Showing papers by "International Institute of Information Technology, Hyderabad published in 2007"

Accelerating large graph algorithms on the GPU using CUDA

Abstract: Large graphs involving millions of vertices are common in many practical applications and are challenging to process. Practical-time implementations using high-end computers are reported but are accessible only to a few. Graphics Processing Units (GPUs) of today have high computation power and low price. They have a restrictive programming model and are tricky to use. The G80 line of Nvidia GPUs can be treated as a SIMD processor array using the CUDA programming model. We present a few fundamental algorithms - including breadth first search, single source shortest path, and all-pairs shortest path - using CUDA on large graphs. We can compute the single source shortest path on a 10 million vertex graph in 1.5 seconds using the Nvidia 8800GTX GPU costing $600. In some cases optimal sequential algorithm is not the fastest on the GPU architecture. GPUs have great potential as high-performance co-processors.

Determining Number of Speakers From Multispeaker Speech Signals Using Excitation Source Information

Abstract: In this letter, we address the issue of determining the number of speakers from multispeaker speech signals collected simultaneously using a pair of spatially separated microphones. The spatial separation of the microphones results in time delay of arrival of speech signals from a given speaker. The differences in the time delays for different speakers are exploited to determine the number of speakers from the multispeaker signals. The key idea is that for a given speaker, the relative spacings of the instants of significant excitation of the vocal tract system remain unchanged in the direct components of the speech signals at the two microphones. The time delays can be estimated from the cross-correlation of the Hilbert envelopes of the linear prediction residuals of the multispeaker signals collected at the two microphones.

Detecting Compositionality of Verb-Object Combinations using Selectional Preferences

Abstract: In this paper we explore the use of selectional preferences for detecting noncompositional verb-object combinations. To characterise the arguments in a given grammatical relationship we experiment with three models of selectional preference. Two use WordNet and one uses the entries from a distributional thesaurus as classes for representation. In previous work on selectional preference acquisition, the classes used for representation are selected according to the coverage of argument tokens rather than being selected according to the coverage of argument types. In our distributional thesaurus models and one of the methods using WordNet we select classes for representing the preferences by virtue of the number of argument types that they cover, and then only tokens under these classes which are representative of the argument head data are used to estimate the probability distribution for the selectional preference model. We demonstrate a highly signicant correlation between measures which use these ‘typebased’ selectional preferences and compositionality judgements from a data set used in previous research. The type-based models perform better than the models which use tokens for selecting the classes. Furthermore, the models which use the automatically acquired thesaurus entries produced the best results. The correlation for the thesaurus models is stronger than any of the individual features used in previous research on the same dataset.

Efficient search in document image collections

Abstract: This paper presents an efficient indexing and retrieval scheme for searching in document image databases. In many non-European languages, optical character recognizers are not very accurate. Word spotting - word image matching - may instead be used to retrieve word images in response to a word image query. The approaches used for word spotting so far, dynamic time warping and/or nearest neighbor search, tend to be slow. Here indexing is done using locality sensitive hashing (LSH) - a technique which computes multiple hashes - using word image features computed at word level. Efficiency and scalability is achieved by content-sensitive hashing implemented through approximate nearest neighbor computation. We demonstrate that the technique achieves high precision and recall (in the 90% range), using a large image corpus consisting of seven Kalidasa's (a well known Indian poet of antiquity) books in the Telugu language. The accuracy is comparable to using dynamic time warping and nearest neighbor search while the speed is orders of magnitude better - 20000 word images can be searched in milliseconds.

IIIT Hyderabad at DUC 2007

Abstract: In this paper we report our performance at DUC 2007 summarization tasks. We participated both in the query-focused multidocument summarization main task and in a pilot update summary generation tasks. This year we used a term clustering approach to better estimate a sentence prior. We used only the sentence prior which is query independent, in the update summarization task and found that it’s performance is comparable with the top performing systems. In the main task our system ranked 1 in ROUGE-2, ROUGE-SU4 and ROUGE-BE scores as well as in pyramid scores.

