Morphological disambiguation

A method for searching a corpus of text made up of words, comprising:

• morphologically analyzing the words in the corpus to generate, for each of at least some of the words, a list of candidate analyses, each candidate analysis comprising a respective lemma and a linguistic pattern relating the lemma to the analyzed word, the linguistic pattern comprising a specification of at least one characteristic of the word, selected from a set of characteristics including a part of speech, prefix, number, gender and person of the word;
• evaluating the pattern in each of the analyses so as to determine a relative frequency of occurrence of the pattern in each of the analyses, independent of the lemma to which the pattern is applied;
• selecting from the list for each of the analyzed words one or more of the analyses that comprise respective patterns whose frequency of occurrence is above a predetermined threshold;
• entering the lemmas of the selected analyses in an index of the corpus;
• and
• applying a search query to the index.

A computer software product, comprising a computer-readable medium in which program instructions are stored, which instructions, when read by a computer, cause the computer to build a statistical base by morphologically analyzing a corpus of text comprising multiple words having lemmas so as to find respective linguistic patterns of the words independently of the lemmas to which the linguistic patterns are applied, each pattern comprising a specification of at least one characteristic selected from a set of characteristics including a part of speech, prefix, number, gender and person, and finding relative frequencies of occurrence of the linguistic patterns in the corpus,

• wherein the instructions further cause the computer to morphologically analyze an input string to generate a list of candidate analyses of the string, each candidate analysis comprising a respective word, having a lemma, and a linguistic pattern of the word, the linguistic pattern comprising a specification of at least one characteristic of the word, selected from a set of characteristics including a part of speech, prefix, number, gender and person of the word, and to evaluate the pattern in each of the analyses so as to determine a relative frequency of occurrence of the pattern in each of the analyses, independent of the lemma to which the pattern is applied, and to select from the list one or more of the analyses that comprise respective patterns whose frequency of occurrence is above a predetermined threshold.

Apparatus for morphological disambiguation, comprising a linguistic processor, which is adapted to build a statistical base by morphologically analyzing a corpus of text comprising multiple words having lemmas so as to find respective linguistic patterns of the words independently of the lemmas to which the linguistic patterns are applied, each pattern comprising a specification of at least one characteristic selected from a set of characteristics including a part of speech, prefix, number, gender and person, and finding relative frequencies of occurrence of the linguistic patterns in the corpus,

• wherein the linguistic processor is further adapted to receive an input string, to morphologically analyze the string to generate a list of candidate analyses of the string, each candidate analysis comprising a respective word, having a lemma, and a linguistic pattern of the word, the linguistic pattern comprising a specification of at least one characteristic of the word, selected from a set of characteristics including a part of speech, prefix, number, gender and person of the word, and to evaluate the pattern in each of the analyses so as to determine a relative frequency of occurrence of the pattern in each of the analyses, independent of the lemma to which the pattern is applied, and to select from the list one or more of the analyses that comprise respective patterns whose frequency of occurrence is above a predetermined threshold.