Posts by Collection

Statistical values alone cannot bring the whole scenario of crime occurrences in the city of Dhaka. We need a better way to use these statistical values to predict crime occurrences and make the city a safer place to live. Proper decision-making for the future is key in reducing the rate of criminal offenses in an area or a city. If the law enforcement bodies can allocate their resources efficiently for the future, the rate of crime in Dhaka can be brought down to a minimum. In this work, we have made an initiative to provide an effective tool with which law enforcement officials and detectives can predict crime occurrences ahead of time and take better decisions easily and quickly. We have used several Fuzzy Inference Systems (FIS) and Adaptive Neuro-Fuzzy Inference Systems (AXFIS) to predict the type of crime that is highly likely to occur at a certain place and time.

Information Extraction allows machines to decipher natural language through using two tasks: Named Entity Recognition and Relation Extraction. In order to build such a system for Bangla Language, in this work a Named Entity Recognition (NER) System is proposed, which requires a minimum information to deliver a decent performance having less dependency on handcrafted features. The proposed model is based on Deep Learning, which is accomplished through the use of a Densely Connected Network (DCN) in collaboration with a Bidirectional-LSTM (BiLSTM) and word embedding, i.e., DCN-BiLSTM. Such a system, specific to the Bangla language, has never been done before. Furthermore, a unique dataset was made since no Named Entity Recognition dataset exists for Bangla language till date. In the dataset, over 71 thousand Bangla sentences have been collected, annotated, and classified into four different groups using IOB tagging scheme. Those groups are person, location, organization, and object. Due to Bangla’s morphological structure, character level feature extraction is also applied so that we can access more features to determine relational structure between different words. This is initially done with the use of a Convolutional Neural Network but is later outperformed by our second approach which is through the use of a Densely Connected Network (DCN). As for the training portion, it has been done for two variations of word embedding which are word2vec and glove, the outcome being the largest vocabulary size known to both models. A detailed discussion in regard to the methodology of the NER system is explained in a comprehensive manner followed by an examination of the various evaluation scores achieved. The proposed model in this work resulted in having a F1 score of 63.37, which is evaluated at Named Entity Level.

Pseudo-cliques (subgraphs with almost all possible edges) have many applications. But they do not satisfy the convertible antimonotone constraint (as we prove here). So, it is hard to reduce the search space of pseudo-cliques and list them efficiently. To our knowledge, only two exact algorithms, namely, ODES and PCE, were proposed for this purpose, but both have high execution times. Here, we present an exact algorithm named Fast Pseudo-Clique Enumerator (FPCE). It employs some pruning techniques we derived to reduce the search space. Our experiment on 15 real and 16 synthetic graphs shows that (i) on real graphs, FPCE is, on average, 38.6 and 6.5 times faster than ODES and PCE, respectively, whereas (ii) on synthetic graphs, FPCE is, on average, 39.7 and 3.1 times faster than ODES and PCE, respectively. We apply FPCE and a popular heuristic method on a PPI network to identify pseudo-cliques. FPCE outputs match with more known protein complexes, are more accurate, and are biologically more significant - suggesting that the exact computation of pseudo-cliques may give better insights. For its speed, FPCE is a suitable choice in such cases.