eSagu™: a data warehouse enabled personalized agricultural advisory system

Abstract: In this paper, we explain a personalized agricultural advisory system called eSagu, which has been developed to improve the performance and utilization of agriculture technology and help Indian farmers. In eSagu, rather than visiting the crop in person, the agricultural expert delivers the expert advice at regular intervals (once in one or two weeks) to each farm by getting the crop status in the form of digital photographs and other information. During 2004-06, through eSagu, agricultural expert advices delivered for about 6000 farms covering six crops. The results show that the expert advices helped the farmers to achieve savings in capital investment and improved the crop yield. Mainly, the data warehouse of farm histories has been developed which is providing the crop related information to the agricultural expert in an integrated manner for generating a quality agricultural expert advice. In this paper, after explaining eSagu and its advantages, we discuss how data warehouse of farm histories is enabling agricultural expert to deliver a quality expert advice. We also discuss some research issues to improve the performance of eSagu.

Working with process variation aware caches

Abstract: Deep-submicron designs have to take care of process variation effects as variations in critical process parameters result in large variations in access latencies of hardware components. This is severe in the case of memory components as minimum sized transistors are used in their design. In this work, by considering on-chip data caches, we study the effect of access latency variations on performance. We discuss performance losses due to the worst-case design, wherein the entire cache operates with the worst-case process variation delay, followed by process variation aware cache designs which work at set-level granularity. We then propose a technique called block rearrangement to minimize performance loss incurred by a process variation aware cache which works at set-level granularity. Using block rearrangement technique, we rearrange the physical locations of cache blocks such that a cache set can have its "n" blocks (assuming a n-way set-associative cache) in multiple rows instead of a single row as in the case of a cache with conventional addressing scheme. By distributing blocks of a cache set over multiple sets, we minimize the number of sets being affected by process variation. We evaluate our technique using SPEC2000 CPU benchmarks and show that our technique achieves significant performance benefits over caches with conventional addressing scheme

Perfectly secure message transmission in directed networks tolerating threshold and non threshold adversary

Abstract: In this paper we study Perfectly Secure Message Transmission (PSMT) between a sender S and a receiver R, connected in a directed synchronous network through multiple parallel edges (called wires), each of which are directed from S to R or vice-versa. The unreliability of the network is modeled by a Byzantine adversary with infinite computing power. We investigate the problem with two different adversarial settings: (i) threshold and (ii) non-threshold. In [1], the authors have characterized PSMT against a t-active threshold adversary in directed networks1. However, their PSMT protocol was exponential both in terms of number of phases2 and communication complexity. In addition, they also presented a polynomial phase PSMT protocol with n′ = max(3t-u+1, 2t+1) wires from S to R. In this paper, we significantly improve the exponential phase protocol and present an elegant and efficient three phase PSMT protocol with polynomial communication complexity (and computational complexity) with n= max(3t-2u+1, 2t+1) wires from S to R. Also with n′ = max(3t - u + 1, 2t + 1) wires from S to R, we are able to further improve the communication complexity of our three phase PSMT protocol. Our second contribution in this paper is the first ever characterization for any two phase PSMT protocol. Finally, we also characterize PSMT protocol in directed networks tolerating nonthreshold adversary. In [3], the authors have given the characterization for PSMT against non-threshold adversary. However, in their characterization, they have only considered the paths from S to R, excluding the feedback paths (i.e paths from R to S) and hence their characterization holds good only for single phase protocols. We characterize multiphase PSMT considering feedback paths.

Optical Character Recognition of Amharic Documents

Abstract: In Africa around 2,500 languages are spoken. Some of these languages have their own indigenous scripts. Accordingly, there is a bulk of printed documents available in libraries, information centers, museums and offices. Digitization of these documents enables to harness already available information technologies to local information needs and developments. This paper presents an Optical Character Recognition (OCR) system for converting digitized documents in local languages. An extensive literature survey reveals that this is the first attempt that report the challenges towards the recognition of indigenous African scripts and a possible solution for Amharic script. Research in the recognition of African indigenous scripts faces major challenges due to (i) the use of large number characters in the writing and (ii) existence of large set of visually similar characters. In this paper, we propose a novel feature extraction scheme using principal component and linear discriminant analysis, followed by a decision directed acyclic graph based support vector machine classifier. Recognition results are presented on real-life degraded documents such as books, magazines and newspapers to demonstrate the performance of the recognizer.

