Alon Itai - Academia.edu (original) (raw)

Papers by Alon Itai

Algorithms for adaptive routing in a high speed networks are examined. In such networks the inter... more Algorithms for adaptive routing in a high speed networks are examined. In such networks the intermediate nodes serve as switching elements with limited computational resources. The mode of operation is to specify an entire path from the source to the destination. If all the links of the path are nonfaulty then the message is transmitted; otherwise an indication is given to the first link that has failed. The source, upon learning of a failure, chooses another path and retransmits the message. The links of the network have independent a priori failure probablities. The status of the links does not change while attempting to transmit a message. First, the greedy algorithm (the algorithm which tries the paths in the order of increasing failure probability) is examined. It is shown to be optimal for a small class of networks, but not for all. The main result concerns series/parallel graphs, for which an optimal algorithm is presented. It is shown that an optimal algorithm can be easily ...

Hebrew, as other ~emitic languages, has a rich morpl1ology, observable in part by the complexity ... more Hebrew, as other ~emitic languages, has a rich morpl1ology, observable in part by the complexity of verb inflections. The primary base of verbs in Hebrew is the past third singular form of tlfe verb. From this base, some twenty eight different inflected forms can be created according to tense, per~on, gender and number. Traditionally, inflection tables were used to describe the various inflected forms derived from the verb 'base. Research done by Oman has managed to describe the verb inflection process using the principles of Generative Grammar. In' this approach, inflCfted verb forms are viewed as constructs of the form preftx+base+sufftx. Verb inflection is described as a s~ries of sequentialpperations. The first stage converts the primary verb base to a secondary'base, when the secondary base is not the same as the primary base. Secondly, the appropriate prefix and/or suffix are concatenated to the base. Thirdly, several morpho-phonemic changes due to the affix concat...

Author(s): Dechter, Rina; Itai, Alon | Abstract: We address the problem of enumerating (producing... more Author(s): Dechter, Rina; Itai, Alon | Abstract: We address the problem of enumerating (producing) all models of a given theory. We show that the enumeration task can be performed in time proportional to the product of the number of models and the effort needed to generate each model in isolation. In other words, the requirement of generating a new solution in each iteration does not in itself introduce substantial complexity. Consequently, it is possible to decide whether any tractably satisfiable formula has more than K solutions in time polynomial in the size of the forumla and in K. In the special cases of Horn formulas and 2-CNFs, although counting is #P-complete, to decide whether the count exceeds K, is polynomial in K.

This paper proposes a new approach for acquiring morpho-lexical probabilities from an untagged co... more This paper proposes a new approach for acquiring morpho-lexical probabilities from an untagged corpus. This approach demonstrates a way to extract very useful and nontrivial information from an untagged corpus, which otherwise would require laborious tagging of large corpora. The paper describes the use of these morpho-lexical probabilities as an information source for morphological disambiguation in Hebrew. The suggested method depends primarily on the following property: a lexical entry in Hebrew may have many different word forms, some of which are ambiguous and some of which are not. Thus, the disambiguation of a given word can be achieved using other word forms of the same lexical entry. Even though it was originally devised and implemented for dealing with the morphological ambiguity problem in Hebrew, the basic idea can be extended and used to handle similar problems in other languages with rich morphology.

Journal of Computer and System Sciences, 1994

The learning model of Valiant is extended to allow the number of examples required for learning t... more The learning model of Valiant is extended to allow the number of examples required for learning to depend on the particular concept to be learned, instead of requiring a uniform bound for all concepts of a concept class. This extension, called nonuniform learning, enables learning many concept classes not learnable by the previous definitions. Nonuniformly learnable concept classes are characterized. Some examples (Boolean formulae, recursive, and r.e. sets) are shown to be nonuniformly learnable by a polynomial (in the size of the representation of the concept and in the error parameters) number of examples, but not necessarily in polynomial time. Restricting the learning protocol such that the learner has to commit himself after a finite number of examples does not affect the concept classes which can be learned. An extension of nonuniform learnability to nonuniform learnability with respect to specific distributions is presented.

