Cross-lingual retrieval system and method that utilizes stored pair data in a vector space model to process queries

A storage medium readable by a computer, storing a program of instructions executable by the computer to perform cross-lingual retrieval, the instructions comprising the steps of:

• receiving a query written in a first language;
• firstly retrieving first language sentences similar to the query from a set of first language sentences in a plurality of pair data each having a first language sentence and a second language sentence having the same meaning stored in a memory area;
• and
• secondly retrieving second language sentences similar to second language sentences paired with the respective first language sentences retrieved by the first retrieval from among the set of the second language sentences in the plurality of pair data;
• wherein the second retrieval of second language sentences comprises the steps of:
• determining and extracting important words from each of second language sentences having the same meaning and paired with the respective first language sentences retrieved by the first retrieval of first language sentences, provides a degree of importance to each of the important words and retrieves similar second language sentences using the important words and their degree of importance, wherein, for the set A of the second language sentences stored in the pair data storing means, a set B of the second language sentences having the same meaning and paired with the respective first language sentences retrieved by the first retrieval of first language sentences and a set C of all words appearing in the set B, a first value that is the number of sentences included in the set B, a second value that is the number of sentences in the set B for each important word candidate containing the important word candidate, supposing that all the words in the set C are the important word candidates, and a third value that is the number of sentences in the set A for each important word candidate containing the important word candidate are computed, and the degree of importance of each important word candidate is obtained using the first, second and third values as variables to determine the important word from among the important word candidates based on the degree of importance;
• and
• retrieval of second language sentences takes one of the following values as the degree of importance of each important word candidate:
• extended mutual information=log2{(Mβ)/(αγ)};
• extended t-score=M{(Mβ−αγ)/(αγ)};
• and
• extended Dice-coefficient=2β(α+γ),
• wherein the number of sentences included in the set A is referred to as M, the first value is referred to as α, the second value for each important word candidate is referred to as β and the third value for each important word candidate is referred to as γ.

A method of cross-lingual retrieval, comprising the steps of:

• receiving a query written in a first language;
• firstly retrieving first language sentences similar to the query from among a set of first language sentences in a plurality of pair data each having a first language sentence and a second language sentence having the same meaning;
• and
• secondly retrieving second language sentences similar to second language sentences paired with the respective first language sentences retrieved by the first retrieving from among a set of second language sentences in the plurality of pair data;
• wherein the second retrieval of second language sentences comprises the steps of:
• determining and extracting important words from each of second language sentences having the same meaning and paired with the respective first language sentences retrieved by the first retrieval of first language sentences, provides a degree of importance to each of the important words and retrieves similar second language sentences using the important words and their degree of importance, wherein, for the set A of the other second language sentences stored in the pair data storing means, a set B of the second language sentences having the same meaning and a set C of all words appearing in the set B, a first value that is the number of sentences included in the set B, a second value that is the number of sentences in the set B for each important word candidate containing the important word candidate, supposing that all the words in the set C are the important word candidates, and a third value that is the number of sentences in the set A for each important word candidate containing the important word candidate are computed, and the degree of importance of each important word candidate is obtained using the first, second and third values as variables to determine the important word from among the important word candidates based on the degree of importance;
• and
• retrieval of second language sentences takes one of the following values as the degree of importance of each important word candidate:
• extended mutual information=log2{(Mβ)/(αγ)};
• extended t-score=M{(Mβ−αγ)/(αγ)};
• and
• extended Dice-coefficient=2β/(α+γ),
• wherein the number of sentences included in the set A is referred to as M, the first value is referred to as α, the second value for each important word candidate is referred to as β and the third value for each important word candidate is referred to as γ.

A cross-lingual retrieval system comprising:

