It is very important to foretell a future ``vogue word" or ``vogue concept" for advertisement, marketing research, CF (Commercial Film) creation, and drama creation activities. The concept making or new word creation is a creative and hard task for creators.
A creator designs many candidate future vogue words or concepts at the beginning step. Next, the creator selects a best candidate word / concept. This selection is very hard, because the selection process is essentially sensual and intuitive. Thus, the creator cannot explain why this candidate word is the best one for the CF in a competition room at the client's office. Often, the selected best one is not accepted by the client and new one is proposed by the client. This client's decision making process is sensual and intuitive, too. Thus, the both of the client and the creator have no confidence in their own decision making process.
To resolve this problem, this paper demonstrates a new approach using natural language processing technique and ``the contemporary word dictionary" published by Jiyukokuminsha in Tokyo to select the best candidate word or concept. This system is based on an assumption that vogue is a reflection of the economical or political situation. This assumption is well-known in advertisement research fields. The proposed system is an application of this assumption.
To use this proposed system, first, the creator make plural candidate vogue words or concepts. Secondly, short description is added into each candidate word/ concept. The each candidate word / concept is an entry of dictionary. Also, added explanation is a description for the entry word in a dictionary. Thirdly, a dictionary ``the contemporary word dictionary (TCWD in the hereafter)" is prepared. The TCWV has an ``upper concept" for the each entry word. There are ten upper concepts: ``politics," ``economy," ``health," ``vogue," and ``science" for instance.
Next, distances between the upper concepts of TCWD and the candidate words / concepts are calculated by ``sense vector method" in natural language processing research field. If a candidate word has small distance for the ``economy" upper concept, for example, the word will be in vogue from the economical viewpoints.
Of course, the vogue changes every year. Thus, ``vogue words" of the every year should be evaluated from the chronological tendency. To calculate chronological tendency, we employed ``Great Prize for Words in Vogue" selected by Jiyukokuminsha. The Great Prize words are regarded as a candidate word / concept for each year. The distances between the Great Prize words and the entry words of TCWD are calculated and the tendency of distance in each upper concept is examined.
Using this system, the Jiyukokuminsha's Great Prize words in the 2000 year are forecasted from 1997-1999 year Great Prize words. The experimental results are also evaluated statistically. The statistical ``test" shows that the distance calculation is valid in some domains, ``economy," and ``politics," for instance. This means that proposed approach is practical and effective in some domains.
On the other hand, the calculated distance is not statistically supported in the other domains: ``vogue," and ``health," etc. A major reason of this limitation is the ``sparseness" of the sense vector space. Thus, use of conventional word dictionaries is required as a further research.