Novel architectures for efficient (m, n) parallel counters

Abstract: Parallel counters are key elements in many arithmetic circuits, especially fast multipliers. In this paper, novel architectures and designs for high speed, low power (3,2), (7,3), (15,4) and (31,5) counters capable of operating at ultra-low voltages are presented. Based on these counters, a generalized architecture is derived for large (m, n) parallel counters. The proposed architecture lays emphasis on the use of multiplexers and a combination of CMOS and transmission gate logic in arithmetic circuits that result in high speed and efficient design. The proposed counter designs have been compared with existing designs and are shown to achieve an improvement of about 45% in delay and a reduction of about 25% in power consumption.

Probabilistic perfectly reliable and secure message transmission - possibility, feasibility and optimality

Abstract: We study the interplay of network connectivity and the issues related to feasibility and optimality for probabilistic perfectly reliable message transmission (PPRMT) and probabilistic perfectly secure message transmission (PPSMT) in a synchronous network under the influence of a mixed adversary who possesses unbounded computing power and can corrupt different set of nodes in Byzantine, omission, failstop and passive fashion simultaneously. Our results show that that randomness helps in the possibility of multiphase PPSMT and significantly improves the lower bound on communication complexity for both PPRMT and PPSMT protocols!!.

Approximate String Matching Techniques for Effective CLIR Among Indian Languages

Abstract: Commonly used vocabulary in Indian language documents found on the web contain a number of words that have Sanskrit, Persian or English origin. However, such words may be written in different scripts with slight variations in spelling and morphology. In this paper we explore approximate string matching techniques to exploit this situation of relatively large number of cognates among Indian languages, which are higher when compared to an Indian language and a non-Indian language. We present an approach to identify cognates and make use of them for improving dictionary based CLIR when the query and documents both belong to two different Indian languages. We conduct experiments using a Hindi document collection and a set of Telugu queries and report the improvement due to cognate recognition and translation.

Relative toxicity of neem to natural enemies associated with the chickpea ecosystem: a case study

Abstract: Neem products are often perceived as harmless to natural enemies, pollinators and other non-target organisms. For this reason, several integrated pest management (IPM) programmes have adopted neem as one of the prime components. This study revealed toxic effects of neem on soil-inhabiting and aerial natural enemies in chickpea to an extent of 41 and 29% population reduction, respectively, compared with 63 and 51% when using a conventional insecticide (endosulfan). Neem also affected the parasitization of Helicoverpa armigera (Hubner) larvae by Campoletis chlorideae Uchida up to 20%. The natural enemy population started building up from the vegetative phase and reached their peak during the reproductive phase, and there was a gradual decline from pod formation to pre-harvest phases of the crop. Adapting the currently used IPM system in chickpea using neem during the vegetative phase, followed by an application of Helicoverpa nuclear polyhedrosis virus (HNPV) at flowering and need-based application(s) of chitin inhibitors like novaluron or flufenoxuron instead of endosulfan during pod formation would strongly augment natural enemy populations. This paper discusses the relative toxicity of neem and other IPM components on soil-inhabiting and aerial natural enemies in the chickpea ecosystem.

Selective shielding: a crosstalk-free bus encoding technique

Abstract: With CMOS process technology scaling to deep submicron level, propagation delay across long on-chip buses is becoming one of the main performance limiting factors in high-performance designs. Propagation delay is very significant when adjacent wires are transitioning in opposite direction (i.e., crosstalk transitions) as compared to transitioning in the same direction. As crosstalk transitions have significant impact on propagation delay, several bus encoding techniques have been proposed in literature to eliminate such transitions. In this work, we propose a technique, namely, selective shielding, to eliminate crosstalk transitions. Compared to the conventional shielding technique, our technique significantly reduces the number of extra wires. We give a lower bound on the number of wires required to encode n-bit data using the selective shielding technique. We show that our technique achieves better energy savings and requires less area as compared to the other techniques.

