Methods, systems, and storage media for automatically identifying relevant chemical compounds in patent documents

A system configured for training and using a chemical entity recognition system to extract and determine one or more chemical compounds from a patent document and determine a relevance of the one or more chemical compounds to the patent document, the system comprising:

• one or more hardware processors;
• and
• a non-transitory, processor-readable storage medium comprising one or more programming instructions thereon that, when executed, cause the one or more hardware processors to:
• obtain a plurality of patent documents from one or more patent databases,
• normalize each patent document of the plurality of patent documents into a unified format to achieve a plurality of unified patent documents,
• generate a chemical patent corpus from the plurality of unified patent documents, the chemical patent corpus comprising one or more chemical entities extracted from the plurality of unified patent document, each of the one or more chemical entities comprising one or more relevancy annotations, the one or more relevancy annotations indicating a relevance to the patent document from which the chemical entity is extracted, wherein a relevant compound for the one or more relevancy annotations is indicated for each of a prophetic compound and a Markush class, an irrelevant compound for the one or more relevancy annotations is a compound mixture part, a mixture part class, a mixture class, a polymer, and a biomolecule, and a mono-component compound and a chemical class for the one or more relevancy annotations are each relevant to the patent when the chemical entity is present in the title or abstract section of the patent document, and the chemical entity is part of a reaction context, the chemical entity or a measured property of the chemical entity is in a claim section of the patent document,
• provide the chemical patent corpus to the chemical entity recognition system, wherein the chemical entity recognition system tags the one or more chemical entities in a corresponding normalized patent document of the plurality of unified patent documents and extracts one or more additional chemical entities from the plurality of unified patent documents,
• calculate, by the chemical entity recognition system, a confidence score for each of the one or more additional chemical entities extracted from an associated patent document,
• determine, by the chemical entity recognition system, based at least in part on the confidence score, a label for each of the one or more additional chemical entities, wherein
• each label indicates the one or more additional chemical entities is either relevant or irrelevant to the associated patent document, and
• assign, by the chemical entity recognition system, the confidence score and the label for each of the one or more additional chemical entities to the associated patent document,
• wherein the confidence score of each additional chemical entity extracted from the associated patent document is a linear combination of features for a compound representing the additional chemical entity, the features comprising a frequency of a compound in a patent document, an occurrence of a compound within predefined sections of a patent document, a length of a term, an occurrence of a compound within special characters, an occurrence of a single compound within a section of a patent document, a compound not containing solvents or laboratory chemicals, and a presence of a compound in one or more predefined groups representing a frequency of compounds in a large set of chemistry patent documents.

A non-transitory storage medium having executable instructions embodied thereon for causing a processing device to:

• obtain a plurality of patent documents from one or more patent databases;
• normalize each patent document of the plurality of patent documents into a unified format to achieve a plurality of unified patent documents;
• generate a chemical patent corpus from the plurality of unified patent documents, the chemical patent corpus comprising one or more chemical entities extracted from the plurality of unified patent document, each of the one or more chemical entities comprising one or more relevancy annotations, the one or more relevancy annotations indicating a relevance to the patent document from which the chemical entity is extracted, wherein a relevant compound for the one or more relevancy annotations is indicated for each of a prophetic compound and a Markush class, an irrelevant compound for the one or more relevancy annotations is a compound mixture part, a mixture part class, a mixture class, a polymer, and a biomolecule, and a mono-component compound and a chemical class for the one or more relevancy annotations are each relevant to the patent when the chemical entity is present in the title or abstract section of the patent document, and the chemical entity is part of a reaction context, the chemical entity or a measured property of the chemical entity is in a claim section of the patent document;
• provide the chemical patent corpus to the chemical entity recognition system, wherein the chemical entity recognition system tags the one or more chemical entities in a corresponding normalized patent document of the plurality of unified patent documents and extracts one or more additional chemical entities from the plurality of unified patent documents;
• calculate, by the chemical entity recognition system, a confidence score for each of the one or more additional chemical entities extracted from an associated patent document,
• determine, by the chemical entity recognition system, based at least in part on the confidence score, a label for each of the one or more additional chemical entities, wherein each label indicates the one or more additional chemical entities is either relevant or irrelevant to the associated patent document, and
• assign, by the chemical entity recognition system, the confidence score and the label for each of the one or more additional chemical entities to the associated patent document,
• wherein the confidence score of each additional chemical entity extracted from the associated patent document is a linear combination of features for a compound representing the additional chemical entity, the features comprising a frequency of a compound in a patent document, an occurrence of a compound within predefined sections of a patent document, a length of a term, an occurrence of a compound within special characters, an occurrence of a single compound within a section of a patent document, a compound not containing solvents or laboratory chemicals, and a presence of a compound in one or more predefined groups representing a frequency of compounds in a large set of chemistry patent documents.