Information Inference Theory and Practice
This event took place on Tuesday 10 May 2005 at 14:00
Dr. Dawei Song KMi, The Open University, UK
In this talk, I will present a information inference framework we developed for mimicking human text based reasoning. The notion of "information inference" refers to the derivation of context sensitive implicit information carried by often short text fragments. Many of us transact information inference as part of our daily information processing tasks, e.g., when scanning subject headings of emails, or captions retrieved from search engines. Due to the information explosion, it is becoming increasingly difficult for humans to keep pace. Thus, the objective of our work is to investigate how to automatically conduct information inference in a way, which is compatible with human reasoning. In short, we aim to construct a computational reasoning system which can draw context-sensitive associations in relation to text which correlates with those we would make, but increasingly can't.
Instead of using traditional symbolic reasoning, we take a psychologistic stance. By drawing on theories from non-classical logic, information retrieval and applied cognition, we have proposed an information inference mechanism via computation of information flow through a high dimensional semantic space, which is automatically constructed from a text corpus. A concept combination heuristic has been developed for priming vector representations according to context. Information flow relationships between concepts or concept combinations are established by computing the degrees of inclusions between their underlying vector representations in the semantic space. Our framework is evaluated, using standard TREC text corpora, topics and the corresponding human relevance judgments, by measuring the effectiveness of query models derived by information flow computations. Experimental results have shown that information inference largely outperforms the co-occurrence and semantic similarity based query expansion.
From a broader point of view, information inference can be applied to many scientific and social knowledge discovery and management problems. In the end of this talk, I will discuss about a global motivation of our work and a list of on-going initiatives derived from our information inference research agenda.
This event took place on Tuesday 10 May 2005 at 14:00
In this talk, I will present a information inference framework we developed for mimicking human text based reasoning. The notion of "information inference" refers to the derivation of context sensitive implicit information carried by often short text fragments. Many of us transact information inference as part of our daily information processing tasks, e.g., when scanning subject headings of emails, or captions retrieved from search engines. Due to the information explosion, it is becoming increasingly difficult for humans to keep pace. Thus, the objective of our work is to investigate how to automatically conduct information inference in a way, which is compatible with human reasoning. In short, we aim to construct a computational reasoning system which can draw context-sensitive associations in relation to text which correlates with those we would make, but increasingly can't.
Instead of using traditional symbolic reasoning, we take a psychologistic stance. By drawing on theories from non-classical logic, information retrieval and applied cognition, we have proposed an information inference mechanism via computation of information flow through a high dimensional semantic space, which is automatically constructed from a text corpus. A concept combination heuristic has been developed for priming vector representations according to context. Information flow relationships between concepts or concept combinations are established by computing the degrees of inclusions between their underlying vector representations in the semantic space. Our framework is evaluated, using standard TREC text corpora, topics and the corresponding human relevance judgments, by measuring the effectiveness of query models derived by information flow computations. Experimental results have shown that information inference largely outperforms the co-occurrence and semantic similarity based query expansion.
From a broader point of view, information inference can be applied to many scientific and social knowledge discovery and management problems. In the end of this talk, I will discuss about a global motivation of our work and a list of on-going initiatives derived from our information inference research agenda.
Future Internet
KnowledgeManagementMultimedia &
Information SystemsNarrative
HypermediaNew Media SystemsSemantic Web &
Knowledge ServicesSocial Software
Future Internet is...
To succeed the Future Internet will need to address a number of cross-cutting challenges including:
- Scalability in the face of peer-to-peer traffic, decentralisation, and increased openness
- Trust when government, medical, financial, personal data are increasingly trusted to the cloud, and middleware will increasingly use dynamic service selection
- Interoperability of semantic data and metadata, and of services which will be dynamically orchestrated
- Pervasive usability for users of mobile devices, different languages, cultures and physical abilities
- Mobility for users who expect a seamless experience across spaces, devices, and velocities
Future Internet from KMi.
List all Future Internet Projects
Check out these Hot Future Internet Technologies:
List all Future Internet Technologies