A multi-level model for activity commitments in e-contracts

Abstract: An e-contract is a contract modeled, specified, executed, controlled and monitored by a software system A contract is a legal agreement involving parties, activities, clauses and payments The goals of an e-contract include precise specification of the activities of the contract, mapping them into deployable workflows, and providing transactional support in their execution Activities in a contract are complex and interdependent They may be executed by different parties autonomously and in a loosely coupled fashion They may be compensated and/or re-executed at different times relative to the execution of other activities Both the initial specification of the activities and the later verification of their executions with respect to compliance to the clauses are tedious and complicated We believe that an e-contract should reflect both the specification and the execution aspects of the activities at the same time, where the former is about the composition logic and the later about the transactional properties Towards facilitating this, we propose a multi-level composition model for activities in e-contracts Our model allows for the specification of a number of transactional properties, like atomicity and commitment, for activities at all levels of the composition In addition to their novelty, the transactional properties help to coordinate payments and eventual closure of the contract

IIIT Hyderabad at CLEF 2007 - Adhoc Indian Language CLIR task

Abstract: This paper presents the experiments of Language Technologies Research Centre (LTRC) 1 as part of their participation in CLEF 2 2007 Indian language to English ad-hoc cross language document retrieval task. In this paper we discuss our Hindi and Telugu to English CLIR system and the experiments using CLEF 2007 dataset. We used a variant of TFIDF algorithm in combination with a bilingual lexicon for query translation. We also explored the role of a document summary in fielded queries and two dierent boolean formulations of query translations. We find that a hybrid boolean formulation using a combination of boolean AND and boolean OR operators improves ranking of documents. We also find that simple disjunctive combination of translated query

A tsunami warning system employing level controlled gossiping in wireless sensor networks

Abstract: This paper deals with a Tsunami warning system based on distributed sensor networks employing level controlled gossip. Level controlled gossip is a technique that is being proposed which employs leveling and gossiping together. The technique reduces the number of messages by transmitting messages in direction of base station only and there by increases the life-time of wireless sensor network. By using various power levels at base station the sensor field is hierarchically partitioned into levels of increasing radius (containing various sensor nodes). The algorithm divides the entire sensor network into logical con-centric zones based on proximity from the base station, whereby the packet is transmitted from a node of higher depth to nodes in the next zone with lesser depth. The transmission probability increases with the proximity of the Tsunami wave to the base station. The primary advantage of the protocol is transmitting a critical event with higher probability and at the same time conserving lifetime of the network for future monitoring.

Modeling and analysis of crosstalk for distributed RLC interconnects using difference model approach

Abstract: On-chip inductive effects are becoming predominant in deepsubmicron (DSM) interconnects due to increasing clock speeds, circuit complexity and decreasing interconnect lengths. Inductance causes noise in the signal waveforms, which can adversely affect the performance of the circuit and signal integrity. The traditional analysis of crosstalk in a transmission line begins with a lossless LC representation, yielding a wave equation governing the system response. This paper proposes a difference model approach to derive crosstalk in the transform domain. A closed form solution for crosstalk is obtained by incorporating initial conditions using difference model approach for distributed RLC interconnects. Simulation results show that the effect of inductive coupling forlong interconnects is significant but is almost negligible for local interconnects. It is also shown that when inductance is neglected, the proposed model reduces to a lumped RC model. Also, the analytical model response agrees very well that obtained with SPICE. All the experiments have been carried out for 90nm technology node using Cadence.s Dynamic Circuit Simulator SPECTRE©.

Area Efficient High Speed Architecture of Bruun's FFT for Software Defined Radio

Abstract: Fast Fourier Transform (FFT) is one of the most basic and essential operation performed in software defined radio (SDR). Therefore designing a universal, reconfigurable FFT computation block with low area, delay and power requirement is very important. Recently it is shown that Bruun's FFT is ideally suited for SDR even when operating with higher bit precision to maintain same NSR. In this paper, authors have proposed a new architecture for Bruun's FFT using a distributed approach for incrementing the number of bits (precision) with successive stages of FFT. It is also shown that proposed architecture further reduces the hardware requirement of Bruun's FFT with negligible changes in it's NSR. The proposed design makes Bruun's FFT, a better option for most practical cases in SDR. A detailed comparison of Bruun's traditional and proposed hardware architectures for same NSR is carried out and results of FPGA and ASIC implementations are provided and discussed.

