Latent Semantic Mapping: Principles And Applications (Synthesis Lectures on Speech and Audio Process

Latent Semantic Mapping: Principles And Applications (Synthesis Lectures on Speech and Audio Processing)

ABSTRACT

Latent semantic mapping (LSM) is a generalization of latent semantic analysis (LSA), a paradigm originally developed to capture hidden word patterns in a text document corpus. In information retrieval, LSA enables retrieval on the basis of conceptual content, instead of merely matching words between queries and documents. It operates under the assumption that there is some latent semantic structure in the data, which is partially obscured by the randomness of word choice with respect to retrieval. Algebraic and/or statistical techniques are brought to bear to estimate this structure and get rid of the obscuring ¡°noise.¡± This results in a parsimonious continuous parameter description of words and documents, which then replaces the original parameterization in indexing and retrieval. This approach exhibits three main characteristics:

* discrete entities (words and documents) are mapped onto a continuous vector space;
* this mapping is determined by global correlation patterns; and
* dimensionality reduction is an integral part of the process.

Such fairly generic properties are advantageous in a variety of different contexts, which motivates a broader interpretation of the underlying paradigm. The outcome (LSM) is a data-driven framework for modeling meaningful global relationships implicit in large volumes of (not necessarily textual) data. This monograph gives a general overview of the framework, and underscores the multifaceted benefits it can bring to a number of problems in natural language understanding and spoken language processing. It concludes with a discussion of the inherent tradeoffs associated with the approach, and some perspectives on its general applicability to data-driven information extraction.

KEYWORDS

natural language processing, long-span dependencies, data-driven modeling, parsimonious representation, singular value decomposition.

CONTENTS

I. Principles

1. Introduction
1.1 Motivation
1.2 From LSA to LSM
1.3 Organization
1.3.1 Part I
1.3.2 Part II
1.3.3 Part III

2. Latent Semantic Mapping
2.1 Co-occurrence Matrix
2.2 Vector Representation
2.2.1 Singular Value Decomposition
2.2.2 SVD Properties
2.3 Interpretation

3. LSM Feature Space
3.1 Closeness Measures
3.1.1 Unit¨CUnit Comparisons
3.1.2 Composition¨CComposition Comparisons
3.1.3 Unit¨CComposition Comparisons
3.2 LSM Framework Extension
3.3 Salient Characteristics

4. Computational Effort
4.1 Off¨CLine Cost
4.2 Online Cost
4.3 Possible Shortcuts
4.3.1 Incremental SVD Implementations
4.3.2 Other Matrix Decompositions
4.3.3 Alternative Formulations

5. Probabilistic Extensions
5.1 Dual Probability Model
5.1.1 Composition Model
5.1.2 Unit Model
5.1.3 Comments
5.2 Probabilistic Latent Semantic Analysis
5.3 Inherent Limitations

II. Applications

6. Junk E¨CMail Filtering
6.1 Conventional Approaches
6.1.1 Header Analysis
6.1.2 Rule-Based Predicates
6.1.3 Machine Learning Approaches
6.2 LSM-Based Filtering
6.3 Performance

7. Semantic Classification
7.1 Underlying Issues
7.1.1 Case Study: Desktop Interface Control
7.1.2 Language Modeling Constraints
7.2 Semantic Inference
7.2.1 Framework
7.2.2 Illustration
7.3 Caveats

8. Language Modeling
8.1 N-Gram Limitations
8.2 MultiSpan Language Modeling
8.2.1 Hybrid Formulation
8.2.2 Context Scope Selection
8.2.3 LSM Probability
8.3 Smoothing
8.3.1 Word Smoothing
8.3.2 Document Smoothing
8.3.3 Joint Smoothing

9. Pronunciation Modeling
9.1 Grapheme-to-Phoneme Conversion
9.1.1 Top¨CDown Approaches
9.1.2 Illustration
9.1.3 Bottom¨CUp Approaches
9.2 Pronunciation by Latent Analogy
9.2.1 Orthographic Neighborhoods
9.2.2 Sequence Alignment

10. Speaker Verification
10.1 The Task
10.2 LSM-based speaker verification
10.2.1 Single-Utterance Representation
10.2.2 LSM-Tailored Metric
10.2.3 Integration with DTW

11. TTS Unit Selection
11.1 Concatenative Synthesis
11.2 LSM-Based Unit Selection
11.2.1 Feature Extraction
11.2.2 Comparison to Fourier Analysis
11.2.3 Properties
11.3 LSM-Based Boundary Training

III. Perspectives

12. Discussion
12.1 Inherent Tradeoffs
12.1.1 Descriptive Power
12.1.2 Domain Sensitivity
12.1.3 Adaptation Capabilities
12.2 General Applicability
12.2.1 Natural Language Processing
12.2.2 Generic Pattern Recognition

13. Conclusion
13.1 Summary
13.2 Perpectives

Bibliography
Author Biography
Password: ebooksclub.org
File size: 0.7 MB
Format: PDF

http://mihd.net/crai6j

http://rapidshare.com/files/54414571/1598291041.rar