Sherlock Holmes technology used to solve in-game chat problem
“Chatbot” technology is being used in an attempt to solve one of “the last uncracked problems” in games design.
221b, released in the run-up to the new Sherlock Holmes movie, harnesses the software to allow conversations between players and in-game characters.
Gamers, who assume the character of either Sherlock Holmes or Dr Watson, must interrogate virtual witnesses and suspects to progress in the game.
Success depends upon getting the right answers from these characters.
“It’s our role to predict what you might know at that point in the game and the questions you might ask,” said Rollo Carpenter of Existor, which provided the technology.
“The ways that you might say things to them are almost unlimited.”
Mr Carpenter is a two-time winner of the Loebner Prize, a competition that challenges computer scientists to build programmes capable of convincingly human conversations.
When a player interrogates a game character in 221b, Carpenter’s technology is used to analyse the question and to provide a relevant response.
Rather than attempting to create an exhaustive list of possible questions and the appropriate response, the characters in the game are capable of making a “fuzzy interpretation” of what is said to them.
Pattern matching is then used to identify the appropriate answer for any given input by a player.
The intention is to remove the frustration, familiar to any who played the old text-based adventure games, of having to guess the right way of asking a question or giving an instruction.
While the technology allows the system to cope with the many different ways a player might attempt to elicit information from a suspect – the responses are limited.
It is not the first game to have explored innovative approaches to language.
Alex Champandard, a programmer who has worked on artificial intelligence for Rockstar Games and Guerrilla Games, believes one of the most creative approaches was used in the interactive story Facade.
Players must interact with two characters, Grace and Trip, whose relationship is experiencing difficulties.
“Since the AI characters are completely interactive, each time you play the outcome depends on your actions, said Mr Champandard.
“Yet in the background there’s a drama manager that makes sure the story keeps going.”
Dr Mike Reddy, who teaches games development and artificial intelligence at the University of Wales, highlights techniques used in the Nintendo DS puzzle game Scribblenauts.
“In this game, the player evokes objects and characters by typing or writing their name,” he explained.
Scribblenauts allows players to summon objects using words
For example, the player can write “helicopter” to summon the vehicle onscreen and use it to collect objects. Multiple objects can be “chained together” to solve problems.
“The clever semantic implementation is to know what would happen when a Dog meets a Lion,” said Dr Reddy.
“The game has 22,000 plus words and has attempted to implement all the possible interactions. Put Death up against God, for example, and you get an interesting surprise.”
For big budget games, creating hi-quality voice acting and animation “on the fly” is a significant technical challenge.
However, some games are attempting to move beyond rigidly scripted dialogue. Mr Champanard highlights the approach of the 2008 release Left4Dead
“Each of the characters has a set of voice samples which can trigger based on events, situations and other dialog lines”, he said.
“This results in completely emergent short conversations depending on the situation.”
Completely convincing language based interaction with non-player characters remains one of the “last uncracked problems”, according to Dr Reddy. But, he said, the games industry is making progress.
“We have come a long way from ‘All your base are belong to us’ and ‘TAKE AXE. THROW AXE AT DWARF'”