A Hybrid Approach to Example based Machine Translation for Indian Languages

Abstract: Corpus based approaches to machine translation namely Example based machine translation and Statistical machine translation have received wide focus in the recent years. Hybrid approaches combining the two further improved the performance. Indian language machine translation has mostly focussed on rule based machine translation. We propose a hybrid approach to Example based machine translation making use of statistical machine translation methods and minimal linguistic resources. Our motive in this paper is to obtain a ’good enough’ translation as opposed to a perfect translation aimed by earlier machine translation efforts. Our approach can be used for translation of english to any indian language. In this paper, we perform experiments for translation of english to hindi and report BLEU scores.

New and improved architectures for Montgomery modular multiplication

Abstract: In this paper an improved Montgomery multiplier, based on modified four-to-two carry-save adders (CSAs) to reduce critical path delay, is presented Instead of implementing four-to-two CSA using two levels of carry-save logic, authors propose a modified four-to-two CSA using only one level of carry-save logic taking advantage of pre-computed input values Also, a new bit-sliced, unified and scalable Montgomery multiplier architecture, applicable for both RSA and ECC (Elliptic Curve Cryptography), is proposed In the existing word-based scalable multiplier architectures, some processing elements (PEs) do not perform useful computation during the last pipeline cycle when the precision is not equal to an exact multiple of the word size, like in ECC This intrinsic limitation requires a few extra clock cycles to operate on operand lengths which are not powers of 2 The proposed architecture eliminates the need for extra clock cycles by reconfiguring the design at bit-level and hence can operate on any operand length, limited only by memory and control constraints It requires 2∼15% fewer clock cycles than the existing architectures for key lengths of interest in RSA and 11∼18% for binary fields and 10∼14% for prime fields in case of ECC An FPGA implementation of the proposed architecture shows that it can perform 1,024-bit modular exponentiation in about 15 ms which is better than that by the existing multiplier architectures

Three models for discriminative machine translation using global lexical selection and sentence reconstruction

Abstract: Machine translation of a source language sentence involves selecting appropriate target language words and ordering the selected words to form a well-formed target language sentence. Most of the previous work on statistical machine translation relies on (local) associations of target words/phrases with source words/phrases for lexical selection. In contrast, in this paper, we present a novel approach to lexical selection where the target words are associated with the entire source sentence (global) without the need for local associations. This technique is used by three models (Bag-of-words model, sequential model and hierarchical model) which predict the target language words given a source sentence and then order the words appropriately. We show that a hierarchical model performs best when compared to the other two models.

Towards simulating billions of agents in thousands of seconds

Abstract: Building multiagent systems that can scale up to very large number of agents is a challenging research problem. In this paper, we present Distributed Multi Agent System Framework (DMASF), a system which can simulate billions of agents in thousands of seconds. DMASF utilizes distributed computation to gain performance as well as a database to manage the agent and environment state. We briefly present the design and implementation of DMASF and present experimental results. DMASF is a generic and versatile tool that can be used for building massive multi agent system applications.

Distributed Wireless Sensor Network Architecture: Fuzzy Logic Based Sensor Fusion.

Abstract: In this research paper, power aware distributed wireless sensor network architecture proposed by us (in the context of design of routing algorithm) is summarized. An approach to combined routing and fusion algorithm is discussed. Fuzzy logic based approach to wireless sensor fusion is discussed. The need for modeling sensor readings using fuzzy uncertainty is discussed. Novel overlap function for sensor fusion is presented. Concept of fuzzy overlap function is proposed. Method of distributed computation of means is proposed.

Bus-encoding technique to reduce delay, power and simultaneous switching noise (SSN) in RLC interconnects

Abstract: Inductance effects cannot be neglected in global interconnect lines as well as in circuits operating at higher frequencies. This paper presents a new spatio-temporal bus-encoding technique to minimize simultaneous switching noise as well as reduce delay and power dissipation in on-chip buses where inductance effects are dominating. Simulation experiments are carried out to find out the delay and SSN reduction for interconnect lines of different lengths (2mm, 5mm and 10mm) at various technology nodes (180nm, 130nm, 90nm and 65nm). Results obtained show that that the proposed bus-encoding scheme provides a delay reduction of about 54% to 73% with respect to the worst case delay. In addition, encoding is combined with wire shaping and its impact on further delay reduction is observed to be 4% to 26%. Further, when encoding was combined with wire shaping and repeater insertion, an additional delay reduction of 9% to 33% is observed. Concerning SSN, the encoding scheme is tested with various SPEC'95 benchmarks and it is found that SSN is reduced by about 33% on an average compared with the un-encoded data. Finally, energy minimization of about 13% on an average is achieved by the application of new spatio-temporal encoding scheme as reflected by the SPEC'95 bench mark tests.

