The hottest IBM innovation in natural language pro

  • Detail

IBM's innovation in natural language processing helps enterprises better understand the business language

according to the introduction of the technicians of zamori company: Waste pen refills can be made into a kind of recycled wood wood plastic composite material. At IBM, we focus on developing and expanding the function of enterprise natural language processing (NLP), aiming to help enterprises have in-depth insight, answer questions and make more wise decisions, Even if they have only a small data set or lack professional knowledge

although human language is easy for children to master, even the most advanced machines are extremely complex. The most challenging work of teaching AI to understand human intentions is that it requires a lot of data, time and professional knowledge

what exactly do you want to say when you ask questions? What do you want to achieve? What information do you really want? Human language is full of subtle differences, resulting in many ways to express specific intentions. For most AI, such as chat robots, this is indeed a problem. When encountering complex grammar, they will make mistakes, because they only focus on specific words and do not relate to a broader context

learn about the new enhancements of Watson assistant and see how our new natural language understanding model stands out like polyarylate par

in order to help enterprises cope with this challenge, IBM introduced a new improved natural language understanding (NLU) model in IBM Watson assistant for intention classification. In the benchmark test, compared with the commercial scheme, the new intention detection algorithm is more accurate. [1]

IBM Research Institute continues to improve natural language processing functions and integrate them into IBM Watson

in addition, we have also introduced new naturallanguageprocessing enhancements in IBM Watson assistant and Watson discovery, which are currently available in beta. These new functions were developed under the leadership of IBM research to improve the automation of artificial intelligence and the accuracy of natural language processing

the function of reading comprehension, which returns specific facts or short answers contained in longer paragraphs. At present, Watson discovery can determine the best paragraph corresponding to the query. Reading comprehension retrieve a large number of candidate paragraphs from the enterprise document set, search the answers to the current questions and return the corresponding answers. Reading comprehension applies the context situation understanding function to understand the query, uses a large number of language models to extract specific answers from the current document, and then the user will receive a confidence score, which represents the confidence of the system for each answer

this function is very suitable for organizations in the financial industry. For example, if you are about to make a loan decision, you may need to find accurate facts from complex documents, which are usually read and reviewed manually. Watsondiscovery will return the suggested paragraph before. With reading comprehension, users will get accurate answers (for example, what is the current loan interest rate? 2.9%), which saves them the time of manually searching a large number of documents. This feature is now available to some Watson discovery users in beta

faq extraction is currently available in beta. It is a novel answer retrieval technology. It crawls on the page to detect common questions and answers, and uses these contents to provide concise and up-to-date answers through Watson assistant

faq extraction is designed to work with Watson assistant's search skill to find answers to end-user questions in documents. This function makes it easier for end users to find the answers they need when interacting with the artificial intelligence virtual agent based in Wayne, Pennsylvania

for example, it may be difficult for enterprises to keep up with the changing public guidance, that is, the provisions that allow them to return to the workplace or reopen physical stores. If there is no such mechanism as FAQ extraction, it will require a lot of resources to keep the AI customer service solution up-to-date. On the contrary, watsonassistant only needs to know the URL of the official FAQ content to keep up with the latest information

explore new naturallanguageprocessing capabilities in Watson discovery and assistant

finally, Watson NLP solutions now support 10 other languages. Ibmwatson discovery now supports Bosnian, Croatian, Danish, Finnish, Hebrew, Hindi, Norwegian (bokm? L), Norwegian (Nynorsk), Serbian, and Swedish. Watson natural language understanding (NLU) now supports Danish, Norwegian (bokm? L), Norwegian (Nynorsk), Finnish, Czech, Hebrew, polish, and Slovak (key words)

these advances are based on the natural language processing innovation channel of IBM Research Institute. At the beginning of this year, we announced that we would adopt some core naturallanguageprocessing technologies that provide support for projectdebater of IBM Research Institute, including advanced emotion analysis (idiom understanding), summarization, topic clustering and key point analysis, and commercialized them in IBM naturallanguageprocessing products such as watsondiscovery

these innovations can help enterprises further understand and obtain real value from their business data, so that they can make smarter decisions and provide more efficient and in-depth insights for customers and employees

the statement about IBM's future direction and intention is only for the purpose of explaining its goals and objectives, and is subject to change or withdrawal without notice

[1] in November 2020, dialog AI software company jio haptik technology published a technical paper, comparing the performance of its products with similar products of Google, Microsoft and rasa. In addition to IBM Watson assistant, the performance of other commercial solutions is obtained from 2020 benchmark research conducted by Arora and other companies. IBM has run the same performance tests as those reported by Arora and other companies on the atsonassistant after ibmw completes the integration of national mandatory standards and the centralized review of recommended industry standards and plans for this analysis. The full results of IBM can be found in the following technical documents:

Copyright © 2011 JIN SHI