DOC PREVIEW
Rational Analysis as a Link between Human Memory and Information Retrieval

This preview shows page 1-2-21-22 out of 22 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 22 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 22 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 22 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 22 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 22 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

IntroductionA Probabilistic Approach to Information RetrievalGoogle and the mind: predicting fluency with PageRankApplying PageRank to Semantic NetworksTopic Models to extract Verbatim and Gist informationTopic ModelsDual Route Topic ModelsExplaining Semantic Isolation EffectsExplaining False Memory effectsApplication to Information RetrievalDiscussionReferencesSteyvers, M. & Griffiths, T.L. (in press). Rational Analysis as a Link between Human Memory and Information Retrieval. In N. Chater and M Oaksford (Eds.) The Probabilistic Mind: Prospects from Rational Models of Cognition. Oxford University Press.Rational Analysis as a Link between Human Memory and Information Retrieval Mark Steyvers1 and Thomas L. Griffiths2 1Department of Cognitive Sciences, University of California, Irvine, CA 92697, USA 2Department of Psychology, University of California, Berkeley, CA 94720, USA. Corresponding author: Steyvers, M. ([email protected]) 1. Introduction Rational analysis has been successful in explaining a variety of different aspects of human cognition (Anderson, 1990; Chater & Oaksford, 1999; Marr, 1982; Oaksford & Chater, 1998). The explanations provided by rational analysis have two properties: they emphasize the connection between behavior and the structure of the environment, and they focus on the abstract computational problems being solved. These properties provide the opportunity to recognize connections between human cognition and other systems that solve the same computational problems, with the potential both to provide new insights into human cognition and to allow us to develop better systems for solving those problems. In particular, we should expect to find a correspondence between human cognition and systems that are successful at solving the same computational problems in a similar environment. In this chapter, we argue that such a correspondence exists between human memory and internet search, and show that this correspondence leads to both better models of human cognition, and better methods for searching the web. Anderson (1990) and Anderson and Schooler (1991; 2000) have shown that many findings in the memory literature related to recognition and recall of lists of words can be understood by considering the computational problem of assessing the relevance of an item in memory to environmental cues. They showed a close correspondence between memory retrieval for lists of words and statistical patterns of occurrence of words in large databases of text. Similarly, other computational models for memory (Shiffrin & Steyvers, 1997), association (Griffiths, Steyvers & Tenenbaum, 2007), reasoning (Oaksford & Chater, 1994), prediction (Griffiths & Tenenbaum, 2006) and causal induction (Anderson, 1990; Griffiths & Tenenbaum, 2005; Steyvers, Tenenbaum, Wagenmakers, & Blum, 2003) have shown how our cognitive system is remarkably well adapted to our environment. Anderson’s (1990) analysis of memory also showed for the first time that there are fundamental connections between research on memory and information retrieval systems. Because information retrieval systems and human memory often address similar computational problems, insights gained from information retrieval systems can be helpful in understanding human memory. For example, one component of Anderson’s first rational memory model involved calculating the predictive probability that items will 1re-occur given their historical pattern of occurrences. The solution to this problem was based on information retrieval models developed for library and file systems (Burrell, 1980; Salton & McGill, 1983). Just as it is useful to know the probability that a book will be needed in order to make it available in short-term or off-site storage, it is useful to know whether a fact is likely to be needed in the future when storing it in memory. Modern information retrieval research provides new tools for modeling the environment in which human memory operates, and new systems to which human memory can be compared. An important innovation has been the introduction of statistical language models to capture the statistics of the regularities that occur in natural language (e.g. Croft & Lafferty, 2003; Ponte & Croft, 1998). The goal of language modeling is to exploit these regularities in developing effective systems to assess the relevance of documents to queries. Probabilistic topic models (e.g. Blei, Ng, Jordan, 2003; Griffiths & Steyvers, 2004; Griffiths, Steyvers, & Tenenbaum, 2007; Hoffman, 1999; Steyvers & Griffiths, 2006; Steyvers, Griffiths, & Dennis, 2006) are a class of statistical language models that automatically infer a set of topics from a large collection of documents. These models allow each document to be expressed as a mixture of topics, approximating the semantic themes present in those documents. Such topic models can improve information retrieval by matching queries to documents at a semantic level (Blei, Ng & Jordan, 2003; Hoffman, 1999; Chemudugunta, Smyth & Steyvers, 2007). Another important problem in information retrieval is dealing with the enormous volume of data available on the world wide web. For any query, there might be a very large number of relevant web pages and the task of modern search engines is to design effective algorithms for ranking the importance of webpages. A major innovation has been the PageRank algorithm, which is part of the Google search engine (Brin & Page, 1998). This algorithm ranks web pages by computing their relative importance from the links between pages. In this chapter, we use these innovations in information retrieval as a way to explore the connections between research on human memory and information retrieval systems. We show how PageRank can be used to predict performance in a fluency task, where participants name the first word that comes to mind in response to a letter cue. We also give an example of how cognitive research can help information retrieval research by formalizing theories of knowledge and memory organization that have been proposed by cognitive psychologists. We show how a memory model that distinguishes between the representation of gist and verbatim information can not only explain some findings in the memory literature but also helps in formulating new language models to support accurate information retrieval. 2. A Probabilistic Approach to Information Retrieval Search engines and human memory are both solutions to challenging retrieval


Rational Analysis as a Link between Human Memory and Information Retrieval

Download Rational Analysis as a Link between Human Memory and Information Retrieval
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Rational Analysis as a Link between Human Memory and Information Retrieval and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Rational Analysis as a Link between Human Memory and Information Retrieval 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?