Language Resources and Evaluation, 2013

Your article is protected by copyright and all rights are held exclusively by Springer Science +B... more Your article is protected by copyright and all rights are held exclusively by Springer Science +Business Media Dordrecht. This e-offprint is for personal use only and shall not be selfarchived in electronic repositories. If you wish to self-archive your article, please use the accepted manuscript version for posting on your own website. You may further deposit the accepted manuscript version in any repository, provided it is only made publicly available 12 months after official publication or later and provided acknowledgement is given to the original source of publication and a link is inserted to the published article on Springer's website. The link must be accompanied by the following text: "The final publication is available at link.springer.com".

This paper presents a new approach for resolving lexical ambiguities in one language using statis... more This paper presents a new approach for resolving lexical ambiguities in one language using statis-tical data from a monolingual corpus of another language. This approach exploits the differences between mappings of words to senses in different languages. The paper concentrates on the prob-lem of target word selection in machine translation, for which the approach is directly applicable. The presented algorithm identifies syntactic relations between words, using a source language parser, and maps the alternative interpretations of these relations to the target language, using a bilingual lexicon. The preferred senses are then selected according to statistics on lexical relations in the target language. The selection is based on a statistical model and on a constraint prop-agation algorithm, which simultaneously handles all ambiguities in the sentence. The method was evaluated using three sets of Hebrew and German examples and was found to be very use-ful for disambiguation. The paper...

A parallel implementation of Ray Tracing is described. Alternative implementation choices are exa... more A parallel implementation of Ray Tracing is described. Alternative implementation choices are examined: organizing local and global memory, paging policies, task allocation, task granularity, and scanning order. These issues are discussed both analytically and experimentally. Extensive experiments were conducted to compare between different methods, and determine the best choice of parameters. The foremost conclusion is that in our parallel implementation, communication is the bottleneck, and that increasing the amount of memory reduces the communication.

This paper describes the process of building the first tree-bank for Modern Hebrew texts. A major... more This paper describes the process of building the first tree-bank for Modern Hebrew texts. A major concern in this process is the need for reducing the cost of manual annotation by the use of automatic means. To this end, the joint utility of an automatic morphological analyzer, a probabilistic parser and a small manually annotated tree-bank was explored.

We consider PAC-learning where the dis-tribution is known to the student. The problem addressed h... more We consider PAC-learning where the dis-tribution is known to the student. The problem addressed here is characterizing when learnability with respect to distri-bution D1 implies learnability with re-spect to distribution Dz. The answer to the above question de-pends on the learnability model. If the number of examples need not be bounded by a polynomial, it is sufficient to require that all sets which have zero probability with respect to Dz have zero probability with respect to D1. If the number of examples is required to be polynomial, then the probability with re-spect to Dz must be bounded by a multi-plicative constant from that of D1. More stringent conditions must hold if we in-sist that every hypothesis consistent with the examples be close to the target. Finally, we address the learnability prop-erties of classes of distributions.

study the problem of distributed leader election in an asynchronous complete network, in presence... more study the problem of distributed leader election in an asynchronous complete network, in presence of faults that occurred prior to the execution of the election algorithm. Failures of this type are encountered, for example, during a recovery from a crash in the network. For a network with n processors, k of which start the algorithm and at most t of which might he faulty, we present an algorithm that uses at most O(n log k + n + kt) messages. We prove that this algorithm is optimal. We also present an optimal algorithm for the case where the identities of the neighbors are known. It is interesting to note that the order of the message complexity of a t-resilient algorithm is not always higher than that of a nonresilient one. The &resilient algorithm is a systematic modification of an existing algorithm for a faultfree network. Index Terms-Complete networks, distributed algorithms, fault-tolerance, leader election, message complexity. I.

Database systems view data in different ways. Some regard the data as nodes in a network, others ... more Database systems view data in different ways. Some regard the data as nodes in a network, others as tuples in a table and others as abstract objects. There are a number of traditional techniques for implementing and indexing each of these database systems. For example, the B-tree is a commonly used index structure. The JS storage engine is a novel way of viewing and representing data. This technology deals with all keys as character strings, which are stored in a layered index (a trie-like data structure). With this technology we can: • Search for an item by any of its keys, within a small number of disk accesses. • Quickly access logically related items. • Ensure referential integrity. • Efficiently implement complex operations such as joins. • Efficiently process the data sequentially according to any of its keys. • Simultaneously view the same physical implementation by several database models (relational, object-oriented, etc). The JS therefore allows a great deal of flexibility...

