Kadri Hacioglu - Academia.edu (original) (raw)
Papers by Kadri Hacioglu
The system described in this document represents CU's first submission to the ACE program for... more The system described in this document represents CU's first submission to the ACE program for the major tasks. In 2004, we submitted English and Chinese systems to the TERN evaluation for temporal expression recognition with competitive performances. Encouraged by our successful participation in TERN we showed interest in the ACE 2005 Evaluation and decided to participate in that evaluation. Our efforts towards that goal have started by the end of 2004 and partially supported by the ongoing ARDA AQUAINT project at CU for developing intelligent question answering systems. The motivation was that the creation of structured text from raw text, as envisioned by the ACE program, would facilitate search, retrieval and ranking of relevant answers sought by users for their information need expressed as complex questions. Our main goals were to first demonstrate competitive performance with the earlier ACE2004 systems and then extend the resulting system for the ACE2005 evaluation. This ...
IEEE Transactions on Signal Processing, 2000
This correspondence corrects and clarifies some notation inconsistencies and the absence of some ... more This correspondence corrects and clarifies some notation inconsistencies and the absence of some definitions in a previous paper by Ong et al.. In addition, we show that the assumptions made in the derivation of the MSE DFRNE-MSE RNE invalidate the final result given in Ong et al.and lead to an unfair comparison between the two equalizers. We present a simulation that best compares the two equalizers.
In this paper, we present a semantic role labeler (or chunker) that groups syntactic chunks (i.e.... more In this paper, we present a semantic role labeler (or chunker) that groups syntactic chunks (i.e. base phrases) into the arguments of a predicate. This is accomplished by casting the semantic labeling as the classification of syntactic chunks (e.g. NP-chunk, PP-chunk) into one of several classes such as the beginning of an argument (B-ARG), inside an argument (I-ARG) and outside an argument (O). This amounts to tagging syntactic chunks with semantic labels using the IOB representation. The chunker is realized using support vector machines as oneversus-all classifiers. We describe the representation of data and information used to accomplish the task. We participate in the “closed challenge” of the CoNLL-2004 shared task and report results on both development and test sets.
A new Decision-Feedback-Equalisation (DFE) technique is introduced, namely, the Sample-Diversity ... more A new Decision-Feedback-Equalisation (DFE) technique is introduced, namely, the Sample-Diversity (SD) DFE. To realise a sample-diversity at the DFE input, N samples are taken from each symbol and these samples are used to drive N different DFE’s. A simple selection mechanism is introduced to select the best sampling phase and DFE to be used in the tracking mode. Through simulations, for normalized-rms-delay-spread over the range from 0.01 to 1, the SDDFE is shown to have a much better performance than a conventional DFE. Being a combination of diversity and equalisation techniques, the SD approach is shown as a promising technique for symbol detection in multipath fading channels.
This paper describes recent work towards development of new corpora and tools for Turkish speech ... more This paper describes recent work towards development of new corpora and tools for Turkish speech research. This effort represents an on-going collaboration between the Center for Spoken Language Research (CSLR) at the University of Colorado and the Department of Electrical Engineering at the Middle East Technical University (METU). A new text corpus developed from Turkish newspapers’ text is described. In addition, a 193-speaker audio corpus and pronunciation lexicon for the Turkish language is developed. We then describe our initial work towards porting Sonic, the CSLR speech recognition system, to the Turkish language. Results are shown for phonetic alignment and phoneme recognition accuracy using the newly constructed corpus and speech tools. It is shown that 91.2% of the automatically labeled phoneme boundaries are placed within 20 msec of hand-labeled locations for the Turkish audio corpus. Finally, a phoneme recognition error rate of 29.3% is demonstrated.
In this paper, we address the lexicon design problem in Turkish large vocabulary speech recogniti... more In this paper, we address the lexicon design problem in Turkish large vocabulary speech recognition. Although we focus only on Turkish, the methods described here are general enough that they can be considered for other agglutinative languages like Finnish, Korean etc. In an agglutinative language, several words can be created from a single root word using a rich collection of morphological rules. So, a virtually infinite size lexicon is required to cover the language if words are used as the basic units. The standard approach to this problem is to discover a number of primitive units so that a large set of words can be created by compounding those units. Two broad classes of methods are available for splitting words into their sub-units; morphology-based and data-driven methods. Although the word splitting significantly reduces the out of vocabulary rate, it shrinks the context and increases acoustic confusibility. We have used two methods to address the latter. In one method, we u...
