Text mining a dataset of electronic documents to discover terms of interest

A method of extracting features from electronic documents for database query processing, the method comprising:

• accessing, by processing circuitry, a memory storing a database composed of a plurality of electronic documents including free-form text;
• text mining, by the processing circuitry, the plurality of electronic documents including the free-form text for terms of interest, including receiving a set of seed nouns as input to an iterative process an iteration of which includes:
• searching the plurality of electronic documents for occurrences of multiword terms linguistically headed by seed nouns of the set of seed nouns, at least some of the multiword terms defining a training set of a machine learning algorithm;
• using the machine learning algorithm and the training set, including one or more linguistic features of words in the multiword terms in the training set, to identify occurrences of additional multiword terms in the plurality of electronic documents, the one or more linguistic features including at least one of a part of speech, lexicon or semantic feature, at least some of the additional multiword terms linguistically headed by nouns outside the set of seed nouns;
• storing the training set and at least some of the additional multiword terms in a collection of terms of interest;
• and
• adding the nouns outside the set of seed nouns to the set of seed nouns and thereby identifying a new set of seed nouns for input to a next iteration of the iterative process;
• unifying, by the processing circuitry, terms of interest in the collection of terms of interest to produce a collection of normalized terms of interest from the terms of interest having common or similar words or modifiers;
• and
• executing, by the processing circuitry, a database query for features of the plurality of electronic documents from the database using the collection of normalized terms of interest for data analytics performed thereon.

A computer-readable storage medium for extracting features from electronic documents for database query processing, the computer-readable storage medium being non-transitory and having computer-readable program code stored therein that in response to execution by processing circuitry, causes an apparatus to at least:

• access a memory storing a database composed of a plurality of electronic documents including free-form text;
• text mine the plurality of electronic documents, including the free-form text, for terms of interest, including the apparatus being caused to receive a set of seed nouns as input to an iterative process an iteration of which includes the apparatus being caused to at least:
• search the plurality of electronic documents for occurrences of multiword terms linguistically headed by seed nouns of the set of seed nouns, at least some of the multiword terms defining a training set of a machine learning algorithm;
• use the machine learning algorithm and the training set, including one or more linguistic features of words in the multiword terms in the training set, to identify occurrences of additional multiword terms in the plurality of electronic documents, the one or more linguistic features including at least one of a part of speech, lexicon or semantic feature, at least some of the additional multiword terms linguistically headed by nouns outside the set of seed nouns;
• store the training set and at least some of the additional multiword terms in a collection of terms of interest;
• and
• add the nouns outside the set of seed nouns to the set of seed nouns and thereby identify a new set of seed nouns for input to a next iteration of the iterative process;
• unify terms of interest in the collection of terms of interest to produce a collection of normalized terms of interest from the terms of interest having common or similar words or modifiers;
• and
• execute a database query for features of the plurality of electronic documents from the database using the collection of normalized terms of interest for data analytics performed thereon.