Breaking lower the barriers of human-computer communication

by Iowa Condition College

artificial intelligence
Credit: CC0 Public Domain

A lot of us regularly ask our smartphones for directions in order to be a musician without giving much considered to we’ve got the technology which makes it all possible – we simply desire a quick, accurate reaction to our voice instructions.

With increased companies using artificial intelligence for a number of applications and also to build relationships consumers, the is scheming to make individuals interactions more human-like, stated Zhu “Came” Zhang, affiliate professor of knowledge systems and Kingland Faculty Fellow running a business Analytics in Iowa Condition University’s Ivy College of economic. Zhang is adding to that particular effort by improving how machines, for example smartphones and computers, understand and generate language.

“Computers weren’t built to handle ambiguity of human language,” Zhang stated. “We’ve subtle methods for saying things having a similar meaning, using different words and linguistic structures and that’s hard for computational models to know and mimic.”

Zhang states running a business scenarios, for instance, consumers can express similar opinions with completely different linguistic forms:

  • First statement: Nothing relating to this camera is exciting.
  • Second statement: I do not such as this product.

Helping machines identify paraphrasing is among the finest challenges in the area of natural language processing. Zhang states trying to code the overwhelming amount of linguistic rules and connected exceptions could be dauntingly impractical. Rather, researchers and industry leaders are utilizing machine learning methods, which utilize considerable amounts of information to educate computers to acknowledge and understand language patterns in tangible use.

New model well accepted

Zhang and Amulya Gupta, an ISU graduate student in information technology, created a new computational model according to deep understanding how to improve precision, precision and recall of discovering statements similar in meaning, but different in wording, structure and length. They tested the model using 50,000 sample sentence pairs, like the above example, and found it 80 to 85 % accurate.

That’s encouraging, but Zhang states it just addresses “a little slice of the bigger problem” in the area of artificial intelligence. He and Gupta lately presented a paper on their own work on the Annual Meeting from the Association for Computational Linguistics around australia. Zhang states it is among the top conferences for industry players, including Google, Microsoft and Amazon . com, in addition to leading academics focusing on this problem.

“Our research was perfectly received,” Zhang stated. “We spoken with individuals at Microsoft and JD, the biggest e-commerce company in China, plus they saw potential within our work.”

Making machines feel human

Zhang states the aim would be to continue increasing the model to really make it more relevant to a number of business scenarios. Customer support bots make the perfect example. Rather of pressing buttons from the menu of options when calling to refill a prescription, most consumers still prefer speaking to someone. Zhang states it isn’t just important the bot seem human, however it should have the versatility to adjust to individual callers.

“This sort of capacity adds an individual touch,” Zhang stated. “The opportunity to manipulate languages is the reason why the bot or any other machines feel human.AI recreates chemistry’s periodic table of elements

Citation: Breaking lower the barriers of human-computer communication (2018, August 16) retrieved 16 November 2020 from
This document is susceptible to copyright. Aside from any fair dealing with regards to private study or research, no part might be reproduced with no written permission.


Breaking Down Communication Barriers | Angela Graves | TEDxEvansville