| Description |
1 online resource (xiii, 307 pages) : color illustrations |
| Series |
Synthesis lectures on human language technologies, 1947-4059 ; #53 |
|
Synthesis lectures on human language technologies ; #53.
|
| Contents |
1. Introduction -- 1.1. Text ranking problems -- 1.2. A brief history -- 1.3. Roadmap, assumptions, and omissions |
|
2. Setting the stage -- 2.1. Texts -- 2.2. Information needs -- 2.3. Relevance -- 2.4. Relevance judgments -- 2.5. Ranking metrics -- 2.6. Community evaluations and reusable test collections -- 2.7. Descriptions of common test collections -- 2.8. Keyword search -- 2.9. Notes on parlance |
|
3. Multi-stage architectures for reranking -- 3.1. A high-level overview of BERT -- 3.2. Simple relevance classification : monoBERT -- 3.3. From passage to document ranking -- 3.4. From single-stage to multi-stage rerankers -- 3.5. Beyond BERT -- 3.6. Concluding thoughts |
|
4. Refining query and document representations -- 4.1. Query and document expansion : general remarks -- 4.2. Pseudo-relevance feedback with contextualized embeddings : CEQE -- 4.3. Document expansion via query prediction : doc2query -- 4.4. Term reweighting as regression : DeepCT -- 4.5. Term reweighting with weak supervision : HDCT -- 4.6. Combining term expansion with term weighting : DeepImpact -- 4.7. Expansion of query and document representations -- 4.8. Concluding thoughts |
|
5. Learned dense representations for ranking -- 5.1. Task formulation -- 5.2. Nearest neighbor search -- 5.3. Pre-BERT text representations for ranking -- 5.4. Simple transformer bi-encoders for ranking -- 5.5. Enhanced transformer bi-encoders for ranking -- 5.6. Knowledge distillation for transformer bi-encoders -- 5.7. Concluding thoughts |
|
6. Future directions and conclusions -- 6.1. Notable content omissions -- 6.2. Open research questions -- 6.3. Final thoughts |
| Summary |
The goal of text ranking is to generate an ordered list of texts retrieved from a corpus in response to a query. Although the most common formulation of text ranking is search, instances of the task can also be found in many natural language processing (NLP) applications.This book provides an overview of text ranking with neural network architectures known as transformers, of which BERT (Bidirectional Encoder Representations from Transformers) is the best-known example. The combination of transformers and self-supervised pretraining has been responsible for a paradigm shift in NLP, information retrieval (IR), and beyond. This book provides a synthesis of existing work as a single point of entry for practitioners who wish to gain a better understanding of how to apply transformers to text ranking problems and researchers who wish to pursue work in this area. It covers a wide range of modern techniques, grouped into two high-level categories: transformer models that perform reranking in multi-stage architectures and dense retrieval techniques that perform ranking directly. Two themes pervade the book: techniques for handling long documents, beyond typical sentence-by-sentence processing in NLP, and techniques for addressing the tradeoff between effectiveness (i.e., result quality) and efficiency (e.g., query latency, model and index size). Although transformer architectures and pretraining techniques are recent innovations, many aspects of how they are applied to text ranking are relatively well understood and represent mature techniques. However, there remain many open research questions, and thus in addition to laying out the foundations of pretrained transformers for text ranking, this book also attempts to prognosticate where the field is heading |
| Bibliography |
Includes bibliographical references |
| Subject |
Text processing (Computer science)
|
|
Information storage and retrieval systems.
|
|
Word Processing
|
|
Information Systems
|
|
Information storage and retrieval systems
|
|
Text processing (Computer science)
|
| Form |
Electronic book
|
| Author |
Nogueira, Rodrigo, author
|
|
Yates, Andrew, author
|
| ISBN |
1636392296 |
|
9781636392295 |
|