The paper discusses searching a corpus for linguistic patterns. Semitic languages have complex mo... more The paper discusses searching a corpus for linguistic patterns. Semitic languages have complex morphology and ambiguous writing systems. We explore the properties of Semitic Languages that challenge linguistic search and describe how we used the Corpus Workbench (CWB) to enable linguistic searches in Hebrew corpora.

The research focuses on automatic construction of multi-lingual domain-ontologies, i.e., creating... more The research focuses on automatic construction of multi-lingual domain-ontologies, i.e., creating a DAG (directed acyclic graph) consisting of concepts relating to a specific domain and the relations between them. The domain example on which the research performed is “Organized Crime”. The contribution of the work is the investigation of and comparison between several data sources and methods to create multi-lingual ontologies. The first subtask was to extract the domain’s concepts. The best source turned out to be Wikepedia’s articles that are under the catgegory. The second task was to create an English ontology, i.e., the relationships between the concepts. Again the relationships between concepts and the hierarchy were derived from Wikipedia. The final task was to create an ontology for a language with far fewer resources (Hebrew). The task was accomplished by deriving the concepts from the Hebrew Wikepedia and assessing their relevance and the relationships between them from th...

Annals of Mathematics and Artificial Intelligence

Acronyms-words formed from the initial letters of a phrase-are important for various natural lang... more Acronyms-words formed from the initial letters of a phrase-are important for various natural language processing applications, including information retrieval and machine translation. While hand-crafted acronym dictionaries exist, they are limited and require frequent updates. We present a new machine-learningbased approach to automatically build an acronym dictionary from unannotated texts. This is the first such technique that specifically handles non-local acronyms, i.e., that can determine an acronym's expansion even when the expansion does not appear in the same document as the acronym. Our approach automatically enhances the dictionary with contextual information to help address the acronym disambiguation task (selecting the most appropriate expansion for a given acronym in context), outperforming dictionaries built using prior techniques. We apply the approach to Modern Hebrew, a language with a long tradition of using acronyms, in which the productive morphology and unique orthography adds to the complexity of the problem.

Siam Journal on Computing, Aug 1, 1993

Encyclopedia of Algorithms, 2016

Algorithms for adaptive routing in a high speed networks are examined. In such networks the inter... more Algorithms for adaptive routing in a high speed networks are examined. In such networks the intermediate nodes serve as switching elements with limited computational resources. The mode of operation is to specify an entire path from the source to the destination. If all the links of the path are nonfaulty then the message is transmitted; otherwise an indication is given to the first link that has failed. The source, upon learning of a failure, chooses another path and retransmits the message. The links of the network have independent a priori failure probablities. The status of the links does not change while attempting to transmit a message. First, the greedy algorithm (the algorithm which tries the paths in the order of increasing failure probability) is examined. It is shown to be optimal for a small class of networks, but not for all. The main result concerns series/parallel graphs, for which an optimal algorithm is presented. It is shown that an optimal algorithm can be easily ...

Hebrew, as other ~emitic languages, has a rich morpl1ology, observable in part by the complexity ... more Hebrew, as other ~emitic languages, has a rich morpl1ology, observable in part by the complexity of verb inflections. The primary base of verbs in Hebrew is the past third singular form of tlfe verb. From this base, some twenty eight different inflected forms can be created according to tense, per~on, gender and number. Traditionally, inflection tables were used to describe the various inflected forms derived from the verb 'base. Research done by Oman has managed to describe the verb inflection process using the principles of Generative Grammar. In' this approach, inflCfted verb forms are viewed as constructs of the form preftx+base+sufftx. Verb inflection is described as a s~ries of sequentialpperations. The first stage converts the primary verb base to a secondary'base, when the secondary base is not the same as the primary base. Secondly, the appropriate prefix and/or suffix are concatenated to the base. Thirdly, several morpho-phonemic changes due to the affix concat...

Author(s): Dechter, Rina; Itai, Alon | Abstract: We address the problem of enumerating (producing... more Author(s): Dechter, Rina; Itai, Alon | Abstract: We address the problem of enumerating (producing) all models of a given theory. We show that the enumeration task can be performed in time proportional to the product of the number of models and the effort needed to generate each model in isolation. In other words, the requirement of generating a new solution in each iteration does not in itself introduce substantial complexity. Consequently, it is possible to decide whether any tractably satisfiable formula has more than K solutions in time polynomial in the size of the forumla and in K. In the special cases of Horn formulas and 2-CNFs, although counting is #P-complete, to decide whether the count exceeds K, is polynomial in K.

