Watson is a computer system that was designed and built by IBM in order to compete on Jeopardy, an American TV quiz show, against human contestants at an expert level live on the show. Watson was named after IBM founder Thomas J. Watson. IBM Research have set out objectives to develop computer science with regard to pushing the boundaries of computer technology so as to advance the areas such as business, society and of course science. In order to achieve, this IBM Research took 3 years and a team of 20 researchers headed up by Dr. David Ferrucci to deeply investigate and develop the project, to have Watson accomplish its task to compete on the show with precision, confidence and speed.
There is a vast amount of information generated …show more content…
There are 6 different categories with 5 clues in each category with broad topics such as science, history, to part clues, such as who am I, to anything goes. The research team in their approach repeatedly tried as many different hypotheses containing a variety of different situations in order to see which generated the most optimal answers that gave a wide spectrum of roughly teamed scoring algorithms. Influencing category information is alternative clear area requiring this approach.
The team attempted to characterise the clues in a variety of ways, some of which consisted of characterising by grammatical constructions, topics, answer type, difficulty, so on and so forth. One of the classifications that the team found to be beneficial was built on the principal technique employed to solve the clue. The majority of the game’s clues are based on factual information about one or more individual elements. With regard to the questions themselves the challenge the team faced was trying to figure out what precisely was being asked and what parts of the clue were valid in formulating the …show more content…
At the beginning of the project they spent a lot of time researching existing baselines and trying to modify them, however, this failed to a substantial bearing on the experiment. As a result the team basically revamped their whole approach and also in collusion with CMU, embarked upon the Open Advancement of Question Answering (OAQA) initiative. DeepQA is the system that the research team built and continue to develop. It is a massively parallel probabilistic evidence-based architecture and they used it for the Jeopardy experiment, where they used over 100 different techniques to analyse natural language, identify sources, find and generate hypotheses, find and score evidence, and merge and rank hypotheses. The most important aspect of the techniques used is that they were combined in DeepQA so that by overlapping them to combined their strengths to bear and augment the enhancements in accuracy, confidence, or speed. They developed DeepQA in such a way that it could be reused on other projects as well as Jeopardy. They have already adapted it to different business applications and additional exploratory experiment problems such as enterprise search, gaming, and