Perfectly Reliable and Secure Communication in Directed Networks Tolerating Mixed Adversary

Abstract: We characterize Perfectly Secure Message Transmission (PSMT) between two nodes S and R in directed wire model, assuming that n wires are directed from S to R (also termed as top band) and u wires are directed from R to S (also termed as bottom band). A mixed adversary (t b , t f ) controls t b wires in Byzantine fashion and t f in fail-stop fashion among these u+n wires with unbounded computing power. S wishes to send a message m from a finite field \(\mathbb{F}\) in a perfectly secure manner to R such that the adversary gets no information whatsoever on m, though he has unbounded computing power. Our characterization is the first ever characterization for PSMT considering mixed adversary and reveals more fault tolerance than the existing results [1]. Our protocols terminates in constant number of phases, performs polynomial computation and have polynomial communication complexity.

Discriminative word alignment by learning the alignment structure and syntactic divergence between a language pair

Abstract: Discriminative approaches for word alignment have gained popularity in recent years because of the flexibility that they offer for using a large variety of features and combining information from various sources. But, the models proposed in the past have not been able to make much use of features that capture the likelihood of an alignment structure (the set of alignment links) and the syntactic divergence between sentences in the parallel text. This is primarily because of the limitation of their search techniques. In this paper, we propose a generic discriminative re-ranking approach for word alignment which allows us to make use of structural features effectively. These features are particularly useful for language pairs with high structural divergence (like English-Hindi, English-Japanese). We have shown that by using the structural features, we have obtained a decrease of 2.3% in the absolute value of alignment error rate (AER). When we add the cooccurence probabilities obtained from IBM model-4 to our features, we achieved the best AER (50.50) for the English-Hindi parallel corpus.

A Vision System for Monitoring Intermodal Freight Trains

Abstract: We describe the design and implementation of a vision based Intermodal Train Monitoring System (ITMS) for extracting various features like length of gaps in an intermodal (IM) train which can later be used for higher level inferences. An intermodal train is a freight train consisting of two basic types of loads - containers and trailers. Our system first captures the video of an IM train, and applies image processing and machine learning techniques developed in this work to identify the various types of loads as containers and trailers. The whole process relies on a sequence of following tasks -robust background subtraction in each frame of the video, estimation of train velocity, creation of mosaic of the whole train from the video and classification of train loads into containers and trailers. Finally, the length of gaps between the loads of the IM train is estimated and is used to analyze the aerodynamic efficiency of the loading pattern of the train, which is a critical aspect of freight trains. This paper focusses on the machine vision aspect of the whole system

Multi-lingual Indexing Support for CLIR using Language Modeling.

Abstract: An indexing model is the heart of an Information Retrieval (IR) system. Data structures such as term based inverted indices have proved to be very effective for IR using vector space retrieval models. However, when functional aspects of such models were tested, it was soon felt that better relevance models were required to more accurately compute the relevance of a document towards a query. It was shown that language modeling approaches [1] in monolingual IR tasks improve the quality of search results in comparison with TFIDF [2] algorithm. The disadvantage of language modeling approaches when used in monolingual IR task as suggested in [1] is that they would require both the inverted index (term-todocument) and the forward index (document-to-term) to be able to compute the rank of document for a given query. This calls for an additional space and computation overhead when compared to inverted index models. Such a cost may be acceptable if the quality of search results are significantly improved. In a Cross-lingual IR (CLIR) task, we have previously shown in [3] that using a bilingual dictionary along with term co-occurrence statistics and language modeling approach helps improve the functional IR performance. However, no studies exist on the performance overhead in a CLIR task due to language modeling. In this paper we present an augmented index model which can be used for fast retrieval while having the benefits of language modeling in a CLIR task. The model is capable of retrieval and ranking with or without query expansion techniques using term collocation statistics of the indexed corpus. Finally we conduct performance related experiments on our indexing model to determine the cost overheads on space and time.