This paper proposes a new approach for acquiring morpho-lexical probabilities from an untagged co... more This paper proposes a new approach for acquiring morpho-lexical probabilities from an untagged corpus. This approach demonstrates a way to extract very useful and nontrivial information from an untagged corpus, which otherwise would require laborious tagging of large corpora. The paper describes the use of these morpho-lexical probabilities as an information source for morphological disambiguation in Hebrew. The suggested method depends primarily on the following property: a lexical entry in Hebrew may have many different word forms, some of which are ambiguous and some of which are not. Thus, the disambiguation of a given word can be achieved using other word forms of the same lexical entry. Even though it was originally devised and implemented for dealing with the morphological ambiguity problem in Hebrew, the basic idea can be extended and used to handle similar problems in other languages with rich morphology.

Journal of Computer and System Sciences, 1994

The learning model of Valiant is extended to allow the number of examples required for learning t... more The learning model of Valiant is extended to allow the number of examples required for learning to depend on the particular concept to be learned, instead of requiring a uniform bound for all concepts of a concept class. This extension, called nonuniform learning, enables learning many concept classes not learnable by the previous definitions. Nonuniformly learnable concept classes are characterized. Some examples (Boolean formulae, recursive, and r.e. sets) are shown to be nonuniformly learnable by a polynomial (in the size of the representation of the concept and in the error parameters) number of examples, but not necessarily in polynomial time. Restricting the learning protocol such that the learner has to commit himself after a finite number of examples does not affect the concept classes which can be learned. An extension of nonuniform learnability to nonuniform learnability with respect to specific distributions is presented.

Language Resources and Evaluation, 2013

Your article is protected by copyright and all rights are held exclusively by Springer Science +B... more Your article is protected by copyright and all rights are held exclusively by Springer Science +Business Media Dordrecht. This e-offprint is for personal use only and shall not be selfarchived in electronic repositories. If you wish to self-archive your article, please use the accepted manuscript version for posting on your own website. You may further deposit the accepted manuscript version in any repository, provided it is only made publicly available 12 months after official publication or later and provided acknowledgement is given to the original source of publication and a link is inserted to the published article on Springer's website. The link must be accompanied by the following text: "The final publication is available at link.springer.com".

This paper presents a new approach for resolving lexical ambiguities in one language using statis... more This paper presents a new approach for resolving lexical ambiguities in one language using statis-tical data from a monolingual corpus of another language. This approach exploits the differences between mappings of words to senses in different languages. The paper concentrates on the prob-lem of target word selection in machine translation, for which the approach is directly applicable. The presented algorithm identifies syntactic relations between words, using a source language parser, and maps the alternative interpretations of these relations to the target language, using a bilingual lexicon. The preferred senses are then selected according to statistics on lexical relations in the target language. The selection is based on a statistical model and on a constraint prop-agation algorithm, which simultaneously handles all ambiguities in the sentence. The method was evaluated using three sets of Hebrew and German examples and was found to be very use-ful for disambiguation. The paper...

A parallel implementation of Ray Tracing is described. Alternative implementation choices are exa... more A parallel implementation of Ray Tracing is described. Alternative implementation choices are examined: organizing local and global memory, paging policies, task allocation, task granularity, and scanning order. These issues are discussed both analytically and experimentally. Extensive experiments were conducted to compare between different methods, and determine the best choice of parameters. The foremost conclusion is that in our parallel implementation, communication is the bottleneck, and that increasing the amount of memory reduces the communication.

This paper describes the process of building the first tree-bank for Modern Hebrew texts. A major... more This paper describes the process of building the first tree-bank for Modern Hebrew texts. A major concern in this process is the need for reducing the cost of manual annotation by the use of automatic means. To this end, the joint utility of an automatic morphological analyzer, a probabilistic parser and a small manually annotated tree-bank was explored.

