Wednesday 10 August 2011

Cracks in Natural Language Processing Watson, Unstructured Data

Craig Rhinehart Enterprise Content Management, IBM strategy officer Chief. If your blog covers a variety of topics of analysis information management of the ECM.

He first started business 22 years 19 years, bought his first company, and seven participated in the purchase. He was vice president of product marketing at FileNet 2006, IBM bought.

Recently, Watson, the IBM computer that beat the competition the candidates have been human occupation in the United States earlier this year.

Hazard was launched in 1964. The answers given in the form of a question is its own form of ownership. Therefore, "the author of the United States negotiated the Louisiana Purchase," you must answer: "Thomas Jefferson,"

On a trip to London a new SearchDataManagement.co.UK Rhinehart said. Below is a modified version of the word.

What is the importance of Watson to win the profession? I have a program here in the UK.

Craig Rhinehart: Watson, I do not think we, a computer game. Think of it as a breakthrough in this process.

There is good information and communication in terms of natural language of unstructured information. This technology offers new solutions, offer new ways to interact with the computer.

Natural language is still one of the five last five years, however, very uncertain. On the contrary, the word, for example, depends on the context - an army, great music, a term, an adjective "great" sense and so on. Natural language, code, abbreviations and references mentioned in pop culture.

Things really easy for hard numbers team are unclear. But information is not structured and 80% of 44 times increase over the next 10 years is welcome.

Knowledge management (KM) on the ground in a dozen years, similar things is said about the importance of the discovery of unstructured content, and tacit knowledge. IBM has participated in many KM. How to 1998, compared to, say, 2011 to?

Rhinehart: Approaches [unstructured information] Heritage failed. The research is a good example. The correct answer to a question from the most popular point that can not be manifested by a search engine. Keywords for search engines to break the question are to you as a user. So and so, the popularity, the influence of advertising, respectively, get pages and pages of results. Then spend more time reading through the answers. This is not a good research experience, and today's decision, using it for support.

Watson, [place] of the natural language interface based on a machine. This could ask questions in natural language. Your confidence score based on the knowledge base of trust, and return to the search results. TV has given the first option, but in real life, work, more than one option, and I want to trust the critics.

This is a new paradigm. Keyword matching is not the way to go at any time. Losing the context of this question. Keywords are not issues in natural language.

Yes, how we have always said, seminars, and podcasts, American English, based representation, but of course a problem for both teams, "natural language" is. - No - I called an official language and cultural-linguistic? Important that you do not know how?

Rhinehart: Watson, an English-language-based technology, the United States and others. Game Show has references to all versions of the English occupation. Watson admits that dialects and slang. IBM, other technology used in call centers, fight against crime, and a lot of different languages ​​are multi-language support. There is a limitation.

In this social context and see the other animals to be exploited. Why?

Rhinehart, a manager for a time dictated a letter to the secretary. Playing word processing. This document. Now that media type is not fast enough. I used to do all the work in Word and Excel. These days, the means of communication for the public to contact me in a much shorter and more comfortable with the basic text - blogs, Twitter.

And the media?

Rhinehart: His work in the basement as long as a few million dollars Yes, I would not have made a video a few years ago. Now you can with a laptop.

Content analysis shows the limits of traditional business intelligence? If so, how?

Rhinehart: You will avoid the full value of BI? No. 1 in the time it takes to deploy the BI projects. No. 2 BI is not only related to structural data. Now the [enterprise content management] ECM girl, I care more structured information. IBM may have a different perspective to one of the leaders of the BI. But we did with the analysis of the value of content produced to shorten the time to solve the first problem. In addition, the data storage such as Netezza, Cognos, or structural data that we have built an integrated environment.

Game of perfect information is a legend but here there is absolutely inaccessible? From a business pure and hard to see, feel and strategy where there is enough information and not when the intestine

Rhinehart: The right to an inability to make a decision very frustrating to find information. Watson decision, the task of finding all this should help the decision maker can decide what time to provide accurate and relevant. In the future, the objective for the management of all information relevant and useful to your business will try to manage. These organizations tend to today is not the case. All information is equal.

No comments:

Post a Comment