In this paper, we propose a machine learning algorithm for shallow semantic parsing, extending th... more In this paper, we propose a machine learning algorithm for shallow semantic parsing, extending the work of Gildea and Jurafsky (2002), Surdeanu et al. (2003) and others. Our algorithm is based on Support Vector Machines which we show give an improvement in performance over earlier classifiers. We show performance improvements through a number of new features and measure their ability to generalize to a new test set drawn from the AQUAINT corpus.
In this paper, a framework for the development of a fast, accurate, and highly portable semantic ... more In this paper, a framework for the development of a fast, accurate, and highly portable semantic chunker is introduced. The framework is based on a non-overlapping, shallow tree-structured language. The derivation of the tree is considered as a sequence of tagging actions in a predefined linguistic context, and a novel semantic chunker is accordingly developed. It groups the phrase chunks into the arguments of a given predicate in a bottom-up fashion. This is quite different from current approaches to semantic parsing or chunking that depend on full statistical syntactic parsers that require tree bank style annotation. We compare it with a recently proposed word-byword semantic chunker and present results that show that the phrase-by-phrase approach performs better than its word-by-word counterpart.
ABSTRACT This paper presents a vision of the near future in which computer interaction is charact... more ABSTRACT This paper presents a vision of the near future in which computer interaction is characterized by natural face-to-face conversations with lifelike characters that speak, emote, and gesture. These animated agents will converse with people much like people converse effectively with assistants in a variety of focused applications. Despite the research advances required to realize this vision, and the lack of strong experimental evidence that animated agents improve human--computer interaction, we argue that initial prototypes of perceptive animated interfaces can be developed today, and that the resulting systems will provide more effective and engaging communication experiences than existing systems. In support of this hypothesis, we first describe initial experiments using an animated character to teach speech and language skills to children with hearing problems, and classroom subjects and social skills to children with autistic spectrum disorder. We then show how existing dialogue system architectures can be transformed into perceptive animated interfaces by integrating computer vision and animation capabilities. We conclude by describing the Colorado Literacy Tutor, a computer-based literacy program that provides an ideal testbed for research and development of perceptive animated interfaces, and consider next steps required to realize the vision
Naacl, 2005
In this paper, we describe an integrated approach to entity mention detection that yields a monol... more In this paper, we describe an integrated approach to entity mention detection that yields a monolithic, almost language independent system. It is optimal in the sense that all categorical constraints are simultaneously considered. The system is compact and easy to develop and maintain, since only a single set of features and classifiers are needed to be designed and optimized. It is implemented using oneversus-all support vector machine (SVM) classifiers and a number of feature extractors at several linguistic levels. SVMs are well known for their ability to handle a large set of overlapping features with theoretically sound generalization properties. Data sparsity might be an important issue as a result of a large number of classes and relatively moderate training data size. However, we report results that the integrated system performs as good as a pipelined system that decomposes the problem into a few smaller subtasks. We conduct all our experiments using ACE 2004 data, evaluate the systems using ACE metrics and report competitive performance.
This paper describes recent work towards development of new corpora and tools for Turkish speech ... more This paper describes recent work towards development of new corpora and tools for Turkish speech research. This effort represents an ongoing collaboration between the Center for Spoken Language Research (CSLR) at the University of Colorado and the Department of Electrical Engineering at the Middle East Technical University (METU). A new text corpus developed from Turkish newspapers' text is described. In addition, a 193-speaker audio corpus and pronunciation lexicon for the Turkish language is developed. We then describe our initial work towards porting Sonic, the CSLR speech recognition system, to the Turkish language. Results are shown for phonetic alignment and phoneme recognition accuracy using the newly constructed corpus and speech tools. It is shown that 91.2% of the automatically labeled phoneme boundaries are placed within 20 msec of hand-labeled locations for the Turkish audio corpus. Finally, a phoneme recognition error rate of 29.3% is demonstrated.
