Context-aware sentence compression

A method comprising:

• operating at least one hardware processor to:
• receive, as input, one or more digital documents, a query statement, and a summary length constraint,
• automatically identify, for each digital document of said one or more digital documents, a sentence subset comprising multiple sentences, to produce one or more sentence subsets, respectively, wherein the automatic identification is based, at least in part, on said query statement, a modified version of said summary length constraint, and a first set of quality objectives,
• automatically generate, for each sentence subset of said one or more sentence subsets, a random forest representation, to produce one or more random forest representations, respectively,
• iteratively:
• (i) generate a plurality of candidate document summaries by automatically sampling, from said one or more random forest representations, a plurality of tokens to create each candidate document summary, based, at least in part, on weights assigned to each of said tokens,
• (ii) automatically assign a quality ranking to each candidate document summary, based, at least in part, on said first set of quality objectives and a second set of quality objectives, and
• (iii) automatically adjust said weights, based, at least in part, on said quality rankings,
• wherein the iterating is performed until an average value of said quality rankings does not improve for a specified number of consecutive iterations, and
• automatically output one of said candidate document summaries having a highest said quality ranking, as a compressed summary.

A computer program product comprising a non-transitory computer-readable storage medium having program instructions embodied therewith, the program instructions executable by at least one hardware processor to:

• receive, as input, one or more digital documents, a query statement, and a summary length constraint;
• automatically identify, for each digital document of said one or more digital documents, a sentence subset comprising multiple sentences, to produce one or more sentence subsets, respectively, wherein the automatic identification is based, at least in part, on said query statement, a modified version of said summary length constraint, and a first set of quality objectives;
• automatically generate, for each sentence subset of said one or more sentence subsets, a random forest representation, to produce one or more random forest representations, respectively,
• iteratively:
• (i) generate a plurality of candidate document summaries by automatically sampling, from said one or more random forest representations, a plurality of tokens to create each candidate document summary, based, at least in part, on weights assigned to each of said tokens,
• (ii) automatically assign a quality ranking to each candidate document summary, based, at least in part, on said first set of quality objectives and a second set of quality objectives, and
• (iii) automatically adjust said weights, based, at least in part, on said quality rankings,
• wherein the iterating is performed until an average value of said quality rankings does not improve for a specified number of consecutive iterations, and
• automatically output one of said candidate document summaries having a highest said quality ranking, as a compressed summary.