Introduction to "This is Watson"
|Introduction to "This is Watson"|
|Authors:||David A. Ferrucci|
|Citation:||IBM Journal of Research and Development 56 (3.4): missing pages. 2012 May-June|
|Database(s):||Google Scholar cites|
|Web:||Bing Google Yahoo! — Google PDF|
|Article:||BASE Google Scholar PubMed|
|Restricted:||DTU Digital Library|
The paper argues that such systems require advances in areas of
- information retrieval
- knowledge representation and reasoning
- machine learning
- human-computer interfaces
The Watson system (AdaptWatson) consists of "more than 100 core algorithmic components."
Some of the components are:
- Question analysis to determine lexical answer type
- Grammatical parser of the question.
- Collection of content, e.g., from encyclopedias.
- Building a resource (PRISMATIC) with knowledge from collected content.
Initial performance was 16% precision@70 (i.e., with 70% questions answered). It rose to 85% Precision@70.
|Task||State of the art|
|Entity disambiguation||Robust disambiguation of named entities in text|
|Relation detection||A composite kernel to extract relations between entities with both flat and structured features|
|Textual entailment||PKUTM participation TAC 2010 RTE and summarization track|
 Related papers
- A framework for merging and ranking of answers in DeepQA
- Automatic knowledge extraction from documents
- Building Watson: an overview of the DeepQA project
- Deep parsing in Watson
- In the game: the interface between Watson and Jeopardy!
- Natural language processing with Prolog in the IBM Watson system
- Question analysis: how Watson reads a clue
- Relation extraction and scoring in DeepQA
- Textual resource acquisition and engineering
- Towards the open advancement of question answering systems