Human Language Technology, 2001
Proceedings of the 2003 Conference of the North American Chapter of the Association for Computational Linguistics on Human Language Technology companion volume of the Proceedings of HLT-NAACL 2003--short papers - NAACL '03, 2003
Proceedings of the 2003 Conference of the North American Chapter of the Association for Computational Linguistics on Human Language Technology companion volume of the Proceedings of HLT-NAACL 2003--short papers - NAACL '03, 2003
The system described in this document represents CU's first submission to the ACE program for... more The system described in this document represents CU's first submission to the ACE program for the major tasks. In 2004, we submitted English and Chinese systems to the TERN evaluation for temporal expression recognition with competitive performances. Encouraged by our successful participation in TERN we showed interest in the ACE 2005 Evaluation and decided to participate in that evaluation. Our efforts towards that goal have started by the end of 2004 and partially supported by the ongoing ARDA AQUAINT project at CU for developing intelligent question answering systems. The motivation was that the creation of structured text from raw text, as envisioned by the ACE program, would facilitate search, retrieval and ranking of relevant answers sought by users for their information need expressed as complex questions. Our main goals were to first demonstrate competitive performance with the earlier ACE2004 systems and then extend the resulting system for the ACE2005 evaluation. This ...
IEEE Transactions on Signal Processing, 2000
This correspondence corrects and clarifies some notation inconsistencies and the absence of some ... more This correspondence corrects and clarifies some notation inconsistencies and the absence of some definitions in a previous paper by Ong et al.. In addition, we show that the assumptions made in the derivation of the MSE DFRNE-MSE RNE invalidate the final result given in Ong et al.and lead to an unfair comparison between the two equalizers. We present a simulation that best compares the two equalizers.
In this paper, we present a semantic role labeler (or chunker) that groups syntactic chunks (i.e.... more In this paper, we present a semantic role labeler (or chunker) that groups syntactic chunks (i.e. base phrases) into the arguments of a predicate. This is accomplished by casting the semantic labeling as the classification of syntactic chunks (e.g. NP-chunk, PP-chunk) into one of several classes such as the beginning of an argument (B-ARG), inside an argument (I-ARG) and outside an argument (O). This amounts to tagging syntactic chunks with semantic labels using the IOB representation. The chunker is realized using support vector machines as oneversus-all classifiers. We describe the representation of data and information used to accomplish the task. We participate in the “closed challenge” of the CoNLL-2004 shared task and report results on both development and test sets.
A new Decision-Feedback-Equalisation (DFE) technique is introduced, namely, the Sample-Diversity ... more A new Decision-Feedback-Equalisation (DFE) technique is introduced, namely, the Sample-Diversity (SD) DFE. To realise a sample-diversity at the DFE input, N samples are taken from each symbol and these samples are used to drive N different DFE’s. A simple selection mechanism is introduced to select the best sampling phase and DFE to be used in the tracking mode. Through simulations, for normalized-rms-delay-spread over the range from 0.01 to 1, the SDDFE is shown to have a much better performance than a conventional DFE. Being a combination of diversity and equalisation techniques, the SD approach is shown as a promising technique for symbol detection in multipath fading channels.
This paper describes recent work towards development of new corpora and tools for Turkish speech ... more This paper describes recent work towards development of new corpora and tools for Turkish speech research. This effort represents an on-going collaboration between the Center for Spoken Language Research (CSLR) at the University of Colorado and the Department of Electrical Engineering at the Middle East Technical University (METU). A new text corpus developed from Turkish newspapers’ text is described. In addition, a 193-speaker audio corpus and pronunciation lexicon for the Turkish language is developed. We then describe our initial work towards porting Sonic, the CSLR speech recognition system, to the Turkish language. Results are shown for phonetic alignment and phoneme recognition accuracy using the newly constructed corpus and speech tools. It is shown that 91.2% of the automatically labeled phoneme boundaries are placed within 20 msec of hand-labeled locations for the Turkish audio corpus. Finally, a phoneme recognition error rate of 29.3% is demonstrated.
In this paper, we address the lexicon design problem in Turkish large vocabulary speech recogniti... more In this paper, we address the lexicon design problem in Turkish large vocabulary speech recognition. Although we focus only on Turkish, the methods described here are general enough that they can be considered for other agglutinative languages like Finnish, Korean etc. In an agglutinative language, several words can be created from a single root word using a rich collection of morphological rules. So, a virtually infinite size lexicon is required to cover the language if words are used as the basic units. The standard approach to this problem is to discover a number of primitive units so that a large set of words can be created by compounding those units. Two broad classes of methods are available for splitting words into their sub-units; morphology-based and data-driven methods. Although the word splitting significantly reduces the out of vocabulary rate, it shrinks the context and increases acoustic confusibility. We have used two methods to address the latter. In one method, we u...
