The InnovationQ application will be updated on Sunday, May 31st from 10am-noon ET. You may experience brief service interruptions during that time.
Browse Prior Art Database

Word-chain Game System

IP.com Disclosure Number: IPCOM000014041D
Original Publication Date: 2001-Dec-29
Included in the Prior Art Database: 2003-Jun-19

Publishing Venue



Disclosed is a system for word-chain games between computer and its users. This system also can be use for collect personal lexicon set of game system users that can be useful for system personalization. This system includes five parts lexical sets, calculation of score for selecting the next reply, collection of user input, game rules, and user interface. Game rules and calculation method can be changed according to game types, so that the system can be applied not only to the question-word-end character to answer-word-head character chaining but also to the any word chaining for instance, "middle" character chaining in the three-character-word, word-head to word-end chaining, and so on. The most accurate method to select next reply in order to win just as planned is to make a complete graph according to the game rules. By referring to the graph, the system can choose the most appropriate lexical entry for the next reply. However, this method takes to much times to make the complete graph and to calculate probabilities of each path in the graph. If calculation time must be saved, one of the other easier selection methods can be adopted. For instance, in Japanese, there are fewer number of words that begin with "ra" "ri" "ru" "re" "ro" than ones that begin with other characters. Therefore in the word-end character to word-head character chain game, the system can get the high probability for winning the game by choosing one of the words that begin with "r-". Any other easier ways can be taken in the calculation part. Lexical entries used in word-chain games are actually very restricted. For instance, concrete and easy words that many three-year-old children understand, such as "apple", "car", "gorilla", are usually used in a word-end to word-head character chain game. However, if a additional rule, such as "to use food name", is added, the property of lexical set used in the game should be different form the "base" set. In addition, there is some difference in what words are used according to the users. The lexical collection part of described system is effective for reflecting such difference. In the collection part, the system collects words not in the current lexicon set and stores them in the additional lexicon pool. By repeating this kind of registration, the system can re-build appropriate game-rule dependent lexicon set and user-dependent lexicon set. This lexicon pool can be the basic data of any other personalizing tuning of natural language processing systems. 1