System of effectively searching text for keyword, and method thereof

A system for searching keywords from texts, the system comprising:

• a first index storage unit storing a first index which designates a list of keywords contained in the texts from identification information of the respective texts, wherein the first index is generated in advance based on a plurality of texts stored in a text database, and the first index is stored in the first index storage unit prior to a search of the keywords;
• the storage unit further storing hash files of categories of keywords in the first index storage unit, with the hash files indicating pointers to lists of keywords contained in the texts;
• a second index storage unit for storing a second index which designates a list of texts containing the keywords from identification information of the respective keywords, in a way that the second index corresponds to a number of texts containing the respective keywords, wherein the second index is generated in advance based on the plurality of texts stored in the text database;
• an input unit
• receiving an input of a text search condition which is a condition of keywords contained in the texts;
• the input unit further receiving at least one of:
• fields in which keywords to be search targets are used, or information showing word classes of the keywords;
• a determination unit
• reading the number of texts containing the respective keywords from the second index storage unit;
• the determination unit further calculating an estimation of search time by the first index and an estimation of search time by the second index, based on the number of texts satisfying the text search condition among a plurality of texts to be searched, and based on the number of the texts stored corresponding to the second index;
• and
• determining which one of the first and the second indexes makes a search faster;
• a search unit for searching for and outputting keywords which appear in texts satisfying the text search condition with higher frequency than a predetermined standard, by using the index which is determined to make the search faster and the search unit outputs the keywords to a user;
• a ratio calculation unit for calculating a condition fulfillment rate indicating an occupancy ratio of texts which satisfy the text search condition among all texts to be searched;
• a function calculation unit for approximately calculating, based on the condition fulfillment rate, probability distributions of probability variables indicating the number of texts which contain the respective keywords and which satisfy the text search condition, and for calculating, based on a calculated probability distribution, a function for a threshold for obtaining the sum of probabilities that the probability variables of the respective keywords are not less than the threshold, wherein the function for calculating the threshold is:
• E(Y)=E(Σi=1, 2, . . . , IYi)=Σi=1, 2, . . . , IE(Yi)
• =Σi=1, 2, . . . I P (Yi>M) wherein M is the threshold, i represents a keyword, and Yi represents a random variable;
• a threshold calculation unit for calculating the threshold at which the calculated function is substantially the same as the standard number;
• a second search time calculation unit for calculating the number of texts containing the respective keywords contained in texts whose number is greater than the calculated threshold, based on the number of the texts stored in the second index storage unit, and for calculating a product of the calculated number of texts and predetermined access time, as an estimation of search time by the second index;
• and
• a first search time calculation unit for calculating a product of the number of texts which satisfy the text search condition among a plurality of the texts to be searched and the predetermined access time, as an estimation of search time by the first index, wherein
• the determination unit determines that a search by the second index is faster, conditional upon the estimation of search time calculated by the second search time calculation unit is smaller than the estimation of search time calculated by the first search time calculation unit.