In this paper, we propose a machine learning algorithm for shallow semantic parsing, extending th... more In this paper, we propose a machine learning algorithm for shallow semantic parsing, extending the work of Gildea and Jurafsky (2002), Surdeanu et al. (2003) and others. Our algorithm is based on Support Vector Machines which we show give an improvement in performance over earlier classifiers. We show performance improvements through a number of new features and measure their ability to generalize to a new test set drawn from the AQUAINT corpus.
In this paper, a framework for the development of a fast, accurate, and highly portable semantic ... more In this paper, a framework for the development of a fast, accurate, and highly portable semantic chunker is introduced. The framework is based on a non-overlapping, shallow tree-structured language. The derivation of the tree is considered as a sequence of tagging actions in a predefined linguistic context, and a novel semantic chunker is accordingly developed. It groups the phrase chunks into the arguments of a given predicate in a bottom-up fashion. This is quite different from current approaches to semantic parsing or chunking that depend on full statistical syntactic parsers that require tree bank style annotation. We compare it with a recently proposed word-byword semantic chunker and present results that show that the phrase-by-phrase approach performs better than its word-by-word counterpart.
ABSTRACT This paper presents a vision of the near future in which computer interaction is charact... more ABSTRACT This paper presents a vision of the near future in which computer interaction is characterized by natural face-to-face conversations with lifelike characters that speak, emote, and gesture. These animated agents will converse with people much like people converse effectively with assistants in a variety of focused applications. Despite the research advances required to realize this vision, and the lack of strong experimental evidence that animated agents improve human--computer interaction, we argue that initial prototypes of perceptive animated interfaces can be developed today, and that the resulting systems will provide more effective and engaging communication experiences than existing systems. In support of this hypothesis, we first describe initial experiments using an animated character to teach speech and language skills to children with hearing problems, and classroom subjects and social skills to children with autistic spectrum disorder. We then show how existing dialogue system architectures can be transformed into perceptive animated interfaces by integrating computer vision and animation capabilities. We conclude by describing the Colorado Literacy Tutor, a computer-based literacy program that provides an ideal testbed for research and development of perceptive animated interfaces, and consider next steps required to realize the vision
Naacl, 2005
In this paper, we describe an integrated approach to entity mention detection that yields a monol... more In this paper, we describe an integrated approach to entity mention detection that yields a monolithic, almost language independent system. It is optimal in the sense that all categorical constraints are simultaneously considered. The system is compact and easy to develop and maintain, since only a single set of features and classifiers are needed to be designed and optimized. It is implemented using oneversus-all support vector machine (SVM) classifiers and a number of feature extractors at several linguistic levels. SVMs are well known for their ability to handle a large set of overlapping features with theoretically sound generalization properties. Data sparsity might be an important issue as a result of a large number of classes and relatively moderate training data size. However, we report results that the integrated system performs as good as a pipelined system that decomposes the problem into a few smaller subtasks. We conduct all our experiments using ACE 2004 data, evaluate the systems using ACE metrics and report competitive performance.
This paper describes recent work towards development of new corpora and tools for Turkish speech ... more This paper describes recent work towards development of new corpora and tools for Turkish speech research. This effort represents an ongoing collaboration between the Center for Spoken Language Research (CSLR) at the University of Colorado and the Department of Electrical Engineering at the Middle East Technical University (METU). A new text corpus developed from Turkish newspapers' text is described. In addition, a 193-speaker audio corpus and pronunciation lexicon for the Turkish language is developed. We then describe our initial work towards porting Sonic, the CSLR speech recognition system, to the Turkish language. Results are shown for phonetic alignment and phoneme recognition accuracy using the newly constructed corpus and speech tools. It is shown that 91.2% of the automatically labeled phoneme boundaries are placed within 20 msec of hand-labeled locations for the Turkish audio corpus. Finally, a phoneme recognition error rate of 29.3% is demonstrated.
Human Language Technology, 2001
Proceedings of the 2003 Conference of the North American Chapter of the Association for Computational Linguistics on Human Language Technology companion volume of the Proceedings of HLT-NAACL 2003--short papers - NAACL '03, 2003
Proceedings of the 2003 Conference of the North American Chapter of the Association for Computational Linguistics on Human Language Technology companion volume of the Proceedings of HLT-NAACL 2003--short papers - NAACL '03, 2003