the method identifies several spatio-temporal clusters within the dataset across both space and time using metadata to form potential candidate events. In the second step, it ranks all the candidates by the amount of hashtag/entity inequality. We used crowdsourcing to evaluate the proposed approach on a data set that contains millions of geo-tagged tweets. The results show that our framework performs reasonably well in terms of precision and discovers local events faster.

Overview of the 4th HistoInformatics Workshop

In line with global trends, historical records are increasingly available in forms that computer can process. These ever expanding records (such as scanned books, large-scale corpora, academic papers, maps, photos, audios, videos)---either digitally born or reconstructed through digitization pipelines---are too big to be read or viewed manually. Historians, like other humanities researchers, have a keen interest in computational approaches to process and study digitized historical information for research, writing, and dissemination of historical knowledge. In Computer Science, experimental tools and methods are challenged to be validated regarding their relevance for real-world questions and applications. The HistoInformatics workshop series is focused on the challenges and opportunities of data-driven humanities and brings together scientists and scholars at the forefront of this emerging field, at the interface between History, Anthropology, Archaeology, Computer Science and associated disciplines as well as the cultural heritage sector. The 4th HistoInformatics Workshop was a half day workshop co-located with the 26th ACM International Conference on Information and Knowledge Management (CIKM 2017) in Singapore.

Place-Type Detection in Location-Based Social Networks

Determining the type of places in location-based social networks will contribute to the success of various downstream tasks such as POI recommendation, location search, automatic place name database creation, and data cleaning. In this paper, we propose a multi-objective ensemble learning framework that (i) allows the accurate tagging of places into one of the three categories: public, private, or virtual, and (ii) identifying a set of solutions thus offering a wide range of possible applications. Based on the check-in records, we compute two types of place features from (i) specific patterns of individual places and (ii) latent relatedness among similar places. The features extracted from specific patterns (SP) are derived from all check-ins at a specific place. The features from latent relatedness (LR) are computed by building a graph of related places where similar types of places are connected by virtual edges. We conduct an experimental study based on a dataset of over 2.7M check-in records collected by crawling Foursquare-tagged tweets from Twitter. Experimental results demonstrate the effectiveness of our approach to this new problem and show the strength of taking various methods into account in feature extraction. Moreover, we demonstrate how place type tagging can be beneficial for place name recommendation services.

Understanding Temporal Query Intent

Understanding the temporal orientation of web search queries is an important issue for the success of information access systems. In this paper, we propose a multi-objective ensemble learning solution that (1) allows to accurately classify queries along their temporal intent and (2) identifies a set of performing solutions thus offering a wide range of possible applications. Experiments show that correct representation of the problem can lead to great classification improvements when compared to recent state-of-the-art solutions and baseline ensemble techniques.