Anupam Basu

Investigation and modeling of the structure of texting language

Language usage over computer mediated discourses, such as chats, emails and SMS texts, significantly differs from the standard form of the language and is referred to as texting language (TL). The presence of intentional misspellings significantly decrease the accuracy of existing spell checking techniques for TL words. In this work, we formally investigate the nature and type of compressions used in SMS texts, and develop a Hidden Markov Model based word-model for TL. The model parameters have been estimated through standard machine learning techniques from a word-aligned SMS and standard English parallel corpus. The accuracy of the model in correcting TL words is 57.7%, which is almost a threefold improvement over the performance of Aspell. The use of simple bigram language model results in a 35% reduction of the relative word level error rates.

Hardware software partitioning using genetic algorithm

Hardware software co-design is gaining importance with the advent of CAD for embedded systems. A key phase in such designs is partitioning the specification into hardware and software implementation sets. The problem being combinatorically explosive, several greedy search algorithms have been proposed for hardware software partitioning. In this paper, we model the hardware software partitioning problem as a Constraint Satisfaction Problem (CSP), and present a genetic algorithm based approach to solve the CSP in order to obtain the partitioning solution.

Optimized array index computation in DSP programs

An increasing number of components in embedded systems are implemented by software running on embedded processors. This trend creates a need for compilers for embedded processors capable of generating high quality machine code. Particularly for DSPs, such compilers are hardly available, and novel DSP-specific code optimization techniques are required. In this paper we focus on efficient address computation for array accesses in loops. Based on previous work, we present a new and optimal algorithm for address register allocation and provide an experimental evaluation of different algorithms. Furthermore, an efficient and close-to-optimum heuristic is proposed for large problems.

Automatic Part-of-Speech Tagging for Bengali: An Approach for Morphologically Rich Languages in a Poor Resource Scenario

This paper describes our work on building Part-of-Speech (POS) tagger for Bengali. We have use Hidden Markov Model (HMM) and Maximum Entropy (ME) based stochastic taggers. Bengali is a morphologically rich language and our taggers make use of morphological and contextual information of the words. Since only a small labeled training set is available (45,000 words), simple stochastic approach does not yield very good results. In this work, we have studied the effect of using a morphological analyzer to improve the performance of the tagger. We find that the use of morphology helps improve the accuracy of the tagger especially when less amount of tagged corpora are available.

Fault diagnosis in discrete time hybrid systems - A case study

A method of analysing diagnosability of discrete time hybrid systems (DTHS), which are similar to the simple n-rate timed automata [R. Alur, C. Courcoubetis, T.A. Henzinger, P. Ho, Hybrid automata: an algorithmic approach to the specification and verification of hybrid systems, in: Hybrid Systems, LNCS 736, Springer Verlag, 1993, pp. 209-229], has been proposed. A state based fault modeling formalism is used. The properties of the DTHS model, under measurement limitations due to inadequacy or non-availability of sensors, are discussed. A definition of diagnosability for DTHS models has been adopted from the one proposed in [M. Sampath, R. Sengupta, S. Lafortune, K. Sinnamohideen, D. Teneketzis, Diagnosability of discrete-event systems, IEEE Transactions on Automatic Control 40 (9) (1995) 1555-1575] for discrete-event system (DES) models. Based on the measurement limited DTHS models, an algorithm for construction of a diagnoser is presented. It is next demonstrated through an example of a chemical reaction chamber that the diagnosability condition (over the diagnoser), which has been shown to be necessary and sufficient for DES diagnosability, fails to hold for many systems. This is so because the abstraction employed in DES modeling obliterates an important feature of the transitions namely fairness. Exploiting the explicit continuous dynamics of the DTHS models, the fairness of transitions is identified and used to demonstrate diagnosability. The diagnosability condition over the diagnoser is suitably modified to encompass the situations typified by the example.

Performance Models for Automatic Evaluation of Virtual Scanning Keyboards

Virtual scanning keyboards are commonly used augmentative communication aids by persons with severe speech and motion impairments. Designers of virtual scanning keyboards face problems in evaluating alternate designs and hence in choosing the better design among alternatives. Automatic evaluation of designs will be helpful to designers in making the appropriate design choice. In this paper, we present performance models for virtual scanning keyboards that can be used for automatic evaluation. The proposed models address the limitations present in the reported work on similar models. We compared the model predictions with results from user trials and established the validity of the proposed models.

Prototype machine translation system from text-to-Indian sign language

This paper presents a prototype Text-To-Indian Sign Language (ISL) translation system. The system will help dissemination of information to the deaf people in India. This paper also presents the SL-dictionary tool, which can be used to create bilingual ISL dictionary and can store ISL phonological information.

Shruti: an embedded text-to-speech system for Indian languages

In India, the recent increase in the number of people with physical impairments has necessitated the need for low-cost, portable, augmentative and alternative communication devices. The authors describe a text-to-speech system for Indian languages, which accepts the text input in two Indian languages, Bengali and Hindi, and produces near-natural audio output. This text-to-speech system has been ported to two common handheld platforms, namely, iPaq from Compaq and Cassiopeia from Casio, both running on a Microsoft Pocket PC.

Modeling the Co-occurrence Principles of the Consonant Inventories: A Complex Network Approach

Speech sounds of the languages all over the world show remarkable patterns of co-occurrence. In this work, we attempt to automatically capture the patterns of co-occurrence of the consonants across languages and at the same time figure out the nature of the force leading to the emergence of such patterns. For this purpose we define a weighted network where the consonants are the nodes and an edge between two nodes (read consonants) signify their co-occurrence likelihood over the consonant inventories. Through this network we identify communities of consonants that essentially reflect their patterns of co-occurrence across languages. We test the goodness of the communities and observe that the constituent consonants frequently occur in such groups in real languages also. Interestingly, the consonants forming these communities reflect strong correlations in terms of their features, which indicate that the principle of feature economy acts as a driving force towards community formation. In order to measure the strength of this force we propose a theoretical information definition of feature economy and show that indeed the feature economy exhibited by the consonant communities are substantially better than that of those where the consonant inventories had evolved just by chance.

Self-organization of the Sound Inventories: Analysis and Synthesis of the Occurrence and Co-occurrence Networks of Consonants∗

Abstract The sound inventories of the worlds languages self-organize themselves, giving rise to similar cross-linguistic patterns. In this work, we attempt to capture this phenomenon of self-organization, which shapes the structure of the consonant inventories, through a complex network approach. For this purpose we define the occurrence and co-occurrence networks of consonants and systematically study some of their important topological properties. A crucial observation is that the occurrence as well as the co-occurrence of consonants across languages follows a power law distribution. This property is arguably a consequence of the principle of preferential attachment. In order to support this argument we propose a synthesis model, which reproduces the degree distribution for the networks to a close approximation. We further observe that the co-occurrence network of consonants shows a high degree of clustering and subsequently refine our synthesis model in order to explain this property. Finally, we discuss how preferential attachment manifests itself through the evolutionary nature of language.