• pair data storing means for storing a plurality of pair data each having a first language sentence and a second language sentence having the same meaning;
• sentence storing means for storing another set of second language sentences;
• query receiving means for receiving a query written in the first language;
• first retrieving means for retrieving first language sentences similar to the query from among a set of the first language sentences stored in the pair data storing means;
• and
• second retrieving means for retrieving second language sentences similar to second language sentences paired with the respective first language sentences retrieved by the first retrieving means from among the other set of the second language sentences stored in the sentence storing means;
• wherein the second retrieving of second language sentences comprises the steps of:
• determining and extracting important words from each of second language sentences having the same meaning and paired with the respective first language sentences retrieved by the first retrieval of first language sentences, provides a degree of importance to each of the important words and retrieves similar second language sentences using the important words and their degree of importance, wherein, for the set A of the other second language sentences stored in the pair data storing means, a set B of the other second language sentences having the same meaning and a set C of all words appearing in the set B, a first value that is the number of sentences included in the set B, a second value that is the number of sentences in the set B for each important word candidate containing the important word candidate, supposing that all the words in the set C are the important word candidates, and a third value that is the number of sentences in the set A for each important word candidate containing the important word candidate are computed, and the degree of importance of each important word candidate is obtained using the first, second and third values as variables to determine the important word from among the important word candidates based on the degree of importance;
• and
• retrieval of second language sentences takes one of the following values as the degree of importance of each important word candidate:
• extended mutual information=log2{(Mβ)/(αγ)};
• extended t-score=M {(Mβ−αγ)/(αγ)};
• and
• extended Dice-coefficient=2β/(α+γ),
• wherein the number of sentences included in the set A is referred to as M, the first value is referred to as α, the second value for each important word candidate is referred to as β and the third value for each important word candidate is referred to as γ.

A storage medium readable by a computer, storing a program of instructions executable by the computer to perform cross-lingual retrieval, the instructions comprising the steps of:

• receiving a query written in a first language;
• firstly retrieving first language sentences similar to the query from a set of first language sentences in a plurality of pair data each having a first language sentence and a second language sentence having the same meaning stored in a first memory area;
• and
• secondly retrieving second language sentences similar to second language sentences paired with the respective first language sentences retrieved by the first retrieval from among another set of second language sentences stored in a second memory area separately to the plurality of pair data;
• wherein the second retrieval of second language sentences comprises the steps of:
• determining and extracting important words from each of second language sentences having the same meaning and paired with the respective first language sentences retrieved by the first retrieval of first language sentences, provides a degree of importance to each of the important words and retrieves similar second language sentences using the important words and their degree of importance, wherein, for the set A of the other second language sentences stored in the pair data storing means, a set B of the other second language sentences having the same meaning and a set C of all words appearing in the set B, a first value that is the number of sentences included in the set B, a second value that is the number of sentences in the set B for each important word candidate containing the important word candidate, supposing that all the words in the set C are the important word candidates, and a third value that is the number of sentences in the set A for each important word candidate containing the important word candidate are computed, and the degree of importance of each important word candidate is obtained using the first, second and third values as variables to determine the important word from among the important word candidates based on the degree of importance;
• and
• retrieval of second language sentences takes one of the following values as the degree of importance of each important word candidate:
• extended mutual information=log2{(Mβ)/(αγ)};
• extended t-score=M{(Mβ−αγ)/(αγ)};
• and
• extended Dice-coefficient=2β/(α+γ),
• wherein the number of sentences included in the set A is referred to as M, the first value is referred to as α, the second value for each important word candidate is referred to as β and the third value for each important word candidate is referred to as γ.

A method of cross-lingual retrieval, comprising the steps of:

• receiving a query written in a first language;
• firstly retrieving first language sentences similar to the query from among a set of first language sentences in a plurality of pair data each having a first language sentence and a second language sentence having the same meaning;
• and
• secondly retrieving second language sentences similar to second language sentences paired with the respective first language sentences retrieved by the firstly retrieving from among another set of second language sentences provided separately to the plurality of pair data;
• wherein the second retrieval of second language sentences comprises the steps of:
• determining and extracting important words from each of second language sentences having the same meaning and paired with the respective first language sentences retrieved by the first retrieval of first language sentences, provides a degree of importance to each of the important words and retrieves similar second language sentences using the important words and their degree of importance, wherein, for the set A of the other second language sentences stored in the pair data storing means, a set B of the second language sentences having the same meaning and a set C of all words appearing in the set B, a first value that is the number of sentences included in the set B, a second value that is the number of sentences in the set B for each important word candidate containing the important word candidate, supposing that all the words in the set C are the important word candidates, and a third value that is the number of sentences in the set A for each important word candidate containing the important word candidate are computed, and the degree of importance of each important word candidate is obtained using the first, second and third values as variables to determine the important word from among the important word candidates based on the degree of importance;
• and
• retrieval of second language sentences takes one of the following values as the degree of importance of each important word candidate:
• extended mutual information=log2{(Mβ)/(αγ)};
• extended t-score=M{(Mβ−αγ)/(αγ)};
• and
• extended Dice-coefficient=2β/(α+γ),
• wherein the number of sentences included in the set A is referred to as M, the first value is referred to as α, the second value for each important word candidate is referred to as β and the third value for each important word candidate is referred to as γ.