We consider PAC-learning where the dis-tribution is known to the student. The problem addressed h... more We consider PAC-learning where the dis-tribution is known to the student. The problem addressed here is characterizing when learnability with respect to distri-bution D1 implies learnability with re-spect to distribution Dz. The answer to the above question de-pends on the learnability model. If the number of examples need not be bounded by a polynomial, it is sufficient to require that all sets which have zero probability with respect to Dz have zero probability with respect to D1. If the number of examples is required to be polynomial, then the probability with re-spect to Dz must be bounded by a multi-plicative constant from that of D1. More stringent conditions must hold if we in-sist that every hypothesis consistent with the examples be close to the target. Finally, we address the learnability prop-erties of classes of distributions.

study the problem of distributed leader election in an asynchronous complete network, in presence... more study the problem of distributed leader election in an asynchronous complete network, in presence of faults that occurred prior to the execution of the election algorithm. Failures of this type are encountered, for example, during a recovery from a crash in the network. For a network with n processors, k of which start the algorithm and at most t of which might he faulty, we present an algorithm that uses at most O(n log k + n + kt) messages. We prove that this algorithm is optimal. We also present an optimal algorithm for the case where the identities of the neighbors are known. It is interesting to note that the order of the message complexity of a t-resilient algorithm is not always higher than that of a nonresilient one. The &resilient algorithm is a systematic modification of an existing algorithm for a faultfree network. Index Terms-Complete networks, distributed algorithms, fault-tolerance, leader election, message complexity. I.

Database systems view data in different ways. Some regard the data as nodes in a network, others ... more Database systems view data in different ways. Some regard the data as nodes in a network, others as tuples in a table and others as abstract objects. There are a number of traditional techniques for implementing and indexing each of these database systems. For example, the B-tree is a commonly used index structure. The JS storage engine is a novel way of viewing and representing data. This technology deals with all keys as character strings, which are stored in a layered index (a trie-like data structure). With this technology we can: • Search for an item by any of its keys, within a small number of disk accesses. • Quickly access logically related items. • Ensure referential integrity. • Efficiently implement complex operations such as joins. • Efficiently process the data sequentially according to any of its keys. • Simultaneously view the same physical implementation by several database models (relational, object-oriented, etc). The JS therefore allows a great deal of flexibility...

The paper discusses searching a corpus for linguistic patterns. Semitic languages have complex mo... more The paper discusses searching a corpus for linguistic patterns. Semitic languages have complex morphology and ambiguous writing systems. We explore the properties of Semitic Languages that challenge linguistic search and describe how we used the Corpus Workbench (CWB) to enable linguistic searches in Hebrew corpora.

The research focuses on automatic construction of multi-lingual domain-ontologies, i.e., creating... more The research focuses on automatic construction of multi-lingual domain-ontologies, i.e., creating a DAG (directed acyclic graph) consisting of concepts relating to a specific domain and the relations between them. The domain example on which the research performed is “Organized Crime”. The contribution of the work is the investigation of and comparison between several data sources and methods to create multi-lingual ontologies. The first subtask was to extract the domain’s concepts. The best source turned out to be Wikepedia’s articles that are under the catgegory. The second task was to create an English ontology, i.e., the relationships between the concepts. Again the relationships between concepts and the hierarchy were derived from Wikipedia. The final task was to create an ontology for a language with far fewer resources (Hebrew). The task was accomplished by deriving the concepts from the Hebrew Wikepedia and assessing their relevance and the relationships between them from th...

Annals of Mathematics and Artificial Intelligence

Acronyms-words formed from the initial letters of a phrase-are important for various natural lang... more Acronyms-words formed from the initial letters of a phrase-are important for various natural language processing applications, including information retrieval and machine translation. While hand-crafted acronym dictionaries exist, they are limited and require frequent updates. We present a new machine-learningbased approach to automatically build an acronym dictionary from unannotated texts. This is the first such technique that specifically handles non-local acronyms, i.e., that can determine an acronym's expansion even when the expansion does not appear in the same document as the acronym. Our approach automatically enhances the dictionary with contextual information to help address the acronym disambiguation task (selecting the most appropriate expansion for a given acronym in context), outperforming dictionaries built using prior techniques. We apply the approach to Modern Hebrew, a language with a long tradition of using acronyms, in which the productive morphology and unique orthography adds to the complexity of the problem.

Siam Journal on Computing, Aug 1, 1993

Encyclopedia of Algorithms, 2016