Get A Generative Theory of Relevance PDF

By Victor Lavrenko

ISBN-10: 3540893636

ISBN-13: 9783540893639

ISBN-10: 3540893644

ISBN-13: 9783540893646

A glossy info retrieval approach should have the potential to discover, manage and current very diverse manifestations of data – reminiscent of textual content, photographs, video clips or database documents – any of that may be of relevance to the consumer. in spite of the fact that, the concept that of relevance, whereas doubtless intuitive, is de facto not easy to outline, and it really is even more durable to version in a proper way.

Lavrenko doesn't try to bring about a brand new definition of relevance, nor offer arguments as to why any specific definition may be theoretically more desirable or extra whole. as a substitute, he's taking a generally approved, albeit slightly conservative definition, makes numerous assumptions, and from them develops a brand new probabilistic version that explicitly captures that inspiration of relevance. With this booklet, he makes significant contributions to the sector of data retrieval: first, a brand new method to examine topical relevance, complementing the 2 dominant types, i.e., the classical probabilistic version and the language modeling method, and which explicitly combines records, queries, and relevance in one formalism; moment, a brand new approach for modeling exchangeable sequences of discrete random variables which doesn't make any structural assumptions in regards to the facts and that could additionally deal with infrequent events.

Thus his ebook is of significant curiosity to researchers and graduate scholars in details retrieval who concentrate on relevance modeling, score algorithms, and language modeling.

Show description

Read Online or Download A Generative Theory of Relevance PDF

Similar structured design books

Scott Klein's Pro Entity Framework 4.0 PDF

Formerly, SQL builders were capable of nearly solely forget about the SQLCLR and deal with it as a peripheral technology—almost an extension to the most product. With the appearance of LINQ and the Entity Framework, this is often not the case, and the SQLCLR is relocating to the heart degree. It’s a robust product yet, for plenty of, it truly is a completely new method of operating with facts.

Euclidean Shortest Paths: Exact or Approximate Algorithms - download pdf or read online

The Euclidean shortest course (ESP) challenge asks the query: what's the course of minimal size connecting issues in a 2- or three-dimensional area? variations of this industrially-significant computational geometry challenge additionally require the trail to go through unique parts and steer clear of outlined stumbling blocks.

Handbook of Video Databases: Design and Applications - download pdf or read online

Know-how has spurred the expansion of massive snapshot and video libraries, many turning out to be into the loads of terabytes. for this reason there's a nice call for between enterprises for the layout of databases that could successfully aid the garage, seek, retrieval, and transmission of video info. Engineers and researchers within the box call for a entire reference that would aid them layout and enforce the main complicated video database initiatives.

Additional resources for A Generative Theory of Relevance

Example text

Consequently, PRP cannot handle issues of novelty and redundancy, or cases where two documents are relevant when put together, but irrelevant when viewed individually. Robertson [114] also cites a curious counter-example (due to Cooper) regarding the optimality of the principle. The counter-example considers the case when we are dealing with two classes of users who happen to issue the same request but consider different documents to be relevant to it. In that case PRP will only be optimal for the larger of the two user classes.

The probability ranking principle, as stated by Robertson is quite broad – it does not restrict us to any particular type of relevance, and the document representation D can be potentially very rich, covering not just the topical content of the document. In fact D could include features determining readability of the document, its relation to the user’s preferences or suitability for a particular task. Similarly, R may well refer to “pertinence” or any other complex notion of relevance. The only restriction imposed by the PRP is that relevance of a particular document be scalar and independent of any other document in the collection.

Multiple-Bernoulli language models Ponte and Croft represent queries in the same space that was used by Robertson and Sparck Jones in the Binary Independence Model. If V is a vocabulary of NV words, the query space Q is the set of all subsets of vocabulary ({0, 1}NV ). The query Q is a vector of NV binary variables Qv , one for each word v in the vocabulary. The components Qi are assumed to be mutually independent conditioned on the language model Md . The language model itself is a vector of NV probabilities pd,v , one for each word v.

Download PDF sample

A Generative Theory of Relevance by Victor Lavrenko


by Thomas
4.4

Rated 4.89 of 5 – based on 43 votes