Information processing apparatus, information processing method, and storage medium

An information processing method comprising:

• acquiring a logical formula representing an inference result which is based on an inference rule for an observation event;
• generating one or more candidate natural language sentences from at least one text element with use of a language model generated on the basis of a corpus which does not contain a document used to generate the inference rule, the at least one text element being included in constituents of the logical formula and being text, wherein the one or more candidate natural language sentences is generated by the language model by calculating a continuous occurrence probability with which words contained in an extracted n-gram occur continuously, and the language model is generated based on a statistical result for a continuous occurrence probability;
• storing the language model in a memory;
• identifying, as a comparison target portion of the document used to generate the inference rule, a portion indicated by ground information associated with a rule applied to obtain the inference result;
• determining, with reference to a result of comparison between each candidate natural language sentence and the comparison target portion, whether the candidate natural language sentence is similar to the document;
• excluding a candidate natural language sentence determined to be similar;
• determining, with reference to a result of comparison between a number of negatives contained in each remaining candidate natural language sentence and a number of negatives contained in the comparison target portion, whether the remaining candidate natural language sentence is consistent with contents of the document;
• excluding a remaining candidate natural language sentence determined to be inconsistent;
• and
• outputting, as the natural language sentence, a candidate natural language sentence remaining after the excluding.

A non-transitory storage medium storing therein a program for causing a computer to function as an information processing apparatus, the program causing the computer to:

• acquire a logical formula representing an inference result which is based on an inference rule for an observation event;
• generate one or more candidate natural language sentences from at least one text element with use of a language model generated on the basis of a corpus which does not contain a document used to generate the inference rule, the at least one text element being included in constituents of the logical formula and being text, wherein the one or more candidate natural language sentences is generated by the language model by calculating a continuous occurrence probability with which words contained in an extracted n-gram occur continuously, and the language model is generated based on a statistical result for a continuous occurrence probability;
• store the language model in a memory;
• identify, as a comparison target portion of the document used to generate the inference rule, a portion indicated by ground information associated with a rule applied to obtain the inference result;
• determine, with reference to a result of comparison between each candidate natural language sentence and the comparison target portion, whether the candidate natural language sentence is similar to the document;
• exclude a candidate natural language sentence determined to be similar;
• determine, with reference to a result of comparison between a number of negatives contained in each remaining candidate natural language sentence and a number of negatives contained in the comparison target portion, whether the remaining candidate natural language sentence is consistent with contents of the document;
• exclude a remaining candidate natural language sentence determined to be inconsistent;
• and
• output, as the natural language sentence, a candidate natural language sentence remaining after the excluding.